This section is not intended to replace a radiation physics course. For that purpose, we refer the reader to an excellent book by Baltas et al.,⁴ which goes into much more detail or the handbook published by Mayles et al.⁵ We will, however, recall the basic concepts necessary to introduce modern brachytherapy radiation sources and dosimetric concepts. In particular, this should help the reader to relate known quantities such as kerma and dose to the AAPM Task Group 43 (TG43) formalism.⁶ Note that in this chapter, our interest is with photon-emitting sources.

Radioactivity and Interactions

Radioactivity defines the process by which unstable nuclei or isotopes undergo decay (with one or more transitions) that lead to stable nuclei species. The number of decays of a radioactive isotope within a finite period of time is given by:

Eq 1

Thus, after a period of time t, the remaining number of atoms, N_t, is given by:

Eq 2

where λ is the radioactive decay constant (probability to undergo a nuclear transformation per unit of time) and N₀ the initial number of atoms. The half-life, corresponds to the time necessary for one half of the initial number of atoms to decay or N = N₀/2. It also corresponds to half of the initial activity. Solving equation 1, one finds:

Eq 3

As for the decay constant, the half-life is specific to a given nucleus. The activity, A, is the number of disintegration per second:

Eq 4

Defining A₀ as λ N₀, one can also predict the activity at time t:

Eq 5

The international system (SI) unit for activity is the Becquerel (Bq) (one disintegration per second). The older unit, the Curie (Ci), corresponds to 3.7 × 10¹⁰ Bq. Note that historically, 1 Ci was defined as the activity of 1 g of radium. However, Marie Curie’s gram of radium was not exactly 1 g; thus the accepted value of 1 g of radium is 0.95 Ci. Finally, a related quantity of interest is the amount of activity per unit of mass (Bq/kg) or the specific activity:

Eq 6

where N_A is the Avogadro constant and M is molar mass. The right term of that equation applies to a pure isotope.

Let us now consider the fate of the products of the decay of a radioactive source. As the emitted photons travel in a given medium, interactions can occur. Among the possible results of an interaction are the scattering of the incident photons or partial loss of energy of the incident photons and complete absorption in the medium. The probability that an incident photon within a unit area will interact with an interaction site of the medium is given by the cross section σ (cm² or barn with 1b = 10⁻²⁴ cm²). The larger the cross section or the number of interaction sites, the more likely the interaction is to occur. Consider a source providing a fluence Φ (number of particle per unit area); the expected number of interaction per interaction site is given by:

Eq 7

Thus the lost of fluence dΦ or the variation dN in the number of incident particle N over a distance dx can be written as:

Eq 8

This is a well-known differential equation, the solution of which describes the exponential attenuation of the primary fluence; µ is the linear attenuation coefficient or the probability of interaction per unit of path length. For a given medium, the number of interaction sites per unit volume depends on the mass density ρ. To remove this dependency, the mass attenuation coefficient µ/ρ is used

Eq 9

in which A is the atomic mass number and N_A is the Avogadro constant. For photon sources, the relevant interactions are Compton, photoelectric, Rayleigh, and pair production, although for most modern brachytherapy sources, this last process is negligible. Each one has its own cross-section, which are combined to give the linear or mass attenuation coefficients (total cross section).

Dosimetry

For dosimetric purposes, we are of course interested by the energy deposited per unit of mass or dose (J/Kg or Gy). However, contrary to charged particles, photons must interact to impart energy to a medium. It is thus the secondary charged particles, electrons in the case of interest, that contribute to the dose. Furthermore, when a photon interacts, a (variable) fraction of its energy can be transferred to a secondary particle and spent in the medium and the remaining fraction is carried away from the interaction site by the scattered photon. The kinetic energy per unit of mass transferred to charged particles by uncharged particles is called kerma (K). The kerma can be related to the basic source characteristics such as photon fluence and energy. For a monoenergetic photon source, the kerma can be expressed as:

Eq 10

in which dE_tr is the average transferred energy, µ_tr/ρ is the mass energy transferred coefficient and ψ is the energy fluence (related to the photon fluence as previously defined). Note that the mass energy transferred coefficient is related to the mass attenuation coefficient through the fraction of the total energy (E) that is transferred per unit of path length.

Eq 11

Of course, the values of both coefficients depend on the energy of the interacting photon and the composition of the interacting medium. We can be even more specific and ask that the energy not only be transferred but also spent in collisions (as opposed to radiative loss by bremsstrahlung). In this case, the quantity of interest is the collision kerma (K^coll):

Eq 12

in which µ_en/ρ (or µ_ab/ρ) is the mass energy absorption coefficient and g is the mean fraction of energy not spent in collision with orbital electrons. In the context of charge particle equilibrium, the dose (D) is equal to the K^coll and thus can be related to the photon fluence and energy fluence.

Eq 13

Note that for materials naturally found in the human body, µ_tr/ρ is equal to µ_en/ρ to within 1% for photon energy up to 1 MeV. Thus, for most modern brachytherapy sources, radiative loss is negligible, K ≈ K^coll, and the last two expressions are basically equivalent. It is important to remember that the dose in a given medium can be related to the dose in a second medium, assuming no change in fluence and charged particle electronic equilibrium, by the following equation:

Eq 14

in which f_as is a factor correcting for attenuation and scattering when the air around the equilibrium mass is replaced by water (adopting the notion of Baltas et al.⁴). In a more compact form, this equation can be written as:

Eq 15

Average mass energy absorption coefficient can also be used based on the effective source energy. Note that the ratio of the mass energy absorption coefficient is sometime referred to as the factor (f_med).⁷

Source Specification and Air Kerma Strength

Source specification or strength has been expressed by various quantities: equivalent mass of radium, activity (in Ci), apparent activity for sealed source (i.e., activity equivalent to an unsealed source in terms of exposure rate), exposure rate constant, and so on. In modern dosimetry protocol, the AAPM recommends the use of air-kerma strength, S_k, to characterize the source strength. As its name indicates, it uses a measure of the collision kerma in air. Although it would be practical to express a source strength in terms of dose rate in water, this is very difficult for a low-energy brachytherapy source because of large dose gradients and extremely small temperature variation resulting from dose deposition. However, for these energies, exposure rate measurements in air are well understood. The exposure (or exposure rate) is related to the collision kerma as follows:

Eq 16

In which W_air/e is 33.97 J/C. The SI unit for exposure (exposure rate) is C Kg⁻¹ (C Kg⁻¹ s⁻¹). Here we have replaced the collision air kerma by the air kerma by neglecting the (1 − g) factor, as discussed before. Furthermore, exposure rate and air-kerma rate are more practical to describe the energy imparted by a radioactive source used because of the activity decay with time. Note that the air-kerma rate will vary as a function of distance to the source because of the inverse square law and attenuation in air. The latter is, however, negligible, leaving the geometric factor the dominant one. Therefore, the product of the air-kerma rate with the square of the distance should be roughly constant over a range of distances that are useful in clinical practice. The air-kerma strength, S_k, has been adopted first as a reference quantity for low-energy sources used in permanent seed implants (PSIs). It is defined as the product air-kerma rate in free space at the distance of measurement and the square of that distance along the perpendicular bisector

Eq 17

The distance r must be large enough compared with the extent of the source so that it can be treated as a point source. The indices δ in the air-kerma rate was included in the revised TG43 publication ⁸ to remove very low-energy photons produced close to the surface of the capsule or source cladding. These do not contribute to the dose beyond the first millimeter. The use of δ is similar to that of the air-kerma rate constant, γ_δ, as defined by the International Commission on Radiation Units and Measurements (ICRU). Values of γ_δ for the most commonly used isotopes can be found here.⁹ The SI unit for the air-kerma strength is cGycm²h⁻¹ (µGym²h⁻¹). The symbol U is also commonly used in practice and publications, but it is not an SI unit. At a distance (r) of 1 cm, the air-kerma strength of a perfect point source will give a dose rate in cGy/h in air. The air-kerma strength can be related to the apparent activity of a source by the air-kerma rate constant (we are not advocating the use of apparent activity):

Eq 18

IN which γ_δ has a value of 1.27 cGy cm² mCi⁻¹ h⁻¹ for ¹²⁵I, 1.30 cGy cm² mCi⁻¹ h⁻¹ for ¹⁰³Pd, and 4.205 cGy cm² mCi−¹ h⁻¹ for ¹⁹²Ir.

From the derivations we have conducted previously, we know how to convert a dose in air to a dose in water, so combining all the related equations, the dose in water for an ideal point source is given by

Eq 19

Let us now consider the dose rate at a reference point. In TG43, this point is defined at r₀ = 1 cm and θ = 90°:

Eq 20

Λ is referred to as the dose rate constant. It is defined for a given medium, water in this case, at the reference point. Its units are cGy h⁻¹ U⁻¹. It can be obtained by measuring directly the dose at the reference point in water and either divided by the air-kerma strength or calculated by the Monte Carlo method. The dose rate constant depends on the isotope and the source design, including the encapsulation.

Finally, let us consider the case in which the effective treatment time is long compared with the source half-life (such as in PSIs). The source strength will decay with time according to:

Eq 21

For a PSI, the integration from t = 0 to infinity of equation 21 is a constant related to the source half-life, and the total deposited dose is given by:

Eq 22

Revised TG43 Formulation

Up to now, we have not addressed the particularity of the various seed designs and how it affects the dose at other points than the reference point for a “prefect” point source. The TG43 formalism is a modular analytical equation in which the effect source geometry, attenuation, scattering, and anisotropy in the medium of interest are represented by different terms and accepted values (measured or calculated) are tabulated. Because these parameters are obtained for each source individually, the formalism can be used for any encapsulated sources, within the limit of the approximations stated previously. The following equation gives the dose rate at any point (r, θ), in polar coordinates:

Eq 23

In which X is replaced by “L” for a line source and “P” for the point source approximation. Let us define the various parts of this equation. We refer the reader the AAPM task group report or to more physics-oriented textbooks for a complete derivation of these terms. Note that the product of g_X(r) · F(r, θ) plays the same role as f_as in our previous derivation. S_k and Λ were defined previously.

G(r, θ) represent the geometry function (i.e., the variation of the radiation field intensity with distance compared with the reference point [r₀, θ₀] already defined). For a perfect point source, there will be complete symmetry in θ and thus G(r) = 1/r², such that

Eq 24

Obviously, at the reference point, this ratio is 1 by definition for both one-dimensional (1-D) and 2-D formalisms. However, for a cylindrical source of finite active length, assuming an ideal line for which the activity is uniformly distributed along the length of the source length, one needs to integrate contribution of each infinitesimal element. The geometry factor for a line (2-D) source is expressed as:

Eq 25

L_s is the active source length and β corresponds to the angle defined by the point of interest (r, θ) and the two ends of the active portion of the source. L_s is specific to the source design and is generally smaller than the physical length of the source. The reader should be reminded that the term at θ = 0° is new to the revised TG43 formalism. It handles the obvious problem of a division by 0 in the second term. However, at this time, most commercially available software only use the second term.

The radial dose function or g_X(r) describes the dependence of the attenuation and scatter of the radiation in the water medium as a function of the distance r on the transverse axis (i.e. θ = 90°):

Eq 26

By definition, equation 26 results in g_X(r = r₀) = 1. Thus, the variation in the radial dose function is given relative to that reference point. The removal from this equation of the geometric dependence and any variation in θ also has to be taken into account, as we describe later in this chapter. The radial dose function g_X(r) is obtained as function of r from measurements or more commonly by the Monte Carlo calculation. The 1-D or 2-D values are extracted using the corresponding point source or line source geometric function. The values are tabulated and used in treatment planning.

Up to now, the variation of the attenuation and scatter has not been modulated according to the angular angle θ of the point of interest. In the TG43 formulation this is taken into account by anisotropy function F(r, θ). For a perfect point source, the values of the radial dose function could be used for all angles θ; in other words, it would be perfectly isotropic or F(r, θ) = 1 for all r and θ. However, because of the linear nature of clinical sources and variation in the composition of internal components and encapsulation (which can modify the energy spectrum), including the drive wire if applicable, the radial dose function must be modulated.

Eq 27

In this equation, the angular dependence dose rate at a given distance r is given relative to that at 90° (the definition for g_X[r]), corrected by the ratio of the geometric factors when going from θ₀ to θ. It should be obvious to the reader that the anisotropy function is always 1 for all r when θ = θ₀. In PSIs, it is not practical in the current state of technology to consider the individual orientation of each seed for the final dose calculation. First, the seeds are small and these orientations are difficult to extract from CT or MRI. Second, it would have to be done consistently for anywhere from 50 to more than 120 seeds depending on the prostate volume and seed activity used. In such cases, correction of dose rate of each seed from 2-D anisotropy function is not relevant. It is, however, possible to average the anisotropy at a given r over 4π (the complete sphere). This is the 1-D approximation of the anisotropy function, which becomes ψ(r).

To conclude this section, let us summarize by providing the possible revised TG43 equations:

Eq 28

Eq 29

Eq 30

Equations 28 and 29 are recommended. Note that the first 1-D formulation uses the line source geometric factors and radial dose function. This is the recommended 1-D formulation by the AAPM TG43 because it provides the greatest accuracy at small distances for sources that are not perfect point sources. It is, however, not systematically implemented in commercial planning systems. A source can be considered as a point source for distances more than three times the active length of the source (r > 3L). The second 1-D formulation uses the point source approximation for the geometric factor. This is the 1-D formulation used in most planning systems at this time. Historically, there was also a 1-D formulation that averaged the anisotropy factor not only over all angles, but also over all distances to obtain the anisotropy constant ψ_an.

Clinical Considerations

The TG43 formalism was originally intended for low-dose-rate (LDR), low-energy seeds used in PSIs. However, as we have shown previously, the dose rate equation has parameters that depend specifically on the source design. As such, it is now widely used for HDR and PDR brachytherapy.

In 1998, an Ad Hoc Subcommittee of the Radiation Therapy Committee of the AAPM published recommendations on the prerequisite for clinical use of new low-energy photon sources. These recommendations advocate that:

1 It should be the manufacturer’s responsibility to provide the end user with air-kerma strength calibrated sources. This calibration should be traceable to National Institute of Standards and Technology (NIST) primary air-kerma strength standard. Furthermore, it is the vendor’s responsibility to ensure periodic validation of its source strength by an independent (primary or accredited) laboratory.

2 The dosimetric characteristics of a new source should be determined by at least one, preferably two, independent experimental studies. In addition, a detailed Monte Carlo study of the source should be conducted. Both the experiment and Monte Carlo studies taken together should enable all relevant dosimetric parameters of the TG43 formalism to be extracted. These recommendations also state that these studies should be published in a peer-reviewed journal.

3 Radiological Physics Center (RPC) maintains a registry of brachytherapy sources that meets the AAPM dosimetric prerequisites.¹⁰ Relevant literature for each source, along with a description of the AAPM prerequisites, are also given on the registry website.

This systematic process has not been officially implemented for HDR and PDR sources at this time. As such, the source registry is limited to iodine and palladium sources currently. However, the practice of performing experimental measurements and Monte Carlo simulations on brachytherapy sources of any type can almost be considered a standard procedure in the scientific community.

Beyond these recommendations, it is the responsibility of the medical physicist to review periodically the scientific literature for published consensus or new dosimetric parameters for the sources used in clinics and to ensure proper entry of the parameters in the planning system. Finally, it is also the responsibility of the medical physicist to experimentally verify, using a calibrated well chamber (traceable to an accredited dosimetry calibration laboratory), the air-kerma strength against the vendor’s calibration. A recent report by the AAPM low-energy brachytherapy source calibration working group has made detailed recommendation in that regards.

Modern Brachytherapy Photon Sources

A number of isotopes have been used throughout the years for cancer treatments. Examples are given in Table 13-1. For radiation protection (safety) reasons, a number of nuclides are not used anymore, such as radium, which has a very long half-life and emits high-energy photons. Furthermore, high-energy sources have a large mean path length. Although this generates relatively uniform dose in the target volume, the penetrating photons are a disadvantage for the protection of OARs compared with the low-energy sources, such as iodine 125 (¹²⁵I) for example.

Table 13-1 Properties of Some Common Radionuclides Used in Brachytherapy

Thus, to produce sources that are easier to handle, to shield, and to subsequently dispose of, the industry has moved toward lower-energy and shorter-half-life isotopes. All the nuclides shown in the bottom portion of Table 13-1 are good examples of isotopes used in modern brachytherapy. For permanent prostate interstitial brachytherapy, the patient body itself constitutes enough shielding, in the energy range of 20 to 30 keV, that the patient is not a danger to his family (even though the “as-low-as-reasonably-achievable [ALARA] principle recommends not to hold small children directly on one’s lap for an extended period). The half-value layer of lead needed for ¹²⁵I photons is 0.025 mm. The latest incarnation of this trend is unequivocally the electronic sources. The latter are miniature x-ray generators, which produce no radiation when the current is off and are sufficiently low-energy to be used in a regular radiology suite.

In the following, the development of brachytherapy sources for PSIs and temporary implants is discussed further.

Permanent Seed Implants

In PSIs, multiple sources or seeds are used. The exact number and their positions within the treatment volume are the results of the planning process. It depends on the prescription dose, the seed strength, and the treatment volume. In this type of treatment, the seeds are left permanently, as the name describes, within the human body. Thus the radioisotopes are encapsulated in a biocompatible shell. Three radioisotopes are used clinically in PSI: ¹²⁵I, palladium 103 (¹⁰³Pd) and more recently cesium 131 (¹³¹Cs). According to the brachytherapy source register at the RPC,¹⁰ 16 seed models meet the general AAPM dosimetric prerequisites at this time. For most seed models, the shells or capsules are made of titanium (Ti), a well-known material in the medical field. All clinically used seeds are between 4.5 and 5 mm in physical length and 0.8 mm in diameter. The shell is usually laser-welded at the tips of the capsules. This can have an affect on the anisotropy function of the seed resulting from the extra attenuation of photon exiting through the welding. Encapsulation provides protection to body tissues or fluids, which are never in direct contact with the radioisotopes. Therefore, the seeds can be used in patients that have allergies to iodine, for example. Furthermore, the Ti melting point is 1941 K. Therefore, they are highly resistant to all types of sterilization procedures and to cremation as well, should a patient die within a few months after the implants are placed.

Recently, another type of shell made of biocompatible polymer was proposed. Polymer has the advantage of being composed of low Z elements (water-like material), attenuating the emitted photons. Consequently, less internal activity should be needed to produce a specific dose at the reference point. Another advantage is that interseed attenuation (ISA)—the attenuation of the photons from one seed passing through another seed—was also shown to be less.

The internal makeup of a brachytherapy seed varies between manufacturers. The radioisotopes are usually deposited in the form of silver halide on metal holders such as beads and rod, as an organic matrix or coated (ion exchanged or other processes) on resin or polymer beads.

A third important component is a high Z radiopaque marker for postinsertion radiologic identification. These can be in the form of rod, beads, and cylindrical bands. Silver, gold, gold-copper alloy, tungsten, and platinum-iridium alloy are the most widely used materials.

Three very different designs are portrayed in Fig. 13-1. The first (left) is a ¹²⁵I seed. The radioactive material is deposited on a silver rod (red) in a welded Ti capsule (black). This design is typical of the 6711 seed and of the SelectSeed. The total seed length is 4.5 mm for both models. The middle seed is model 2335 ¹⁰³Pd seed. The radioactive material is deposited on three polymer beads (red) on each side of a 1.2-mm long tungsten marker (yellow). A welded, double-wall Ti capsule is used (black) to seal the components. The total seed length is 5 mm (4.25 mm active length). Finally, the last design represented is the IBt OptiSeed 1031L ¹⁰³Pd seed. The innovative part of this seed is its biocompatible polymer shell made only of carbon and hydrogen (orange). A 2-mm long gold marker (yellow) is placed in the middle of the 5-mm long seed. On each side of the marker, a palladium tube (red) 0.7 mm in length emits the low-energy photons. The active length is 3.7 mm.

FIGURE 13-1 • A Monte Carlo study of seed design on the interseed attenuation in permanent prostate implants.

(From Afsharpour H, D’Amours M, Coté B, et al: A Monte Carlo study on the effect of seed design on the interseed attenuation in permanent prostate implants, Med Phys 35:3671, 2008.)

Let us stress that many more designs exist.¹¹ Because of the internal configuration of the marker and nuclides, and the use of different marker and encapsulation materials, it should be obvious to the reader that each seed will have specific photon emission patterns: intensity as a function, polar angles, and energy spectrum. Therefore, we emphasize once more, that TG43 dosimetric parameters are different for each seed.

Iodine 125

¹²⁵I used in an LDR source for brachytherapy is produced mainly in a neutron capture process, through xenon 124 (¹²⁴Xe) gas target: ¹²⁴Xe(n, γ)¹²⁵Xe. In turn, ¹²⁵Xe decays into ¹²⁵I by electron-capture (EC) transition: ¹²⁵Xe → ¹²⁵I + ν + E_b. The half-life of the ¹²⁵Xe is 16.9 hours. ¹²⁵I also decays by EC into an excited state ¹²⁵Te*, producing the maximum photon energy of 35.5 keV by gamma decay (6.7% of the time). In addition, the transition leads to characteristic x-rays of energy between 27.2 to 31.7 keV (K-shells) as a result of internal conversion (93.3%). Note that a very low-energy x-ray of 3.8 keV is also possible (15%), but in practice is attenuated within the source capsule. These low-energy x-rays are omitted in the following. The same is true for the various electrons emitted in the decay process; these are low-energy and are filtered out by the various components of the source.

The half-life and photon energies are given in Table 13-1. The specific activity of ¹²⁵I is more than 600 GBq/mg⁴ and iodine seeds are currently being used not only for PSIs (with air-kerma strength between 0.4 U to 0.8 U) but also eye plaque (up to 5 U).

Palladium 103

¹⁰³Pd can be produced in the neutron activation of ¹⁰²Pd: ¹⁰²Pd(n,γ)¹⁰³Pd and by nuclear reaction with a proton beam: rhodium 103 (¹⁰³Rh)(p,n)¹⁰³Pd. Because stable ¹⁰²Pd occurs only at 0.9% level, an enrichment process (and potentially a further separation process following the activation) is needed to obtain high specific activity of ¹⁰³Pd via neutron activation. The other avenue, p-n nuclear reactions, has the advantage of using ¹⁰³Rh (natural abundance 100%). In practice, this isotope can be produced with a very high specific activity, more than 2500 GBq/mg.

¹⁰³Pd decays by EC to excited states of ¹⁰³Rh, 99.9% to the first excited state. Multiple gamma decays are possible from the excited states (39.7 keV to 497 keV). However, the emission probabilities for the gamma are all less than 0.07%. Internal conversion is the preferred mode of de-excitation from the first excited state to the ground stage. The probable emission energies for this isotope are the resulting characteristic x-ray emission between 20 to 23 keV. Overall, the mean photon energy of a ¹⁰³Pd seed is lower than ¹²⁵I (21 versus 28 keV). However, its shorter half-life results in higher initial dose rate compared with ¹²⁵I. ¹⁰³Pd has been promoted to be more biologically effective against rapidly proliferating tumors; however, limited support can be found for this notion. Up to now, clinical results do not show significant difference in survival rate.

Cesium 131

Two modes of production have been used to obtain ¹³¹Cs radioisotopes. Both first produced the parent nuclide barium 131 (¹³¹Ba), which decays into ¹³¹Cs by EC.

1 Neutron activation: ¹³⁰Ba(n,γ)¹³¹Ba

2 Nuclear reaction: ¹³³Cs(p,3n)¹³¹Ba ¹²

The natural abundance of ¹³⁰Ba is small (0.11%) and ¹³²Ba (0.10%) will lead to ¹³³Ba, a long-lived radioisotope (10.52 years) that decays by EC to ¹³³Cs with a high proportion of gamma rays between 53 and 384 keV. Note that ¹³⁴Ba, ¹³⁵Ba, ¹³⁶Ba, and ¹³⁷Ba are all stable isotopes. Thus, ¹³⁰Ba target enrichment is necessary. On the other hand, ¹³³Cs is a stable isotope with 100% natural abundance. Therefore, the nuclear reaction (p,3n) appears to be an interesting alternative. However, ¹³³Cs(p,n)¹³³Ba reaction is also possible. However, the production cross-section for (p,3n) reaction can be significantly higher than for (p,n) by choosing the appropriate incident proton beam energy.

Independent of the production mode, ¹³¹Ba decays into the daughter product ¹³¹Cs by EC ( = 11.50 days). ¹³¹Cs also proceeds by EC in ¹³¹Xe with a half-life of 9.69 days. This last process results in low-energy x-rays between 29.46 keV and 34.42 keV. There is no gamma emission.

Note that this isotope possesses the shortest half-life of the three isotopes available for PSI. Thus the initial dose rate for ¹³¹Cs is higher than that of ¹⁰³Pd and ¹²⁵I.¹³^,¹⁴ It is thus more susceptible to postimplant edema. Finally, the controversies remain in comparing the dose to both the prostate and the OARs with different radionuclides, in particular when the biologic effective dose is considered.¹⁵

High-Dose-Rate and Pulsed-Dose-Rate Brachytherapy

HDR brachytherapy treatments deliver doses of more than 12 Gy/hr. As such, HDR brachytherapy treatment usually lasts a few minutes. For its obvious advantages, our interest is toward afterloading delivery. Afterloader units are composed of an encapsulated radioactive source (Ti or stainless steal) at the tip of a stainless steel drive wire. While not in use, the source sits into a shielded area of the unit. The unit is controlled from outside of a shielded treatment room, decreasing direct manipulation and increasing safety. For treatment, the source travels through various types of applicators or inside needles and catheters. The source can stop at various positions, called dwell positions, for a given amount of time, called dwell time. The determination of the individual dwell position and dwell time is part of the planning process. Physicians have used modulations of these degrees of freedom for decades before the advent of external-beam, intensity-modulated radiation therapy. However, the full extent can only be explored with the recently available inverse planning algorithms.

A special case of HDR is PDR brachytherapy. In PDR, the treatment time is limited to brief irradiation for a given period to mimic LDR brachytherapy, in the range of 1 to 3 Gy/hr. The same type of equipment is used.

The most widely used nuclide for HDR or PDR brachytherapy is iridium 192 (¹⁹²Ir). Compared with cobalt 60 (⁶⁰Co) or ¹³⁷Cs, ¹⁹²Ir’s lower average photon energy makes it easier to shield. This offsets the disadvantage of a shorter half-life. Furthermore, the high specific activity of ¹⁹²Ir (341 GBq/mg) compared with ¹³⁷Cs allows for its use for both LDR and HDR brachytherapy with relatively small sources.

Iridium 192

¹⁹²Ir is produced in the nuclear reactor in the reaction ¹⁹¹Ir(n,γ)¹⁹²Ir. ¹⁹¹Ir composes 37.3% of natural iridium, ¹⁹³Ir making 62.7%. The material can be enriched so as to reduce the contamination of radioactive ¹⁹⁴Ir from ¹⁹³Ir and increase ¹⁹²Ir production. Interestingly, ¹⁹²Ir has a high neutron capture cross section, allowing double-neutron capture by ¹⁹¹Ir into ¹⁹³Ir. Hence, the production rate of ¹⁹²Ir is limited by its own decay and by its transformation into ¹⁹³Ir.

¹⁹²Ir decays into platinum 192 (¹⁹²Pt) via β⁻ decay 95.1% of the time and the remaining 4.9% in osmium 192 (¹⁹²Os) by EC. This leads to a complex decay pattern resulting in 29 gamma emission peaks from 0.110 to 1.378 MeV, various characteristic x-rays, and numerous electrons up to 1.377 MeV. For the β⁻ decay channels, de-excitation of ¹⁹²Pt* to ¹⁹²Pt (ground state) proceeds mainly via gamma emission and in a smaller fraction by internal conversion. Gamma rays of 0.296, 0.309, 0.317, and 0.468 MeV are the most probable (29% to 83%). This decay chain leads to 2.2 photons per β⁻ decay. The second, less probable, branch leading to ¹⁹²Os* by EC also decays mainly via gamma emission to the ground state. The most probable energies are the 0.206 and 0.485 MeV, but they make up only a small percentage.

The active ¹⁹²Ir core is usually made of an Ir-Pt alloy (10%-30% Ir and 70%-90% Pt) to make it more mechanically robust. The core itself is approximately 3 to 4 mm long and 0.3 to 0.4 mm in radius.

Ytterbium 169

Battista et al.¹⁶ have proposed ytterbium 169 (¹⁶⁹Yb) as an alternative radionuclide for brachytherapy. Its specific activity is approximately 2.5 times higher than ¹⁹²Ir, making it a good candidate for production of a small source of high strength. With an average energy of 93 keV, ¹⁶⁹Yb would lower the shielding requirements of an HDR brachytherapy room. This nuclide is produced in a reactor from the reaction ¹⁶⁸Yb(n,γ)¹⁶⁹Yb. Because ¹⁶⁹Yb composes only 0.13% of natural Yb element, an enrichment process must be performed. ¹⁶⁹Yb decays by EC in excited states thulium 169 (¹⁶⁹Tm). Return to the ground state proceeds via gamma emissions (45%) or internal conversion (55%). Seventy-four γ-ray transitions (from 63 keV to 782 keV), 13 characteristic x-rays, and more than 100 different electron energies (the most probable all being less than 130 keV) are emitted. The most common γ-rays are 63.12 keV (44%), 109.8 keV (17.5%), 130.5 keV (11.3%), 177.2 keV (22.2%), 198 keV (35.8%), and 307.7 keV (10%). A potential downside of this isotope from a clinical perspective is that its half-life is less then 50% of that of ¹⁹²Ir, and it therefore needs more frequent source change.

Thulium 170

Thulium 170 (¹⁷⁰Tm) is an interesting radioisotope, first because of its half-life of 128.6 days seems practical from a clinical perspective. Second, the average x-ray energy of 66 keV, lower than ¹⁶⁹Yb and ¹⁹²Ir, would allow for easier shielding and greater protection for the OARs. Finally, its specific activity is only 30% lower than ¹⁹²Ir. This enables the making of compact sources with high strength.

This radioisotope is made by neutron activation through the reaction ¹⁶⁹Tm(n,γ)¹⁷⁰Tm. The abundance of stable ¹⁶⁹Tm is 100%. ¹⁶⁹Tm is the rarest of the rare-earth elements (although found in quantities similar to silver) composing less than 0.01% of the ores from which it is extracted.¹⁷ Thus obtaining high purity materials is expensive. The decay of ¹⁷⁰Tm proceeds mainly (99.87%) via β⁻ decay. It can decay directly to the ground state (81.6% of the time) with a maximum beta energy of 968 keV (average energy of 323 keV). 18.3% of time the β⁻ decay will go to the first excited state and the associated beta energy will be lower (maximum of 883 keV, average of 290.5 keV). Final decay to the ground state can proceed directly by gamma emission of 84.25 keV photons, 2.48% of the time or via internal conversion leading to a series of six x-rays and six low-energy electrons. Apart from the 84.25 keV, the total x-ray emission probability is small at 6.25% for energies ranging from 7.42 keV (2.93%) to 60.96 keV (0.12%).

The other decay branch is EC to erbium 170 (¹⁷⁰Er). The probability of this decay is low at 0.13%. The x-rays (seven rays between 6.95 and 78.7 keV) and low-energy electron (six from 5.5 to 78.3 keV) spectra are a factor of 10 to 100 less intense than those from the main decay branch.

However, it should be noted that because of the low gamma and x-ray emission intensities, the air-kerma rate constant for ¹⁷⁰Tm is approximately 200 times lower than ¹⁹²Ir.⁹ Thus, a thulium source might very well be limited to PDR applications. Moreover, the effect of the high-energy electron emitted in beta decay probably needs to be taken into account both in terms of bremsstrahlung emission¹⁸ and also in terms of the possibility of insufficient encapsulation thickness to stop all betas. In such a case, an increased dose could be expected in the first few mm outside of the source capsule.

Electronic Sources

The latest in the field of brachytherapy are electronic sources.¹⁹ These are composed of a miniature x-ray source, which emits x-ray only when the tube current is on. Xoft has commercialized the first electronic sources, Axxent, and these are Food and Drug Administration–approved for specific treatments. It is a low-energy source of 50 kV. As such, the dose fall-off is greater with electronic sources than ¹⁹²Ir. Because the energy is lower than current radiology type devices, the body itself constitutes sufficient shielding. This allows for treatment personnel to be present in the room with the patient during treatment. So in principle no special (shielded) treatment room is needed. Finally, there is no radioactive waste handling.

As with any x-ray device, an electronic brachytherapy source produces a continuous energy spectrum. For the Axxent model, that spectrum has a mean energy at or lower than that of ¹⁰³Pd,²⁰ but a dose rate high enough for HDR treatments. The x-ray tube is 2.25 mm in diameter and the complete assembly is 5.4 mm in diameter. The tube is water-cooled and disposable after usage.

Three-Dimensional Image Acquisition for Planning

3-D imaging modalities, mainly US and CT but also MRI for prostate ²¹^,²² or gynecologic imaging²³^,²⁴ and functional imaging such as magnetic resonance spectroscopy (MRS)²⁵ or PET, are rapidly being adopted as the standard for dose planning in brachytherapy. Similarly to external-beam radiation therapy, the anatomic structures of interest are contoured and used to define planning target volumes (PTVs) and OARs. Here PTVs and clinical target volumes (CTVs) are often considered the same in brachytherapy. A discussion to differentiate them is beyond the scope of this chapter. Clearly, the availability of 3-D imaging improves tumor targeting and OAR avoidance, and allows for a more precise reporting of dose delivered and dose-response curves.

The physics underlying the imaging mechanism vary widely between the different imaging modalities. Obvious examples are US, CT, and MRS images of the prostate, each showing a completely different aspect of the gland. They complement each other, providing a new insight at each stages of brachytherapy: diagnosis, planning, delivery, and follow-up. The benefit of 3-D imaging is clear and has been discussed in the introduction to this chapter. Proper use of these modalities is progressing rapidly, in particular because of the availability of image registration (rigid and non-rigid) and fusion tools.

The limitations and uncertainties introduced using 3-D imaging (CT and MRI or MRS) must be understood and taken into account. A few examples affecting precision and uncertainty are presented in the following section. However, one must keep in mind that the additional knowledge from 3-D imaging for target definition, OAR identification, catheter and applicator reconstruction, dose planning, quality assurance, and safety carries significantly more weight in helping to achieve the overall treatment quality.

Precision and Uncertainty

In CT-based HDR brachytherapy, the uncertainty in the localization of the longitudinal catheter-tip positions caused by the discrete slice thickness results in an uncertain dose being delivered. Intuitively, there is a trade-off between the treatment planning time and the number of CT slices employed in HDR brachytherapy treatment planning. If we make use of the fine-thickness CT scan to minimize the dose error induced by uncertainty of CT reconstructions, the treatment planning needs additional time and effort from an oncologist who contours the target volume of the tumor, as well as a dosimetrist or medical physicist who defines the catheter position for the increased number of CT slices resulting from thinner slices of CT scan. In a recent study,²⁵ different scenarios were considered to mimic the longitudinal imprecision of the catheter tips; the dwell positions, either with all catheters displaced forward, backward, or, more realistically, randomly distributed within the space defined by one CT slice thickness, were for thicknesses ranging from 2 mm to 5 mm. It was found that the dose uncertainties caused by the finite slice thickness increase linearly with the slice spacing, from 3% to 8% for the slice thickness values ranging from 2 to 5 mm, respectively. The more realistic scenario yields average errors ranging from 0.7% to 1.7%. The apex and the base show larger dose errors and variability of dose errors than the median of the prostate. No statistical difference was observed among different transversal sections of the prostate. A CT slice thickness of 3 mm appears to be a good compromise, showing an acceptable average dose uncertainty of 1%, without unduly increasing the number of slices.

The proper dose delivery of fractionated HDR prostate brachytherapy relies on the stability of the implant and on reproducible catheter positions for each fraction. This is particularly critical with the current trend to reduce the number of fractions and increase the dose per fraction. Typically, doses ranging from 9.5 to 10.5 Gy per fraction are used.

The tumor volume may change because of the insertion of catheters or the delivery of radiation. The dosimetric errors resulting from these volume changes are not related to the introduction of 3-D image planning, but the anatomic information provided by 3-D imaging exposes the extent of the problem, and may offer solutions for improvements. For instance, for permanent prostate implants, the needle insertion itself is known to induce a trauma that changes the prostate volume.²⁶^–²⁸ These changes occur rapidly, within hours or a few days after implantation and can take several weeks to resolve.²⁹ This effect of edema on dose coverage has been investigated by many authors.^{26–28,30–32} For HDR, although the number of catheters inserted is usually significantly less than the number of needles required for PSI, one may expect a similar trauma from the catheter insertion. However, because the dose is entirely delivered in a few fractions within the first 2 days after catheter insertion, the significance of the edema is different. One can ask whether the dose plan used for the first fraction is adequate for the second fraction, usually delivered on the next day. A study³³ of the change of prostate volume between two HDR fractions showed either an expansion or a contraction of up to 17% of the gland. This translated into small-implant DVH variations of less than 5% for the prescription dose. The global volume changes observed in this study for patients treated with HDR brachytherapy were less significant than volume changes usually observed after permanent prostate implants.

The factors discussed previously translate into a discrepancy between the source dwell positions observable on planning CT or MRI and the actual dwell positions during dose delivery by the afterloader. In addition to the dose uncertainty attributable to the intrinsic characteristics of the finite CT slice thickness and the dosimetric affect of prostate volume change resulting from the trauma caused by the insertion of catheters and from resolution of edema between fractions,³⁴ craniocaudal catheter displacement attributable to patient and prostate motion between fractions can also affect the dose distribution.

A study performed by Kim et al.³⁵ to evaluate the Hsu free-hand technique developed at the University of California–San Francisco (UCSF) showed an average catheter displacement of 4.1 mm between the first and second fractions (on average, 19.5 hours difference). This is small compared with the average displacement noted in several reports.³⁶^–³⁹ Using fluoroscopy, Martinez et al.³⁶ measured a mean displacement of 20 mm between the first and second fractions (at least 6 hours difference, and 36 hours between the first and the fourth fractions). They reported that needle movement decreased between the subsequent fractions to an average of 4 mm (between the third and fourth fractions). Using measurements of catheter tips by plain films taken before treatment, Damore et al.³⁷ reported a mean displacement of 7.6 mm and a maximum displacement measurement by the bony marker method of 28.5 mm between the first and second fractions (40 hours for a total of four HDR fractions). They also reported a reduction in needle movement after the first day (an average of 2 mm between the third and fourth fractions). Using a 5-mm CT scan, Hoskin et al.³⁸ reported that the average template movement was 1 mm and that catheter movement relative to the prostate was 9.7 mm between the first and second fractions (over a period of 18-24 hours). Mullokandov and Gejerman³⁹ reported that no displacement of catheters relative to the template occurred and that the mean consecutive catheter displacement was 2, 8, and 10 mm before the second, third, and fourth fractions. Because the time interval between fractions was 6 hours in their study, the measured displacement (10 mm) before the fourth fraction (minimum 18 hours difference) can be compared with our measurement (4.1 mm) before the second fraction.

These four HDR fraction studies in the literature ³⁶^–³⁹ showed time-dependent catheter displacement between fractions. The maximum catheter displacement occurred up to approximately 12 hours after the first fraction (20 mm before the second fraction),¹³ 7.6 mm before the second fraction,²² and 6 mm before the third fraction.²⁴ The dosimetric effect and how to compensate for these catheter displacements must be understood.

Uncertainties in Postimplant Dose Assessment

Prostate Imaging

Imaging is the cornerstone of postimplant dose assessment in PSIs. On this subject, the literature is overwhelmingly associated with PSIs. Although most procedures are performed under US guidance for convenience, some centers rely on slower but more precise MRIs. The superiority of one or the other in terms of survival and toxicity outcomes has not been shown in the literature. However for postimplant dose assessment, the imaging modality does make a difference in terms of quantification of dose distributions, especially to the prostate volume.

CT is currently a widely used imaging modality to accomplish this task. The problem in using CT-based images for prostate brachytherapy after plan evaluation is twofold. The contrast between the prostate boundary and the surrounding tissues is not always clear; it is very difficult to assess the prostate base and apex on CT.⁴⁰^,⁴¹ As such, a prostate volume contoured on CT images tends to result in larger volumes than on MRI.⁴²^,⁴³ Furthermore, there are large inter- and intraobserver variabilities⁴⁴ with CT imaging. These variability are smaller on 3-D US imaging and MRI than CT.⁴⁵ The dose distribution, and the derived dose metrics used to evaluate the implant quality, is related to delineated organ volume. Thus, the literature has shown that CT-based volume definition can lead to uncertainties on the order of 10%⁴⁶^,⁴⁷ for a clinical metric such as D90.⁴⁸^,⁴⁹

Factors other than volume definition on a particular imaging modality also contribute to decrease the target dose coverage or increase OAR doses. Of importance, we should mention prostate edema, seed misplacements, seed migrations, and seed position detection accuracy. Although we present them independently, in clinical practice it is the combined effect that is ultimately measured by the postimplant dosimetry.

Prostate Edema

Prostate edema is characterized by a swelling of the prostate associated with the trauma from needle insertions. Because the dose rate at a given point within the volume of interest (VOI) obeys the inverse square law, a larger target volume will effectively receive less dose. Moreover, in prostate seed implant the central region is generally planned “colder” to spare the urethra. Thus this central portion can receive a lower dose than expected. In general, the overall effect on the clinical dose parameters depends on two factors: the magnitude of the swelling and its resorption with time.²⁸^,³⁰

Various studies have shown that the volumes after implantation are larger than that at 1 month after implant, up to a factor of 2.²⁸^,⁵⁰ When looking at serial MRI, Crook et al.⁴³ showed that the prostate volume is approximately 10% higher at day 30 postimplant relative to baseline. Furthermore, a sizable fraction of the patients have much larger residual edema at their postimplant volume study. From their serial CT study, Waterman et al.²⁸ have extracted an edema half-life of approximately10 days (time for a 50% reduction in volume from its maximum). On the other hand, Dogan et al.⁵¹ and Leclerc et al.⁵² have obtained average edema half-life values in the range of 20 to 30 days. Thus, on the one hand, Badiozamani et al.⁵³ have concluded that the magnitude of edema cannot be predicted using a single parameter such as age, prostate volume, number of needles, number of seeds, and so on. On the other hand, the results of Crook et al.⁴³ also point toward patient-dependent edema resorption or half-life. The affect of edema on the final dosimetry can therefore be expected to vary greatly from patient to patient.⁵²^,⁵⁴

Another effect to consider for dosimetric purposes in permanent implants is the obvious relationship between the edema half-life and the isotope decay half-life when the magnitude of the edema is important. A shorter half-life isotope, such as ¹⁰³Pd will be more affected than an isotope like ¹²⁵I that has a longer half-life.²⁶^,³² For a very short half-life isotope, ¹³¹Cs, three recent studies have indicated the critical need to carefully evaluate the necessary dose level at the time of planning because the edema half-life is similar or greater than the isotope half-life.⁵⁰^–⁵² For loose seed ¹²⁵I implants with no seeds implanted outside the prostate capsule, Villeneuve et al.⁵⁵ have used an inverse planning software to come up with a set of dose objectives that was shown to be robust to a wide range of edema scenario, namely prostate V100 of 99% (with a margin of 3-4 mm), V150 at approximately 65%, and V200 of 25% to 30%, while keeping the urethra V150 at 0% and D5 below 200 to 220 Gy.⁵⁴

Seed Misplacements

Seed misplacement is usually reserved to the phenomena by which the final position of the seeds within the prostate is different than was originally planned. It is therefore a problem associated with the delivery process. This difference in position can be characterized by displacement in each axis. However, the process is slightly more complex. Taschereau et al.⁵⁶ have categorized seed misplacements according to the notion of needle-related misplacements and other factors. Needle-related misplacements account for any change to a group of seeds (inside a needle) because the needle path during insertion is not perfectly parallel to the insertion axis.⁵⁷ Any angulations result in a number of seeds misplaced relative to planning. Suction during needle retraction can also result in clustering of seeds and displacement of seeds along the needle track. Prostate motion and deformation during needle insertion and retraction also contributes to seed misplacements. Finally, an individual seed can move about its deposition position. Postimplant needle reconstructions results in Gaussian distribution of relative distances compared with planned seed positions, having a 1σ of approximately 2.2 mm,⁵⁸ with needle-related misplacements being the most important factor.⁵⁶ Simulation of seed misplacements have been used in a number of studies to look into the dosimetric consequences of the choice of a particular isotope and seed activity⁵⁹^,⁶⁰ or the robustness of various implantation techniques.⁵⁶^,⁶¹

Seed Migration

Seed migration can be differentiated from seed misplacement in the general sense that a seed moves at a distance from the target volume, mostly lung, such that it does not contribute to the target dose anymore. Seed migration was documented as early as 1991 by Steinfeld et al.⁶² The incidence varies from 5.9% to 36.2% of the patients, for typically less then 0.5% of the seeds implanted. Some specific case reports have documented deleterious effects.⁶³^–⁶⁵ In theory, the loss of a seed constitutes a degradation of the dosimetric plan and could lead to a cold spot. In practice, it is the seeds in the periphery of the prostate that are the most likely to be affected. A significant margin of 3 to 4 mm is usually recommended for the 100% isodose line in this area. As such, the clinical consequence of a single seed loss is unclear.⁵⁹

According to the American Brachytherapy Society (ABS) and the International Commission on Radiological Protection, each patient should be evaluated for seed migration at the time of the first follow-up visit, generally corresponding to the postimplant imaging study.⁶⁶^,⁶⁷ A standard chest x-ray is recommended by the ABS⁶⁶; however, a new detection method not involving ionizing radiation and using an external scintillation detector has also been proposed⁶⁸ and clinically validated.^68a

Seed Detection Accuracy

The precise detection of seed positions is crucial for accurate postimplant dose calculations. For example it is known that even with a state-of-the-art algorithm, seed detection on US images is a difficult task.⁶⁹ On the other hand, there is a large body of literature on modern multiprojection x-ray imaging⁷⁰^–⁷³ or tomosynthesis and cone-beam imaging⁷⁴^,⁷⁵ for accurate seed detection. These modalities provide an accurate count, position, and even angle of each seed. However, these latter techniques do not offer the same organ-definition capability; if at all possible, they should be compared with CT or MRI. As such, seed positions are often extracted on CT. Most commercial treatment planning system (TPS) use simple level and thresholding algorithms followed by component breaking. This step usually involves knowing precisely the number of postimplant seeds in the area of interest. Some advanced algorithms have been proposed for CT images.⁷⁶^,⁷⁷ A study by De Brabandere et al.⁷⁸ has looked at the accuracy of seed position reconstruction based on a phantom study for both CT and MRI. They concluded that the accuracy increases as the slice thickness decreases (increasing sampling and decreasing partial volume effect) and that it is usually better for CT (<1 mm) than MRI (1.5 to 2 mm) for slice thickness less than 5 mm. The effect on postimplant dose distribution is usually small (e.g., less than 5% on D90) when the reconstruction errors are smaller than 2 mm.⁷⁹^,⁸⁰

On the other hand, the detection of small objects on MRI can be complicated by the presence of distortions and artifacts.⁵⁰ For clinical cases, Bloch et al.⁸¹ have explored contrast-enhanced MRI and T1- and T2-weighted MRI acquisition for seed detection. They found that MRI was missing from 12% to 29% of the seeds depending on the MRI sequence. To provide an accurate evaluation of the various dose parameters in PSIs, at least 95% of the seeds should be detected.⁸² To overcome the deficiencies of both organ definition on CT and seed reconstruction on MRI, a combination of x-ray and MRI scans has been proposed for postimplant dose dosimetry.⁸³ However, Beaulieu and Verhaegen have pointed out that neglecting the density information from CT images might eventually lead to dose calculation errors of 10% for low-energy photon sources.⁸⁴

Optimal Seed Strength in Prostate Permanent Implants

A systematic planning study was conducted by Beaulieu et al.⁵⁹ using different source models and a wide range of source strengths. In that work, unbiased plans were generated using an inverse planning algorithm⁸⁵; clinically acceptable plans were generated based solely on dose objectives: dose coverage to the target, margins, and protection to the urethra. Moreover, seed misplacement simulations were added to create postplans.⁸⁶ As shown in Pouliot et al.,⁸⁵ the inverse planning algorithm naturally resulted in robust implants for source strength ranging from 0.38 U (0.3 mCi) to 0.89 U (0.7 mCi), with a small advantage for activity in the range of 0.7 U (0.55 mCi) with more total strength implanted with increasing source strength for the same planning objectives. It also resulted in better “simulated” postimplant D90.⁸⁵ This is directly in line with the clinical randomized trial conducted by Narayana et al.,⁸⁷ which demonstrated a dosimetric advantage of the higher strength at the time of postimplant evaluation. The results of Beaulieu et al.⁵⁹ and Narayana et al.⁸⁷ should also be looked at in the context of papers by Sloboda et al.⁸⁸^,⁸⁹ Within a completely different inverse planning software, these authors have also shown that increasing source strength leads to implants with a higher total implanted strength,⁸⁸ and resulted in better clinical postimplant dose coverage.⁸⁹

Based on the literature and clinical results (low and high source strength), it should be emphasized that there is no such thing as a single optimal source strength, but rather a range of clinically usable source strengths, as defined in Beaulieu et al.⁵⁹

Three-Dimensional Dose Distribution

Standard Volumetric Dose Indices

We have shown in a previous section how important 3-D imaging is for brachytherapy. Similarly, reporting dose to a few single points, often defined geometrically only, is no longer appropriate in brachytherapy. Dose distributions can now be characterized using a variety of standard dosimetric criteria. With a 3-D planning system, the dose distribution of any contoured organ can be described by its DVH. Although it is always desirable to globally evaluate the DVH, it is also useful to derive a few volumetric dose indices from the DVH. The ABS has published general guidelines for different anatomical sites ⁹⁰ that consist of a set of dose-limit specifications. For prostate cancer, Stock et al.⁴⁸ have demonstrated that the dose delivered to 90% of the gland (D90) had a critical affect on the subsequent risk of prostate-specific antigen–diagnosed recurrence. For prostate PSIs (PPIs), we will build on these guidelines and those used clinically at UCSF⁹¹ that have yielded reproducible results: mature 5-year biochemical control rates of 96% (median follow-up of 63 months) reported for 118 consecutive patients.⁹²

For HDR prostate brachytherapy, a multicenter study, Radiation Therapy Oncology Group (RTOG) 0321 protocol,⁹³ which was completed in 2007, was tasked with developing a quality assurance process for HDR prostate brachytherapy. The RTOG 0321 recommendations are (1) deliver the prescription dose to at least 90% of the primary target volume (V_Prostate100 > 90%), (2) restrict the volume of bladder and rectum tissue receiving 75% of the prescription dose to less than 1 cc (V_Rectum75 and V_Bladder75 < 1 cc), and (3) restrict the volume of urethra receiving 125% of the prescription dose to less than 1 cc (V_Urethra125 < 1 cc). At UCSF, we also strive for the prostate homogeneity index (HI) to be greater than 0.60 (HI = [V100 − V150]/V100 > 0.60; this can also be expressed as V_Prostate150 < 40% of the prescription dose). For PPI, the clinical criteria are V_Prostate100 > 90% and V_Urethra120 < 1 cc. The dose to healthy organs—including the urethra, bladder, rectum, and penile bulb—and the boost achievable to areas of higher cancer density are also computed.

Similarly, for gynecologic cancers treated with HDR, Nag et al.⁹⁴ have proposed guidelines for 3-D planning and reporting of intravascular brachytherapy. Haie-Meder et al.²³ and Sahgal and Roach⁹² have shown that in addition to traditional point dose and volume parameters, DVH analysis enables further possibilities for prescribing and reporting. In their case, the DVH constraints for OARs allowed reproducible treatment plans, helping to detect and avoid severe overdosage not necessarily detectable with the superseded point A.

Inverse Planning Dose Optimization

The inverse planning approach always benefits from additional anatomic information without increasing the complexity of the planning process. Another advantage of the inverse planning is that the burden on the physicist does not increase with the number of targets and OARs. This facilitates and therefore encourages the inclusion in the planning process of more protection to OARs that were neglected before. Examples of additional OARs are the penile bulb and the neurovascular bundles in the treatment of prostate cancer.

The inverse planning approach can be described as a method of radiation treatment planning in which one starts with the desired dose distribution, or clinical objectives, and then determines the treatment parameters that will achieve it. This is opposed to the conventional forward planning approach in which the treatment parameters are first chosen and then the resulting dose distribution is calculated and evaluated. Because inverse treatment planning begins with the description of the desired dose distribution, it represents a change of paradigm in the planning process. CT or MRI contours of the CTVs and OARs are used not only to define the anatomy for visual assessment and DVH calculation, but also to optimize the dose distribution. Therefore, they provide the physician with added flexibility and control to shape the dose distribution. In inverse planning, the anatomic features together with the dose objectives constitute the starting point of the dose optimization process. This requires that the multiple targets and OARs are known, and that the admissible and required dose coverage is specified for each organ. The computer knows the possible source positions because they need to be previously digitized, but from the user’s perspective, the emphasis is on the anatomy and the dose constraints. At the planning stage, the dwell times (for afterloader) or source positions (for permanent implants) become irrelevant to the user. This is different from forward planning, in which the dose distribution is iteratively adjusted by modifying the dwell times or the source positions until an acceptable dose distribution is produced. The main benefit of the inverse planning approach is that all dosimetric requirements (dose coverage, dose homogeneity, extent of hot spots, OAR protection, etc.), are simultaneously and automatically taken into account in the planning process.

The ability to define multiple targets, each with its own dose prescription, permits the definition of complex dose gradient within one tumor and between different tumors. Examples of targets within a target are the boost (delivery of higher dose than the prescription dose) of cancer-validated areas in prostate defined by MRS or of the positive biopsy areas.²⁴^,⁹¹

Different mathematical methods can be applied to minimize the objective function. They can be divided in two groups: deterministic methods and stochastic methods. Deterministic optimization algorithms travel downhill continuously on the objective function surface until a minimum is reached. Deterministic optimization algorithms have been applied for HDR brachytherapy planning optimization ⁹⁵^,⁹⁶ and permanent implant planning optimization.⁹⁷^,⁹⁸ Although deterministic optimization algorithms are faster, they yield a solution at the nearest local minimum depending on the starting conditions and when the best starting condition of the nonlinear objective function is unknown. Consequently, the solution may get trapped in a local minimum that is not the global or near the global minimum except by pure chance. An option to overcome this generic problem is to execute repeatedly the optimization several times using randomly selected starting conditions. However, the physician still has to select one treatment plan from a pool that may contain 100 treatment plans.

Alternatively, stochastic optimization algorithms such as genetic algorithms (GAs) or simulated annealing (SA) algorithms can process any form of objective functions, including nonlinear objective functions. Stochastic optimization algorithms apply a random search and therefore have the ability to overpass the local minima. A GA-class algorithm applies the principles of natural selection for the computation of complex problems. A GA-class algorithm is interesting in its own right and has been previously applied for high-dose-rate brachytherapy planning optimization ⁹⁹^–¹⁰¹ and permanent implant planning optimization.¹⁰²^–¹⁰⁴ However, GA was not originally developed as an optimization algorithm,¹⁰⁵ and GA does not offer any statistical guarantee of global convergence to an optimal solution.¹⁰⁶ Nevertheless, it should be expected that GA might be better suited for some problems than an SA algorithm.¹⁰⁷ On the other hand, an SA-class algorithm does offer a statistical guarantee of global convergence to an optimal solution.¹⁰⁸^–¹¹¹ For that reason, an SA approach seems also to be a reasonable choice for the present combinational optimization problem. Previous investigations have shown that an SA algorithm can be used to govern an optimization process to automatically and rapidly produce plans for permanent implant treatments in the prostate.^85,^86,¹¹²^–¹¹⁴ The same concept has also been applied for prostate HDR brachytherapy.¹¹⁵ SA optimization engines carefully designed for a specific task can find an adequate solution in a very short time, opening the door to intraoperative treatment planning optimization. The next section describes a comprehensive inverse planning simulated annealing (IPSA) algorithm for brachytherapy planning.

Inverse Planning Simulated Annealing Algorithm

IPSA is an inverse planning optimization system based on SA that is now wildly available.¹¹⁵ The IPSA optimization tool has been designed to produce plans for any type of brachytherapy automatically and in a few seconds. These treatment plans are specially optimized for each single patient because the whole routine is controlled by the patient’s specific anatomy. IPSA determines the active dwell positions and their optimal dwell times based on the digitized catheters, the anatomy, and the dose constraints.

Dose Objective Parameters

The total dose D_i at the point i is determined by summing the dose contributions from all source positions j with their respective dwell times t_i. This is expressed as:

in which d_ij represents the dose rate at point i from the source position i. The global dose distribution described by D_i everywhere in space has to be translated into a quantitative value to measure its degree of fulfillment with the clinical objectives. The first step toward the evaluation of the dose distribution is to mathematically describe the physician’s requests by means of dose objectives. The dose objective parameter is a set of numerical values that defines the acceptable dose limits for a specific organ and their relative importance over the other clinical criteria. They convert the dose delivered to a dose point i (D_i) into a penalty value W_i. This conversion is defined by the following penalty relation and illustrated in the following equations:

The coefficients D^min and D^max represent the lower and the upper range of acceptable doses. If the dose is within the permissible dose range, the penalty is null. If the dose goes below or above the range, the penalty increases at rates m^min and m^max respectively. Adjustment of the weights m^min and m^max sets the relative importance between the clinical criteria. The bigger the weight, the stronger the penalty. The standard notation used to abridge the penalty relation is given here:

The advantage of using this type of penalty relation is that the translation from the physician expectations of the ideal dose distribution to a mathematical form is straightforward. The relation defines a clear border between acceptable and unacceptable doses and the physical meaning of dose objective is clear. Physicians have much clinical experience that can be translated into such dose objectives. In addition, this approach of defining the objectives induces an intuitive understanding of the optimization results. Moreover, this method is flexible and can describe different clinical objectives such as target minimal dose coverage and OAR protection. A set of dose objectives for HDR prostate brachytherapy is presented in the Table 13-2. In that case, dose objectives of two targets are defined—the prostate and a dominant intraprostatic lesion¹¹⁶—and the three OARs—the urethra, the rectum, and the bladder.

Table 13-2 Dose Objective Parameters (e.g., Prescription Dose: 950 cGy Per Fraction × 2 Fractions)

The optimization process is guided by dose objective parameters defined for each organ or contours delineated from CT or MRI. D^min and maximum dose (D^max) objectives are defined for each volume with weighting factors emphasizing the relative importance of each dose objective. The higher the weighting factor value, the stronger the penalty to doses outside of the acceptable range between the D^min and D^max objectives.

Organ Type

Each VOI can either be a target (or reference target), an OAR, or ignored. When a VOI is set as a target, all dwell positions within the VOI (or, if a catheter margin is set, the VOI + margin) are activated and dwell times are determined by IPSA. No other dwell positions are activated. This means that dwell positions are never activated in an OAR, or in uncontoured normal tissue.

There are cases in which dwell positions outside a volume must be used to obtain coverage. A catheter margin can be set to include dwell positions close to but outside a target. An additional VOI can be drawn in an area where additional dwell positions should contribute. Dose objectives may or may not be specified for this VOI (e.g., ring applicator). This implies that some VOI may be distinguished not by their organ type but by their functional use. VOI can be used to guide the optimization, others to report the dose, the DVH, and the dosimetric indices.

Modern planning tools such as the anatomy-based IPSA,¹¹⁵^,¹¹⁸ efficiently use the information from the 3-D imaging modalities at all stages of the treatment procedure. IPSA automatically and rapidly produces conformal dose coverage of the target volume while limiting the dose to OARs in the delivery of HDR brachytherapy or permanent implants. With this inverse planning approach, the focus is on the physician’s prescription and dose constraints instead of on the technical limitations. Consequently, the physician’s control of the treatment is improved. With clear anatomic images, functional images to locate the cancerous tissues, and rapid optimization to take all this information into account within the clinical time frame, plans can be generated with dose distribution that fulfill multiobjective dose constraints, sparing organs often not considered before—personalized for each patient—a task inherently difficult to achieve with forward planning.

Plans obtained with IPSA show only a limited dosimetric sensitivity variability resulting from changes in the number of catheters, positioning, and prostate volume and shape, providing additional flexibility to the physician. Inverse planning can achieve dosimetrically equivalent treatment plans using fewer catheters than standard approaches. IPSA can maintain the quality of dosimetry (all within the RTOG 0321 criteria) of unevenly implanted catheters with standard deviations of up 5 mm and compensates for variation in catheter position, making brachytherapy less operator-dependent.

Challenges and Opportunities

Limitation of the Task Group 43 Dose Calculation Formalism

The TG43 formalism, presented in a previous section, has numerous advantages, the first one being a certain standardization of dose calculation formalism. More importantly, the current dosimetric practices, which include TG43, have led to a consistent source strength definition with NIST traceability and uniform prerequisite information about seeds (low-energy sources only initially) and sources. One key clinical advantage of TG43 is that all parameters are either tabulated or can be represented by simple analytic equations. This enables fast dose calculation and its use for computerized inverse planning dose optimization that necessitates hundreds of thousand iterations.

However, the data used extract the parameters assuming that there is only one seed, in a water medium, the medium having a radius of 15 cm. The final dose calculation is then performed by the simple summation of the contribution of all seeds or source positions in an implant. Taking a step back, these conditions make obvious the areas where the formalism could be less accurate: any differences in scatter condition between the data acquisition for parameter extraction and the treatment condition, the presence of any high-Z, high-density shielding materials, any significant differences (this will be defined later in this chapter) between the tissues and water, and attenuation of one source emission by another one in multiple source implants or ISA. An excellent summary can be found in Table 1 in Rivard et al.,¹¹⁹ which is presented by sites for low (¹²⁵I range) and high (e.g., ¹⁹²Ir) energy brachytherapy.

Scatter Condition

As stated previously, the condition set forth for TG43 parameter extraction is precise and requires a water equivalent medium of 15-cm radius. Furthermore, we stress that this radius condition was chosen for low-energy sources such as ¹²⁵I and ¹⁰³Pd. Although this condition can be easily respected in interstitial prostate implants, the geometry might differ significantly for breast implants, head and neck implants, or any other superficial implants. This particular issue has been studied extensively by Perez-Calatayud et al.,¹²⁰ Melhus and Rivard,¹²¹ and Granero et al.¹²² For low-energy sources, the effect of a reduced radius can become important when the distance to the tissue–air junction falls below 10 cm, and can be quite significant for a distance of 5 cm or less. However, as shown in Fig. 13-2, the effect increases with photon energy. As for ¹⁹²Ir, significant differences in the dose deposition can be expected within the first 10 cm of the source, even for a phantom radius of 15 cm.¹²² In conclusion, the effect of the difference in scatter condition increases as the photon energy increases and the distance decreases.

FIGURE 13-2 • Ratio of the radial dose function, g(r) for a phantom size R to that of the unbounded radial dose function for ¹²⁵I (top) and ¹⁹²Ir (bottom).

(From Perez-Calatayud J, Granero D, Ballester F: Phantom size in brachytherapy source dosimetric studies, Med Phys 31:2075, 2004.)

Shielding and Interseed Attenuation

Shielding is commonly found in brachytherapy. Usually shielded applicators are used to increase the protection to normal tissue. A number of examples can be found in gynecologic brachytherapy,¹²³^,¹²⁴ for rectal HDR treatment,¹²⁵ and eye plaque.¹²⁶^,¹²⁷ In all cases, taking into account the presence of the shielding material has shown a dramatic change in dose deposition relative to the current clinical standard.

The physical explanation can be found by looking at the mass attenuation coefficients as a function of photon energy, shown in Fig. 13-3, for various materials used for shielding or that are part of the source composition. It is quite obvious that even for ¹⁹²Ir brachytherapy, the effect of having high-Z materials leads to a difference in the attenuation relative to water. As expected, this effect increases as the photon energy decreases. It is due to the increasing importance of the photoelectric effect.

FIGURE 13-3 • Mass attenuation coefficients of water and various shielding and source composition materials as a function of energy.

(From National Institute of Standards and Technology, retrieved 16 March 2009 from www.nist.gov.)

Another important form of shielding comes from the attenuation of the radiation of one source by the presence of multiple other sources nearby. This happens particularly in PSIs in which 50 to 125 small, low-energy (20 to 30 keV) sources are used, and it is referred to in the literature as the interseed attenuation (ISA). An early study based on regular geometry configuration predicted a decrease of 6% to 10% in the minimum peripheral dose resulting from ISA.¹²⁸^,¹²⁹ More recently, clinical seed configurations along with prostate anatomies were studied.¹³⁰^–¹³² The results indicate that a decrease of 2% to 5% of the clinical parameter D90 can be expected from ISA. This effect of the dose parameters was shown to depend on the radioisotope,¹³⁰ seed density or interseed distances,¹³¹ and seed design.¹³³

Tissue Heterogeneity

Tissue heterogeneity affects the dose calculation on two fronts. First, it affects the attenuation of the radiation as it is transported into the medium. From a physical perspective, this is characterized by the difference in the mass attenuation coefficient between water (TG43 reference medium) and the tissue of interest. Secondly, the difference in dose deposition between tissue and water is given by the mass-energy absorption coefficient, assuming electronic equilibrium. In this context, the differences for both cases depend on the energy and tissue composition (Z). Fig. 13-4 shows the mass-energy attenuation coefficient for water, adipose, breast (ICRU-44), bone, and soft tissue (ICRU-44).¹³⁴ At more than 200 keV, small variations of composition from most tissues found in the human body makes little differences in the attenuation coefficients. This is because Compton scattering dominates the interaction cross-section. However, for lower photon sources such as is found in ¹⁰³Pd, ¹²⁵I, ¹³¹Cs, and low-energy electronic brachytherapy sources¹³⁵ and even ¹⁶⁹Yb, there are significant variations relative to water. Again, this is due to the increasing importance of the photoelectric effect.

FIGURE 13-4 • Mass energy attenuation coefficients of water and various tissues as a function of energy.

(From National Institute of Standards and Technology, retrieved 16 March 2009 from www.nist.gov.)

It has been shown that for ¹⁹²Ir brachytherapy, the ratio of the dose to tissue to that of water is within 1% of a ratio of 1 for the first 10 cm from the source.¹²¹ This is in strong contrast to the study of Reniers et al.,¹³⁶ which looked at the term g(r) of TG43 for ¹⁰³Pd seeds in various tissues. The latter study has demonstrated a strong dependence with distance, resulting in differences to water of −40% at 5 cm for breast tissue and +10% for muscle at the same distance from the source for g(r). For prostate seed implants, a retrospective postimplant Monte Carlo dose calculation study on 28 consecutive cases has shown that tissue heterogeneity accounted for a 3% decrease in D90.¹³² However, there were large variations from patient to patient, which, combined with ISA, resulted in a decrease of D90 from 3% to 14%. Using a lower energy isotope, ¹⁰³Pd instead of ¹²⁵I, decreases even further D90.¹³⁰ Given the current knowledge, it should be obvious that low-energy electronic sources (lower than 100 kVp) will be affected similarly.

Potential Avenues

A few avenues are being explored to solve some of the limitations described in this section. They can be divided roughly into three categories, namely analytical algorithms, the stochastic approach or Monte Carlo, and the deterministic approach, which solve the transport equation. It is beyond the scope of this chapter to get into the details of each, and we refer the reader to recent reviews.^119,^137,¹³⁸ However, because some of these are getting close to clinical release, a few lines should be given. The most common advance analytical methods for brachytherapy proceed by separation of the primary and scatter contribution to the dose. A well-known method is that of the collapsed–cone,¹³⁹ which has been proposed for brachytherapy dose calculation.¹⁴⁰^,¹⁴¹ For the deterministic approach, a solution is proposed by a commercial venture, Transpire Inc. The product named Attila has been recently tested for external-beam radiotherapy,¹⁴¹^,¹⁴² and Acuros, also by Transpire Inc., was commercialized by Varian for brachytherapy. Finally, on the Monte Carlo side, a number of codes are available to perform dose calculation within a research environment: MCNP,¹⁴³^,¹⁴⁴ GEANT4¹⁴⁵^,¹⁴⁶ EGS4,¹⁴⁷^,¹⁴⁸ and PTRAN.¹⁴⁹^,¹⁵⁰ However, recently new, dedicated, and extremely fast MC codes have been promoted for low-energy brachytherapy dose calculations: MCPI¹⁵¹ and BrachyDose.¹⁵² Thus the next few years might see a number of these different approaches used on a regular basis in our clinics.

Dose-Guided Radiation Therapy Real-Time In Vivo Dosi

Adaptive radiation therapy (ART) strategies seek to improve treatment delivery accuracy and patient dose distributions by introducing feedback into the treatment process. This is done using increasingly sophisticated imaging capabilities at the time of treatment. The availability of an image depicting the patient anatomy and the brachytherapy applicator and catheters immediately before dose delivery brings new information that challenges some of the a priori processes of the treatment practice. Changes in anatomy caused by tumor regression, weight loss, or other factors become observable and can be quantified. This questions the validity of using the pretreatment planning CT to represent the patient in treatment position throughout the various fractions of radiotherapy. The assessment of a daily treatment could be performed based on the dose that each organ would receive if the delivery were performed as planned, bringing back the focus on the true end-point of radiation therapy: the delivered dose.

Image-Guided Radiation Therapy

The initial form of ART is image-guided radiation therapy (IGRT). It refers to patient 3-D imaging in the treatment room. The objective of IGRT is to verify the proper positioning of the patient or of the tumor immediately before delivery, with the intent to increase the conformity of the delivered dose. This allows for the reduction of the size of the margin around the target. By minimizing the dose delivered to normal tissue, dose escalation may increase tumor control.

Possibly the most important goal of ART is to manage the real dose delivered globally to the patient and each organ of interest. Effectively, the ultimate strategy of IGRT extends beyond the geometrical verification of the patient anatomy. Dose-guided radiation therapy (DGRT) is a form of ART in which dosimetric considerations constitute the basis of treatment modification and validation. By verifying and tracking the key parameter for treatment outcomes, DGRT has great potential to improve dose distributions and treatment results. Although the concept of DGRT was first introduced in external-beam radiation therapy,¹⁵² it is more easily applied to brachytherapy. In fact, quasi real-time planning for permanent prostate implants is an attempt to perform DGRT. The concept requires performance of a postimplant dosimetry while the patient is still on the operating table, opening the possibility of modifying the dose distribution (inserting more seeds) if the dose distribution is suboptimal. This requires that seeds be visible under US imaging, technically challenging at the moment. Other possibilities include combining other imaging modalities available in the brachytherapy suite. A promising avenue is to use cone beam CT to determine the location of the seeds and US to delineate the prostate. Generally speaking, if the dose distribution can be recalculated before the patient is released, then a comparison with the intended plan can be performed and corrected if necessary.

The treatment outcome is related to the delivered dose. The ability to identify the tumor and delineate the target has been improving at the same pace as the development of new anatomic and functional imaging modalities. More aggressive treatments, dose escalation, and nonuniform dose distribution can be increasingly considered thanks to modern treatment units that can precisely deliver dose. The consequences of delivering the dose at the wrong location are becoming more important for the same reasons.

The ability to evaluate the daily dose distribution and compare it with the planned dose will affect three different fronts. First, it can provide a global quality assurance; the comparison of the reference dose distribution and the current distribution provides a patient-specific quality control based on 3-D in vivo dosimetry. More globally, it provides an in situ, patient-specific global quality control of the entire chain of radiation therapy, including an unprecedented level of treatment documentation. Second, it enables DGRT, allowing it to be an adaptive approach. With external-beam radiotherapy, the evolution of the dose differences observed as the days and weeks pass during the radiation treatment can be used to decide whether the initial plan is still valid, and to determine whether it should be modified or whether the treatment should be replanned. Similar to DGRT, dose-guided brachytherapy is used for treatment adaptation to mitigate dosimetric effect before the dose is delivered or based on the dose the patient has already received. The action level is based on the clinically relevant dosimetric effect. Finally, the knowledge of the delivered dose will enhance our knowledge of dose response (treatment outcomes versus delivered dose). Currently, the actual clinical evidences are effectively based on the planned dose. Because the planned and delivered dose differ, a better understanding of the dose response can be achieved with a more accurate knowledge of the delivered dose. As new biological markers are developed and molecular imaging becomes more specific and precise, long-term patient follow-up will be used to correlate the treatment outcome with the actual delivered dose, hence establishing a more accurate organ-specific dose response relation.

Robotic Brachytherapy

The placement of seeds or catheters requires precise positioning using a reference system registered to the anatomy of the patient. This is why so-called real-time brachytherapy is currently limited to areas of the body (e.g., prostate) that (1) can be imaged by US and (2) allow positioning to be registered using rigid mechanical grids.

Robot technology has the potential to greatly expand the applicability of brachytherapy to treat cancer in new areas of the body. Robot mechanisms achieve precise positioning and registration using precision mechanisms and mathematic coordinate transformations. Robots have the potential to guide seeds to many more locations in the body because they do not rely on rigid mechanical grids or real-time imaging for human guidance.

Robots are increasingly being introduced into clinical practice. These robotic surgical assistants have the potential to provide greater precision and accuracy compared with manually controlled surgical devices, and in many cases can be used with real-time imaging. A pioneer in this area has been the commercially successful da Vinci Surgical System, a robotic surgical assistant for laparoscopic surgery developed by Intuitive Surgical that has been installed in more than 700 locations worldwide.¹⁵³ In addition to the da Vinci system, numerous robotic hardware systems are being developed commercially and in academia for specialized medical procedures. Surveys of recent advances in medical robotics have been written by Taylor and Stoianovici¹⁵⁴ and Cleary et al.¹⁵⁵ Image-guided medical needle insertion procedures in particular can benefit from the more precise control of needle position and velocity made possible through robotic surgical assistants. One of the first medical applications of a robot was to precisely position a needle for a brain biopsy using a programmable universal machine for assembly industrial robot and a CT scanner in 1985.¹⁵⁶ More recently, custom robotic devices have been developed to address the clinical challenges of space constraints, safety requirements, and imaging modality compatibility. Masamune et al.¹⁵⁷ developed an easily sterilized device with detachable electrical parts for needle insertion in stereotactic neurosurgery. Fichtinger et al.¹⁵⁸ developed a robot for transperineal needle insertion for prostate biopsy and therapy that is capable of operating inside the gantry of a CT scanner. Maurin et al.¹⁵⁹ developed CTBot, a stereotactic-guided robotic assistant for percutaneous procedures in the abdomen. Fichtinger et al.,¹⁵⁸ Schneider et al.,¹⁶⁰ and Phee et al.¹⁶¹ developed needle-insertion robots specially designed for prostate brachytherapy with transrectal US guidance. The robot of Phee et al. achieved placement errors ranging from 2.3 mm to 6.5 mm in nine clinical biopsy cases. More recently, several groups have developed robots capable of operating in an MRI scanner, which is particularly difficult because of the limitations placed on robot materials resulting from the magnetic field required for MRI. Chinzei et al.¹⁶² and DiMaio et al.¹⁶³ developed an MRI-compatible surgical assistant designed to operate in an open MRI magnet. Stoianovici et al.¹⁶⁴ developed MrBot, a compact robotic device that uses a novel pneumatic step motor that can operate inside a closed MRI magnet. MrBot was applied to brachytherapy seed implantation in live dogs and achieved a seed placement error in the range of 1.45 to 10.54 mm, with most seeds being placed with an error of 1 to 3 mm.¹⁶⁵ These errors are most likely the result of tissue deformations and seed migration because the robot itself achieved positioning accuracy of the needle tip in the range of 0.86 to 3.18 mm.¹⁶⁵

The delivery of seeds or catheters using robots offer the unique advantage of performing real-time or quasi real-time imaging during the procedure. This allows the validation of the correct location of the needle relative to the anatomy before seeds are dropped, potentially eliminating the errors discussed previously. A series of experiments were performed at UCSF in August 2008 to establish a number of proofs of principle required to clinically deploy image-guided robot brachytherapy. MrRobot was set up in two imaging environments, in a closed MRI 3T magnet, and in a brachysuite equipped with cone beam CT imaging. It was successfully shown that the several reference systems of the imaging systems (MRI or CT), the planning system, and the robot control could be registered to guide the delivery of a seed within 1 mm of a target. Similarly, MRS information was used to guide the robot to certain cancer-validated areas. To demonstrate an adaptive image-guided delivery workflow, MRI or cone beam CT images were acquired during the process to visualize the needle position before a seed was dropped, allowing needle adjustment.

Recently, dosimetric planning studies showed that inverse planning ¹⁶⁶ could be used to optimize new seed distributions, such as divergent needles for permanent implants,¹⁶⁷ or new conical or firework catheter patterns for HDR¹⁶⁸ in the limited space inside a closed MRI. Those new needle or catheter patterns (nonparallel) can potentially deliver equivalent dose distribution to the target while considerably reducing the physical trauma to certain organs. The experiments performed at UCSF with the 3T MRI unit successfully demonstrated that inverse-planned novel catheter patterns could be accurately delivered by the robot.

It is reasonable to assume that robot hardware will soon offer the ability to drive a needle along any desired path in 3-D to reach a desired point that is consistent with prior imaging. This is a unique opportunity to improve the precision of brachytherapy procedures to a level never achieved. This unprecedented precision and consistency in image interpretation, organ avoidance, and tumor targeting, combined with inverse-planning providing implant optimization, a safe needle path, and customized seed distribution accurately delivered by robot will allow high precision of dose delivery resulting in increased treatment effectiveness, reduced side effects, and expanded treatment availability.

References

1 Thomadsen BR, Rivard MJ, Butler WM, editors. Brachytherapy physics, ed 2, Seattle: American Association of Physicists in Medicine, 2005. Medical Physics Monograph No. 31

2 Krempien RC, Dacuber S, Hensley FW, et al. Image fusion of CT and MRI data enables improved target volume definition in 3D-brachytherapy treatment planning. Brachytherapy. 2003;3:164.

3 Milicokovic N, Giannouli S, Baltas D, et al. Catheter autoreconstruction in computed tomography based brachytherapy treatment planning. Med Phys. 2000;27(5):1047.

4 Baltas D, Sakelliou L, Zamboglou N. The physics of modern brachytherapy for oncology. New York: Taylor & Francis; 2007.

5 Mayles P, Nahum AE, Rosenwald J–C. Handbook of radiotherapy physics: theory and practice. New York: Taylor & Francis; 2007.

6 Nath R, Anderson LL, Luxton G, et al. Dosimetry of interstitial brachytherapy sources: recommendation of the AAPM Radiation Therapy Committee Task Group No. 43. Med Phys. 1995;22:209.

7 Khan FM. The physics of radiation therapy, ed 4. Philadelphia: Lippincott Williams & Wilkins; 2009.

8 Rivard MJ, Coursey BM, DeWerd LA, et al. Update of AAPM Task Group No. 43 Report: a revised AAPM protocol for brachytherapy dose calculations. Med Phys. 2004;31:633.

9 Ninkovic MM, Raicevic JJ, Adrovic F. Air kerma rate constants for gamma emitters used most often in practice. Radiat Prot Dosimetry. 2005;115(1–4):247.

10 MD Anderson Cancer Center, retrieved August 2009 from http://rpc.mdanderson.org/