Insertion Tube Flexibility

As previously mentioned, the handling characteristics of the insertion tube are extremely important, particularly for colonoscopes. For easy insertion, the instrument must be capable of accurately transmitting all of the subtle movements and torque applied by the endoscopist. Any rotation that the endoscopist applies to the proximal portion of the shaft (torque) must be transferred to the distal tip of the instrument in a 1 : 1 ratio, although this capability is lost when the instrument is looped. The torquing ability of the instrument is facilitated by flat, spiral metal bands that run just under the skin of the insertion tube (see Fig. 3.2). Because these bands are wound in opposite directions, they lock against one another as the tube is torqued, accurately transmitting rotation of one end of the tube to the other. At the same time, gaps between these spiral bands allow the shaft to flex freely. The bands also give the insertion tube its round shape. Their stiffness prevents the internal components of the insertion tube from being crushed by external forces. These spiral bands are covered by fine strands of stainless steel wire, braided into a tubular mesh. A plastic polymer layer, typically black (or dark green on colonoscopes), is extruded over this wire mesh to create the smooth outer surface of the insertion tube. The polymer layer provides an atraumatic, biocompatible, and watertight surface for the insertion tube. It is usually marked with numbers to gauge the depth of insertion.

Experience has shown that a more rigid insertion tube is optimal for examining the fixed anatomy of the upper GI tract. The colon, with its tortuosity and freely moving loops, is best examined by a more flexible instrument. The instrument should be sufficiently floppy (nonrigid) to conform easily to the tortuous anatomy of the patient and to exert minimum force on the colon wall and attached mesentery. The instrument also should have sufficient column strength to prevent buckling when the proximal end of the instrument is pushed. In addition to its flexibility, the colonoscope should have sufficient elasticity to pop back into a straightened condition when it is pulled back; this aids in removing loops.

Obtaining the best combination of flexibility, elasticity, column strength, and torquing ability is the art and science of insertion tube design. Improvements in one of these characteristics often negatively affect one or more of the others. The final design is usually a compromise between these ideal characteristics, confirmed by months of clinical testing. To improve insertion further, the flexibility of both gastroscope and colonoscope insertion tubes typically varies from end to end. As Fig. 3.3 illustrates, the distal 40 cm of a colonoscope insertion tube is significantly more flexible than the proximal portion. This variation in flexibility is achieved by changing the formulation of the tube’s outer polymer layer as it is extruded over the wire mesh during manufacturing. As Fig. 3.4 illustrates, the extruder contains two types of resins, one significantly harder than the other. Initially, as the distal end of the insertion tube passes through the machine, a layer of soft resin is applied to the distal 40 cm of the wire mesh. This soft resin is gradually replaced by the hard resin within a transition zone near the middle of the tube. The proximal portion of the insertion tube (50 to 160 cm) is constructed of only the hard resin.¹ The end result is an insertion tube that has a soft distal portion for atraumatically snaking through a tortuous colon, with a stiffer proximal portion that is effective at preventing loop reformation in the portions of the colon that have already been straightened by the colonoscope.

Fig. 3.3 Graphic representation of insertion tube stiffness. Note the variation—the distal 40 cm is more flexible. Note extra stiffness when the instrument is in the “stiff” setting (dashed line) versus in the “soft” setting (solid line).

Fig. 3.4 Composition of insertion tube cover.

Adjustable Flexibility

Clinical experience has shown that endoscopists often disagree over what constitutes the ideal insertion tube. This disagreement may be due to differences in training, insertion technique, or past experience. In addition, some endoscopists have expressed a desire to change the characteristics of the insertion tube during the procedure, based either on insertion depth or on the patient’s anatomy, which has led to the development of an insertion tube with adjustable stiffness.²

Colonoscopes with adjustable stiffness have a tensioning wire that runs the length of the insertion tube (see Fig. 3.2). The amount of tension in this wire is controlled by rotating a ring at the proximal end of the insertion tube, just below the control section (Fig. 3.5). When the pull wire in this stiffening system is in the “soft” position, the stiffening system provides no additional stiffness to the insertion tube beyond that provided by the wire mesh and polymer coat. As Fig. 3.5 illustrates, when the control ring is rotated to one of the “hard” positions, an angled slot in the control ring pulls on the slide pin at the end of the pull wire, stretching the pull wire and placing it under heavy tension. This tension stiffens the coil wire that surrounds the pull wire and adds significant rigidity to the insertion tube. As Fig. 3.3 illustrates, although the base stiffness of the insertion tube is established by varying the mixture of hard and soft resins in the polymer base layer, the insertion tube can be stiffened further at will during the procedure by rotating the stiffness control ring. The variable stiffness mechanism does not run the entire length of the insertion tube so that the distal portion of the endoscope is not affected.

Fig. 3.5 Mechanism for stiffening a variable-stiffness colonoscope.

Distal Tip

Fig. 3.6 illustrates the components found in the distal tip of a typical end-viewing endoscope, such as a gastroscope or colonoscope. The larger of the circular glass lenses on the distal tip is the objective lens. This lens focuses a miniature image of the GI mucosa on the surface of a solid-state CCD image sensor. The image sensor sends a continuous stream of images back to the videoprocessor via a collection of very fine electrical wires. The objective lens and CCD unit are tightly sealed to prevent condensation from fogging the image and to protect the imaging system from damage if fluid were to enter the instrument accidentally. Light to illuminate the interior of the body travels through the instrument via fiberoptic illumination fibers. This light is evenly dispersed across the endoscope’s field of view via a light guide lens system.

Fig. 3.6 Colonoscope distal tip assembly.

Some endoscopes have a single illumination system (as shown in Fig. 3.6). Other endoscope models have two fiberoptic bundles and two light guide lenses to improve illumination on both sides of the biopsy forceps (e.g., snare) and to facilitate the packing of components within the insertion tube. The channel used for biopsy and suction exits close to the objective lens on the distal tip. The position of the biopsy channel relative to the objective lens determines how accessories appear in the image as they enter the visual field. On some instruments, the snare or biopsy forceps appears to emanate from the lower right corner of the image; on other instruments, these accessories enter the visual field from the lower left corner, and so forth. When planning difficult procedures, such as piecemeal polypectomy or hemostasis, it is crucial that the endoscopist know where the accessories will enter into his or her field of view.

The insertion tube also contains small tubes that carry air and water through the instrument (see Fig. 3.2). These tubes typically merge into a single tube a few inches from the distal tip (see Fig. 3.9 further on). This combined air/water tube connects to the air/water nozzle on the tip of the instrument. Under control of the endoscopist, water can be fed across the objective lens to clean it, or air can be fed from the nozzle to insufflate the GI tract. Some gastroscopes and colonoscopes have an additional water tube and a water-jet nozzle on the distal tip for washing debris from the mucosa (see Fig. 3.6).

Fig. 3.7 illustrates the components found in the distal end of a typical duodenoscope. Fig. 3.7A is a schematic cross section through the optical and illumination systems found in the distal tip of the duodenoscope. The objective lens for viewing the tissue is now located on the side of the distal end rather than on the very tip of the instrument. A prism is used to deflect the angle of view 90 to 105 degrees and to convert the instrument into a side-viewing endoscope. The illumination fibers are likewise steeply bent at the tip of the instrument, directing the light to emanate from the side. As in end-viewing instruments, an air/water nozzle positioned near the objective lens directs water across the lens to clean it, followed by air to blow away any remaining water droplets. Air from this nozzle is also used to insufflate the patient. All duodenoscopes have a forceps elevator to deflect actively the tip of any accessory passed through the channel. The elevator mechanism is shown in Fig. 3.7B. This elevator normally lies in a recess within the tip of the endoscope (lowered position). When the endoscopist wishes to raise the accessory up into the field of view, he or she operates a thumb control on the control section of the instrument (not shown). This thumb control pulls on the elevator wire, lifting the elevator out of its recess into a raised position, deflecting the tip of the accessory up into the field of view. In some instruments, such as the V-scope (Olympus Corp, Melville, NY), this elevator has a small groove that entraps a 0.035-inch guidewire to aid in wire stabilization during accessory exchanges.

Fig. 3.7 A and B, Duodenoscope distal tip assembly.

Bending Section and Tip Angulation

The distal tip of the insertion tube of the endoscope can also be manipulated by the endoscopist. The deflectable portion, referred to as the bending section, is constructed differently from the rest of the insertion tube. As Fig. 3.8 illustrates, the bending section is composed of a series of oddly shaped metal rings, each one connected to the ring on either side of it via a freely moving joint. These joints are constructed using a series of pivot pins, each one displaced from its neighbors by 90 degrees. One set of pivots allows the bending section to curl in the up-and-down direction. A second set allows the bending section to curl in the right-and-left direction. Together, they enable the bending section to curl in any direction. The direction of the curl is controlled by four angulation wires that run the length of the insertion tube (see Fig. 3.2). These four wires are firmly attached to the tip of the bending section at the 3 o’clock, 6 o’clock, 9 o’clock, and 12 o’clock positions. Pulling on the wire attached at the 12 o’clock position causes the bending section to curl in the up direction and achieves what endoscopists refer to as “up tip deflection.” Pulling on the wire attached at the 3 o’clock position causes right tip deflection. Pulling the other two wires causes down and left deflection. The endoscopist is able to pull on each of these wires in turn by rotating either the up-and-down or right-and-left angulation knobs. (For simplicity, Fig. 3.8 illustrates only the up-and-down angulation system.) Rotating the up-and-down and right-and-left knobs together produces a combined tip movement (e.g., upward and to the right) and allows the endoscopist to sweep the tip of the endoscope in any direction.

Fig. 3.8 A, Construction of bending section. B, Rotation of angulation knob caused deflection of bending section.

Trichromatic Vision

It has been discovered that nearly any color to which the human eye is sensitive can be matched by mixing light of three colors—red, green, and blue (RGB). If three light projectors are fitted with RGB filters and the projected spotlights are overlapped, the resulting image would appear similar to that shown in Fig. 3.15. The color resulting from the overlap of the red and green projectors is indistinguishable from monochromatic yellow light. Likewise, light from the overlapping green and blue projectors produces the mental sensation of looking at pure cyan light. The overlap of red and blue produces magenta. Surprisingly, where all three of the projectors overlap in the center, the observer sees an area of pure white, with no hint of the three component colors. If the intensities of each of the three projectors are accurately controlled and varied, it is possible to reproduce virtually any spectral color in the central area of the overlap.

Fig. 3.15 Illustration of additive primary colors.

All video images are reconstructed using the three component colors of RGB. Because these three colors can be additively combined to mimic all other spectral colors, they are commonly referred to as the three additive primary colors. These three colors, RGB, are the colors of the phosphors used to create full color images on the face of a video monitor (Fig. 3.16). The colors yellow (Ye), cyan (Cy), and magenta (Mg) also play an important role in video imaging and are referred to as complementary colors. Commercial videoendoscopes currently use two different systems for recreating color. The first commercial video image endoscope, the VideoEndoscope introduced by Welch Allyn in 1983, was based on an RGB sequential imaging system.⁵ Many current instruments continue to use this system. The second system, the so-called color-chip endoscope, has now become the predominant system worldwide. Each color reproduction system has its own advantages and disadvantages, as explained later.

Fig. 3.16 Schematic of red, green, and blue (RGB) sequential endoscope imaging system.

Red, Green, and Blue Sequential Imaging (“Color Wheel”)

The components of an RGB sequential videoscope system are schematically illustrated in Fig. 3.16. The endoscope has a monochromatic (black and white) CCD mounted in its distal tip. The objective lens at the tip of the endoscope focuses a miniature image of the endoscope’s field of view on the photosensitive surface of this CCD. This image is illuminated via a fiberoptic bundle running through the endoscope. This fiberoptic bundle carries light from a lamp within the light source to the distal tip of the endoscope. In contrast to the light used for fiberoptic endoscopes or the light used for color-chip endoscopes, this light is not continuous but is strobed or pulsed. The high-intensity xenon lamp within the light source emits continuous white light with the approximate color temperature of sunlight. A rotating filter wheel with three colored segments (RGB) is placed between this lamp and the light guide post of the endoscope. This filter wheel chops and colors the light falling on the light guide bundle into alternating bursts of red, black (no light), green, black, blue, and black. When observed at the distal tip of the endoscope, this illumination appears to the eye to be a flickering white light, rather than the actual sequential bursts of RGB. Rotating at 20 to 30 revolutions/sec, these three primary colors appear to merge, creating white illumination when observed with the unaided eye.

The purpose of this unique illumination system is to produce three separate monochromatic images, each obtained when the field of view is sequentially illuminated by the three primary colors in turn. During the fraction of a second when the red filter is in the light path, the GI mucosa is illuminated only by red light. The CCD image sensor captures a monochromatic (black and white) image of the mucosa as it appears under this red illumination (Fig. 3.17). Tissue that is naturally reddish in color reflects heavily under red light and appears to be bright. Areas of the tissue with less red reflect red light weakly and appear dark under red illumination. After a monochromatic image of the mucosa is obtained under red illumination, the filter wheel rotates to the adjacent opaque area of the wheel. At this point, the endoscopic illumination goes momentarily dark, and the image on the CCD is read out, directed through a processing and switching circuit, and stored in the “red image” memory bank of the videoprocessor (see Fig. 3.16). After the red image is stored, the filter wheel rotates to place the green filter in the light path. A monochromatic image of the mucosa as it appears under green illumination is obtained by the CCD (see Fig. 3.17). This image is read out and sent to the videoprocessor for storage in the “green image” memory bank. In a similar manner, a third monochromatic image is obtained when the filter wheel rotates to the blue segment. This image is correspondingly stored in the “blue image” memory bank. This sequence of capturing a set of images for each of the three primary colors is repeated 20 to 30 times each second—the precise speed being determined by the specifications of the videoprocessor. Synchronization circuitry matches the rotation of the filter wheel with the readout of the CCD and sequences the switching circuit to direct each new image to the proper memory bank.

Fig. 3.17 Image capture with red, green, and blue (RGB) sequential illumination.

Color-Chip Imaging

As an alternative to the RGB sequential imaging system, some videoendoscopes use a color-chip imaging system. A color-chip CCD is essentially a black and white image sensor with a custom-fabricated, multicolored microfilter bonded to its surface. This filter allows the CCD to resolve directly and simultaneously the component colors of the image. The term instantaneous single-plate CCD is sometimes used for this device to emphasize that all three color components are obtained concurrently by a single “plate,” or CCD. A color-mosaic filter of the type shown in Fig. 3.18 is commonly used in color-chip CCDs. It is possible to design a mosaic filter with myriad different color configurations; however, the color choices shown in Fig. 3.18 are very common. The colors used in this mosaic filter are yellow, cyan, magenta, and green (Ye/Cy/Mg/G). These segments are arranged in a 2 × 2 pixel box pattern that regularly repeats over the face of the CCD.

Fig. 3.18 Color mosaic filter matched to the pixels in a color-chip charge-coupled device (CCD) allows the image sensor simultaneously to capture full color information from the image projected on its photosensitive surface.

Because the final output signals sent to the video monitor must be the standard RGB component images, the image produced behind this Ye/Cy/Mg/G filter must be converted into its primary RGB components before display. This conversion is done by adding and subtracting information from neighboring pixels until the specific values for RGB are calculated for each pixel in the sensor. As Fig. 3.19 illustrates, yellow filter elements absorb blue light but pass red and green light; this enables the pixels behind all yellow filter elements to receive both red and green information. Fig. 3.19 also shows that pixels behind cyan filter elements receive both blue and green portions of the color spectrum. Magenta pixels receive both red and blue light. In a representative block of four Ye/Cy/Mg/G pixels, two pixels receive red information, three pixels receive green information, and two pixels receive blue information (see Fig. 3.18).

Fig. 3.19 Effect of filter color on transmission of red, green, and blue (RGB) light.

By adding and subtracting the information obtained from adjacent pixels using an appropriate algorithm within the videoprocessor, it is possible to derive the individual RGB component values for each block of Ye/Cy/Mg/G pixels. The blue component value can be determined by subtracting the number of charges generated in the green pixel from the number of charges generated in the cyan pixel (B = Cy − G or B = [B+G] − G). The red component value can be determined by subtracting the number of charges generated in the green pixel from the number of charges generated in the yellow pixel (R = Ye − G or R = [R+G] − G). Calculations such as these are repeated for all 2 × 2 pixel blocks across the entire face of the CCD. When the process is completed, all required RGB component values for each pixel block in the matrix will have been calculated. One may question why it is desirable to use a mosaic filter composed of complementary colors (Ye, Cy, and Mg) and to employ a processing algorithm if using primary RGB filters to cover the pixels would yield the RGB component values directly without calculation. The answer lies in the fact that a Ye/Cy/Mg/G mosaic filter has a significant advantage in brightness over an RGB mosaic filter.

When RGB filter segments are used, each pixel is filtered to receive only one of the three primary colors (see Fig. 3.19). A cyan-filtered pixel is exposed to both blue and green light. It is more heavily illuminated than a pure blue or pure green pixel. Likewise, pixels behind a yellow filter (red + green) or a magenta filter (blue + red) receive more photons (light) than pixels behind a pure RGB filter. Because of the increased light intensity passing through a Ye/Cy/Mg/G mosaic filter, a CCD with this construction exhibits far greater light sensitivity. The clear advantage of complementary colored mosaic filters is that because of the increased light sensitivity, these filters allow the videoscope designer to construct an endoscope with a smaller light guide fiber bundle, to maximize the angle of view of the endoscope and to increase the depth of field of the endoscope. All of these features improve performance, but each requires additional light. For this reason, all commercial color-chip endoscopes use complementary color mosaic CCDs.

Reproduction of Motion

The color-chip videoscope has an inherent advantage over the RGB sequential videoscope in reproducing motion. The filter wheel in RGB videoprocessors typically rotates at 20 to 30 rps (revolutions per second). Because each of the color component images is captured individually, in sequence, it takes second (with a 30-rps filter wheel) to capture the three component images that make up a single video image. If there is relative motion between the endoscope and the object being viewed, as often occurs during endoscopy, the three component images may differ slightly with respect to object size and position. When these three RGB images are subsequently superimposed on the video monitor, they are likely to be misaligned. This misalignment is clearly visible if the endoscopist happens to freeze the image while it is moving rapidly. Although the RGB sequential videoscope has difficulty reproducing motion, the color-chip videoscope is excellent at imaging moving tissue because all three color components of the image are captured simultaneously. Because the illumination is continuous and nonstrobed and the frame rate is consistent with contemporary television standards, reproduction of moving images with a color-chip videoscope is smooth and natural.

Another unique advantage of a color-chip videoscope is a feature that allows its effective shutter speed to be shortened to increase the sharpness of captured images. The color-chip system normally captures a new video image every second. Although this time period is relatively short, quickly moving subjects that are frozen may appear to be slightly blurred (but with no color separation) because of movement during the capture period. To reduce this blur, it is advantageous to shorten the electronic capture period to a fraction of its normal time (e.g., from second to second). As in traditional film photography, the shorter the exposure period, the sharper the subject, but the more brightly the subject must be illuminated to prevent underexposure. The fast-shutter mode found on many color-chip endoscopes may not provide enough light for distant panoramic images, but in situations in which it is truly needed, the fast-shutter capture mode is very effective at producing bright, sharp, frozen images (i.e., close-up still images of quickly moving mucosa).

Advantages of a Color-Chip Videoscope

The color-chip videoscope has several inherent advantages over the RGB sequential system (Table 3.1). Advantages discussed previously include (1) a smooth, natural reproduction of motion; (2) the absence of color separation on frozen images; and (3) a fast-shutter mode that prevents image blur of even the fastest moving subjects. Additional advantages include (4) compatibility with standard (nonstrobing) xenon light sources, (5) increased transillumination, and (6) superior performance during laser therapy. With RGB sequential endoscopes, abdominal transillumination is problematic because its strobed light output is substantially weaker than that of nonstrobed systems. Many RGB sequential light sources have a means for temporarily removing the spinning filter wheel from the light path when operating in the “transillumination” mode. A steady, intense white light that is ideal for transillumination is produced. However, once the filter wheel is removed, the image is lost because in most cases the illumination is so intense that it saturates the CCD, producing a largely white image. Even if an image is visible, it is in black and white because the filter wheel must be in its proper position to reproduce color.

Table 3.1 Advantages and Disadvantages of Videoscope Imaging Systems

RGB, Red, green, and blue.

Advantages of the Red, Green, and Blue Sequential Videoscope

A major advantage of the RGB sequential videoscope is the opportunity for increased resolution. Image resolution depends heavily on the number of pixels in the original image. The color-chip system requires information from several pixels, which is processed via an algorithm, to obtain the RGB component values for each individual point within the image. In the RGB system, each pixel is illuminated by RGB light sequentially. Each pixel provides information on each of the three color components in turn. The fact that a single pixel can provide all three color components is an advantage for small imaging devices such as endoscopes. In practice, this advantage is not significant for most videoendoscopes, but it is a significant advantage when the thinnest possible endoscope is required (e.g., videocholedochoscope). Because the RGB sequential videoscope uses primary-colored filters and because the color components are isolated, captured, and processed separately within the videoprocessor, this type of videoscope provides very accurate color information. This potential advantage is not usually apparent with routine endoscopy; however, in image analysis research, the RGB sequential system is better than the color-chip system.

Narrow Band Imaging with a Red, Green, and Blue Sequential Videoscope

In recent years, Olympus (Tokyo, Japan) has introduced videoendoscopy systems with a new observation mode called narrow band imaging (NBI).^6–9 The purpose of this feature is to enhance the observation of mucosal surface detail (e.g., pit patterns) and to increase the contrast between microvascular structures and the surrounding tissue through the selective manipulation of the imaging of hemoglobin. Hemoglobin is a major chromophore in tissue. Fig. 3.20 shows the relative absorbance of light by oxyhemoglobin over the visible spectrum. Peak absorption occurs at 415 nm (blue light). A secondary peak is observed around 540 nm (green light). It is clear from Fig. 3.20 that hemoglobin reflects rather than absorbs red light, giving hemoglobin its characteristic color. Using such information regarding the characteristic absorbance of hemoglobin, it is possible to design an imaging system that increases image contrast based on the relative presence or absence of hemoglobin in the tissue being endoscopically studied.

Fig. 3.20 Absorbance spectra of oxyhemoglobin.

Fig. 3.21 schematically illustrates the implementation of NBI imaging in an RGB sequential videoscope system. The light source has the same xenon lamp and the same rotating RGB filter wheel as the standard endoscopy system illustrated in Fig. 3.16. However, for NBI imaging, a special NBI filter has been temporarily placed in front of the xenon lamp. This device filters the broad-spectrum white light produced by the xenon lamp (Spectrum 1 in Fig. 3.21) and allows only a very narrow band of blue light with wavelengths centered around 415 nm and a very narrow band of green light with wavelengths centered around 540 nm to pass through the filter (Spectrum 2 in Fig. 3.21). It is this use of narrow bands of light (i.e., very limited wavelengths of light) to illuminate the tissue that gives NBI its name.

Fig. 3.21 Narrow band imaging (NBI) with red, green, and blue (RGB) sequential videoendoscope system.

(Endoscopic images courtesy of Y. Sano and S. Yoshida.)

The imaging sequence of an RGB sequential NBI system is as follows: When the red filter segment is in the light path, no light enters the endoscope because there is no red light in the NBI-filtered light. No image is created, and no image is stored at this point. The filter wheel rotates to the green filter segment. The green light (540 nm) passing through the NBI filter also passes through the green segment of the rotating filter wheel, passes through the endoscope, and illuminates the tissue. (The 415-nm blue light emitted by the NBI filter is blocked by the green filter in the rotating filter wheel.) A monochromatic (black and white) image of the tissue as it appears under this narrow band of 540-nm green illumination is captured by the CCD (see Fig. 3.21A); however, rather than storing this image in the expected green image memory bank, the videoprocessor switching circuit sends the 540-nm image to the empty red image storage bank (see Fig. 3.21C). This action is an intentional reassignment of color.

Finally, the filter wheel rotates to the blue filter segment. This filter segment passes the 415-nm blue light from the NBI filter, and the tissue is illuminated by a narrow spectrum of blue light centered at 415 nm. The CCD captures a black and white image of the mucosa as it appears under 415-nm illumination (see Fig. 3.21B), and the videoprocessor stores the image as expected in the blue memory bank (see Fig. 3.21E). The videoprocessor also stores an identical copy of this 415-nm image in the green memory bank (see Fig. 3.21D), a second intentional reassignment of color. The red component image now consists of an image of the tissue as it appeared under 540-nm narrow band illumination (green light), and the green and blue component images now consist of images of the tissue as it reflected the 415-nm narrow band illuminating light (blue light). After one complete rotation of the RGB filter wheel and the assignment of images to each of the RGB image storage areas, the three RGB component images are sent simultaneously to the video monitor, where the three component images (see Fig. 3.21C-E) are superimposed as a final full-color NBI image (see Fig. 3.21F). This final NBI image is quite different from the image of the tissue as it would appear under normal white light illumination (see Fig. 3.21G). The color reassignment has altered the natural coloration of the tissue, and the selective illumination of the tissue using only the wavelengths that are highly absorbed by oxyhemoglobin has greatly increased the surface pattern of the mucosal lesion.

Narrow Band Imaging with a Color-Chip Videoscope

NBI can also be implemented on color-chip video platforms. Fig. 3.22 illustrates the insertion of an NBI filter in the light source of a color-chip endoscope. The NBI filter filters the broad-spectrum white light emitted by the xenon lamp (Spectrum 1 in Fig. 3.22) to the same NBI spectrum used in the RGB sequential system. Color-chip NBI illumination consists of continuous, simultaneous illumination of the tissue by 415-nm and 540-nm narrow band light. The color-chip CCD in the endoscope captures an image of the tissue reflecting under this special NBI illumination. Following image capture, the image information is processed by the videoprocessor before display. Fig. 3.23 summarizes the processing of the color-chip NBI image. The color mosaic filter on the CCD filters the image as explained earlier. However, because there are no red wavelengths in the NBI illumination, there is no red image information entering the color processing circuits in the videoprocessor—only blue and green information is available. Similar to the RGB sequential NBI system, the green color information is intentionally reassigned by the videoprocessor for display as a red component image. Also similar to the RGB sequential system, the blue image information is displayed as a blue component image and is intentionally reassigned as a green component image. When these RGB component images, as defined, are simultaneously displayed on the video monitor, the color-chip videoscope system displays an NBI image that is similar to that of the RGB sequential videoscope system.

Fig. 3.22 Narrow band imaging (NBI) with color-chip videoendoscope system.

Fig. 3.23 Narrow band imaging (NBI) color processing with color-chip videoendoscope system.

The NBI function in both types of systems is easily turned on and off via a switch on the endoscope or via a button on the videoprocessor. The endoscopist can quickly alternate between standard white light and NBI imaging at will.

Image Processing

All videoprocessors convert the endoscopic image to a digital format during processing. When in a digital format, it is very easy to manipulate the image via various algorithms. Some more typically used algorithms are designed to produce various levels of “edge enhancement” or “texture enhancement.” These algorithms do not intentionally change the color of the tissue but add image contrast to enhance the delineation of edges or enhance tissue topography. In a sense, they attempt to “sharpen” the image.

Postprocessing Image Enhancement

Efforts have been made more recently to manipulate the image (postprocessing) to obtain data similar to that obtained with NBI. Fujinon (Saitama, Japan) has introduced a feature called FICE (Fuji Intelligent Color Enhancement), and Pentax (Montvale, NJ) has introduced a feature called i-Scan, which intentionally manipulate the color of the endoscopic image.^10–13 These postprocessing features are designed to enhance the display of mucosal blood vessels and surface structure. Olympus’ NBI feature does this through illuminating the tissue at selected wavelengths corresponding to the natural absorption of hemoglobin. FICE and i-Scan use normal white light illumination, capture a normal white light image, and perform digital manipulations on the resulting normal endoscopic image. This process is variously referred to as “optimal band imaging” or “multiband imaging.” When the endoscopic image is in a digital form, it is easy to identify the specific color of each point (pixel) in the captured image and to select only the pixels that fall within a designated color range. These selected colors can be displayed as a new image, manipulated to perform a desired enhancement, or reassigned for display as a different color.

The FICE system allows the user to select from the white light image certain narrow ranges of color that are then used as “optimal band images.” Three such optimal band images can be selected by the user and assigned to the RGB monitor inputs. Ten factory-determined optimal band presets are available in FICE configured processors; however, presets can be customized by the user. Any of 60 estimated wavelengths (color hues) can be input into any of the three RGB channels. A button on the control section of the endoscope allows the operator to switch from conventional white light imaging to real-time multiband imaging.

The optimal multiband imaging presets for tissue diagnosis or differentiation are yet to be determined; however, the concept is easily illustrated. Although endoscopic diagnosis typically consists of interpreting subtle changes in the various hues of red seen in endoscopic images, postprocessing is more easily shown using a photograph containing vivid colors; Fig. 3.24A is such an image of a hot-air balloon launch. Once the image is in digital form, it is easy for a software algorithm to select and extract only certain specific hues from the image. In the case of Fig. 3.24B, the software (Adobe Photoshop) has selected a specific range of blues and a specific range of greens as colors of interest. This range of colors could be chosen to be very narrow (e.g., a certain shade of dark blue) or could be chosen to be very broad (e.g., ranging from light blue to dark blue). These selected colors can be displayed as they would normally appear or reassigned to a different color to create added emphasis. In Fig. 3.24C, the selected blue colors were reassigned to be green, and the selected green colors were reassigned to be blue.

Fig. 3.24 Example of image processing. A, Original image. B, Selection of colors with blue and green hues. C, Color reassignment of selected colors.

As one can imagine, the ways in which images can be manipulated by digital postprocessing are endless. However, in every case, the colors of the original white light image are being manipulated. NBI manipulates the interaction of the illuminating light with the tissue itself by illuminating the tissue with only specific wavelengths that achieve the desired contrast change in the tissue. In the postprocessing example shown in Fig. 3.24C, although the software effectively exchanged the colors of the sky and the grass, it also changed the colors of other objects in the scene that had similar colors in the white light image (e.g., the blue band on the highest central balloon). If certain endoscopic structures, such as capillaries, glands, and pits, always had a unique characteristic color, it would be possible to design a software algorithm to identify and enhance them. Although no image processing schema has been validated or standardized for routine clinical use, research is ongoing to identify valuable algorithms.