Physiology: On gross examination, bone marrow may be red because of hemoglobin in the erythrocytes and their precursors, indicating active hematopoietic marrow, or it may be yellow as a result of the presence of carotenoid derivatives dissolved in fat droplets within adipocytes.¹ Hematopoietic marrow is rich in vascular sinusoids, whereas fatty marrow is considerably less vascular. During periods of decreased hematopoiesis, the fat cells increase in size and number. During periods of increased hematopoiesis, the fat cells atrophy.

The cellularity of bone marrow diminishes most rapidly in the first two decades of life. The cellularity of hematopoietic marrow is near l00% at birth and decreases to 50% to 75% by 15 years of age. By adulthood, the hematopoietic marrow is composed of approximately 40% fat, 40% water, and 20% protein, with 60% of cells being hematopoietic and 40% being adipocytes.¹ In contrast, the fatty marrow is composed of 80% fat, 15% water, and 5% protein, with 95% of cells being adipocytes (e-Fig. 142-1).^1,2

Imaging: Conventional radiography, computed tomography, and ultrasonography are of limited value in the assessment of bone marrow.³ Since the first reports of magnetic resonance imaging (MRI) of the marrow in children were published in 1984, MRI has emerged as the primary imaging modality to evaluate the bone marrow. It provides a noninvasive method for visualizing the gross anatomic structure of a large sample of the bone marrow and for inferring alterations in its chemical and cellular composition related to a variety of physiologic and pathologic processes. Furthermore, MRI can provide valuable information about regions of the bone marrow that may be inaccessible or difficult to biopsy.

The constituents of bone marrow that contribute to the signal characteristics on MRI are fat, water, and, to a lesser extent, mineralized matrix.³ Fat is the dominant contributor to both hematopoietic and fatty marrow signal intensity patterns. Most fat protons are in hydrophobic methylene (−CH₂-) groups of relatively heavy molecular complexes, conferring very efficient spin-lattice relaxation; this results in a short T1 relaxation time, which results in high signal intensity on T1-weighted sequences.¹ The T2 relaxation time of fat is much shorter than that of free water protons, and water contributes much more to the signal intensity of hematopoietic marrow than fatty marrow.¹ The mineralized matrix of bone has a low density of hydrogen protons that lack mobility within the crystalline structure of bone, accounting for very long T1 and short T2 relaxation times and low signal intensity of mineralized matrix on T1- and T2-weighted sequences. In addition, local field gradients at the trabecular surface are generated by the fixed dipole from the immobile protons and cause magnetic field inhomogeneity. This magnetic susceptibility effect, as well as that resulting from iron deposition, can play a large role in the signal characteristics of hematopoietic and fatty marrow on gradient echo (GRE) images.¹

The normal intervertebral disks, skeletal muscle, and subcutaneous fat have little interindividual and intraindividual variation in signal intensity on T1-weighted sequences during childhood and consequently serve as convenient internal reference standards for comparison with the signal intensity of the marrow.1,4 Fatty marrow has high signal intensity on conventional spin echo T1-weighted sequences. The relative amounts of fat, water, and protein contribute in a complex fashion to produce a longer T1 relaxation time in hematopoietic marrow, with signal intensity that ranges from intermediate to low (less than that of muscle or intervertebral disks) in fat-poor hematopoietic marrow to intermediate to high (greater than that of muscle or the intervertebral disks but less than that of subcutaneous fat) in hematopoietic marrow with larger proportions of fat.^1,5 In neonates, hematopoietic marrow contains minimal fat and has low signal intensity on T1-weighted sequences. With aging, hematopoietic marrow signal intensity progressively increases on T1-weighted sequences, reflecting a progressive increase in fat content. After the neonatal period, hematopoietic marrow has signal intensity equal to or slightly greater than muscle and the intervertebral disks but much less than subcutaneous fat on T1-weighted sequences, whereas fatty marrow approaches the signal intensity of subcutaneous fat on T1-weighted sequences.⁵ Because of the similar proton densities of hematopoietic and fatty marrow, proton density sequences without fat suppression are less useful than T1-weighted sequences.⁶

On fat-suppressed fast spin echo (FSE) T2-weighted and short tau inversion recovery (STIR) sequences, normal hematopoietic marrow in childhood shows higher signal intensity than fatty marrow and higher signal intensity than muscle; the signal intensity of hematopoietic marrow decreases with age and approaches that of muscle by adolescence.

The signal characteristics of marrow are highly variable on GRE sequences, with which images that exploit chemical shift can be obtained by choosing an echo time in which the phases of relaxing water and fat protons are either opposed 180° (out-of-phase or opposed-phase sequences) or coincide (in-phase sequences) related to differences in their resonance frequencies.¹ When hematopoietic marrow contains approximately equivalent amounts of water and fat, as in normal adults, the signal intensity is markedly diminished on opposed-phase T1-weighted sequences, compared with in-phase T1-weighted sequences, as a result of intravoxel chemical shift effect.¹ When the amount of fat and water is no longer balanced, for example, in fatty marrow or in edematous or hypercellular hematopoietic marrow, the marrow signal intensity does not show such a profound difference between the opposed-phase and in-phase T1-weighted images.¹ Magnetic susceptibility effect from trabecular bone and iron leads to a lower signal intensity of the marrow on GRE T2*-weighted sequences than on spin echo sequences.

The degree of contrast enhancement of normal marrow varies with the contrast dose, timing of image acquisition following contrast administration, the age of the patient, and the composition of the marrow. Marrow enhancement peaks within a minute of contrast administration and then slowly declines.⁷ Enhancement is greater in children than in adults, greater in the metaphyses than in the epiphyses, and greater in hematopoietic marrow than in fatty marrow. Enhancement can be imperceptible by visual inspection in the marrow of adults and in the marrow of the epiphyses in patients aged more than 2 years. Interindividual variation in the degree of enhancement is also substantial.⁷ If gadolinium-enhanced T1-weighted sequences are acquired without fat suppression, contrast between hematopoietic and fatty marrow is decreased, which potentially obscures marrow lesions or age-related marrow conversion changes.⁸ Image-subtraction and fat-suppression techniques facilitate the detection of abnormal marrow enhancement on postcontrast T1-weighted sequences.⁶

A combination of T1-weighted and fat-suppressed FSE T2-weighted or STIR sequences is sufficient for the detection and characterization of most marrow lesions.⁹ Contrast-enhanced T1-weighted sequences increase the cost and duration of the MRI exam while providing only modest incremental added sensitivity; because of this, they should be reserved for cases with unclear findings on the precontrast sequences. GRE sequences are valuable for the assessment of the iron content of the marrow, and chemical-shift techniques are useful to detect subtle changes in the fat and water fractions of the marrow.

Practical time constraints previously limited MRI scanning of the marrow to only sections of the skeleton. This disadvantage has been mitigated by the development of fast, whole-body MRI scanning techniques that include FSE and single-shot sequences, parallel imaging, rolling table platforms with a large field of view, and global matrix coil concepts.10–12 Obscuration of lesions by motion artifact may be overcome to some extent by the use of respiratory triggering or other motion-suppression techniques. The lower signal-to-noise ratios in smaller children on whole-body MRI exams can be ameliorated by using MR systems with a higher field strength.¹³ Diffusion-weighted whole-body MRI holds particular promise as a technique to globally assess the bone marrow for hematologic disorders and tumor metastases.^14,15

Molecular diffusion is a stochastic process characterized by brownian motion.¹⁶ Diffusion-weighted imaging (DWI) is based on the MR signal attenuation caused by the brownian motion of water molecules. In biologic tissues, the diffusion of water molecules is influenced by the microstructure of the surrounding environment. Bone marrow is semifluid in consistency and is confined within spaces defined by the bony trabeculae and supported by reticulum cells and adipocytes. DWI is not pure diffusion imaging, and the apparent diffusion coefficient (ADC) reflects both the molecular diffusion of water and the blood perfusion of the microvasculature. At b-values of 30 and 300 seconds/mm², the ADC values of marrow are more affected by perfusion effect than by diffusion.¹⁷

A positive correlation exists between the degree of marrow cellularity and the marrow ADC. Hematopoietic marrow or marrow infiltrated by neoplastic cells has more abundant microvasculature and more intracellular and interstitial free water than fatty marrow, and it exhibits higher ADC values.¹⁸

Motion and susceptibility artifacts are especially problematic in DWI of the marrow, which is contained by bone and in proximity to physiologic motions, such as cerebrospinal fluid pulsations in the case of the spine.¹⁶ Recent advances have largely overcome these technologic challenges, and diffusion-weighted whole-body MRI is being used to evaluate an increasing number of benign and malignant conditions that include chronic recurrent multifocal osteomyelitis, Langerhans cell histiocytosis, and bony metastases.^19–21

Physiology: Hematopoiesis occurs in the yolk sac in the early stages of fetal development; later in gestation, it shifts to the liver, and to a lesser extent, to the spleen. Bone marrow begins hematopoiesis in the fourth intrauterine month, overtakes the liver in this function by the sixth month, and is entirely responsible for hematopoietic cell production by birth.²² Shortly before birth, conversion from hematopoietic to fatty marrow begins in the distal phalanges of the hands and feet and proceeds in a centripetal fashion from the distal to the more proximal portions of the appendicular skeleton.⁴ Within the long bones, conversion from hematopoietic to fatty marrow proceeds from the mid diaphyses to the distal metaphyses and then to the proximal metaphyses. Conversion also progresses from the central medullary canal to the endosteum. Fatty transformation in the epiphyses and apophyses begins almost as soon as they begin to ossify. In infants, the skull and limbs contain about half of the total amount of hematopoietic marrow.^1,4 By early adulthood, hematopoietic marrow becomes confined to the vertebrae, sternum, ribs, pelvis, skull, proximal humeri, and proximal femurs. Approximately one half of the bone marrow volume is fatty marrow in early adulthood and is located primarily in the appendicular skeleton. The involution of hematopoietic marrow continues throughout adult life, although at a slower pace than during childhood.^1,4,23 Unlike skeletal maturation, there are generally no gender differences in the rate of marrow conversion during childhood.² Knowledge of the normal age-related changes in the distribution of hematopoietic and fatty marrow is necessary for the recognition of abnormal conversion and reconversion patterns and for the detection of marrow infiltration by other pathologic processes.

Imaging: Conversion from hematopoietic to fatty marrow is readily detected by MRI because of the high sensitivity of T1-weighted spin echo sequences to fat (e-Table 142-1).¹ In fact, marrow conversion is observed earlier by MRI than by gross pathologic inspection because of the capability of MRI to detect microscopic fat present in marrow.^2,4,5,24 Numerous publications detail the temporal and spatial sequence of marrow conversion revealed by MRI. Although there are discrepancies among these publications in the precise ages of transformation, the sequence of conversion is consistent along the long axes of individual bones and in the skeleton as a whole. Conversion occurs at a faster pace in the appendicular skeleton (extremities, shoulders, and pelvic girdle) than in the axial skeleton (skull, spine, ribs, and sternum).^1,4

Overview: Reconversion is a process by which established fatty marrow is replaced by hyperplastic hematopoietic marrow in response to conditions that create a demand for increased hematopoiesis. In children, reconversion is commonly encountered in patients with chronic anemia—such as from sickle cell disease, thalassemia, and spherocytosis—and in those administered hematopoietic growth factors, such as G-CSF, granulocyte-macrophage colony-stimulating factor (GM-CSF), and erythropoietin.

Etiology, Pathophysiology, and Clinical Presentation: The stimulus for increased oxygen-carrying capacity of the blood in cyanotic congenital heart disease patients, endurance athletes, heavy smokers, and high-altitude dwellers also induces hematopoietic marrow hyperplasia and reconversion. Because marrow conversion is a process that occurs throughout childhood, some of what is construed as reconversion in children actually represents an arrest or delay in conversion from hematopoietic to fatty marrow and is attributable to increased hematopoietic demands (e-Fig. 142-9).⁴

Reconversion occurs in the reverse order of normal marrow conversion, beginning in the axial skeleton and proceeding sequentially to the proximal metaphyses, distal metaphyses, and diaphyses of the long bones of the appendicular skeleton.4,6 The more distal long bones are the last to undergo this process. The epiphyses are usually spared but can undergo reconversion in response to very high hematopoietic demands.

Imaging: Positron-emission tomography (PET) with 18-fluorodeoxyglucose (FDG) may reveal increased FDG uptake in areas of marrow hyperplasia because of the metabolic demands of increased hematopoiesis. The marrow reconversion revealed by MRI is far more extensive than that suggested by technetium-99m (⁹⁹mTc) phosphonate bone scintigraphy. Marrow reconversion usually begins in the metaphysis. On MRI, reconverted marrow follows the signal intensity of hematopoietic marrow and tends to involve the appendicular skeleton in a symmetric fashion.1,4 The reconverted hematopoietic marrow may be distributed in either a homogeneous or patchy pattern and may have an appearance overlapping with that of pathologic processes, such as leukemia and storage disorders. Although MRI is very sensitive to changes in marrow fat content, disease specificity is low, and the distinction between reconverted and diseased marrow often requires clinical correlation. Also, some of the conditions that exhibit marrow reconversion have additional superimposed marrow processes that complicate MRI interpretation, such as transfusion hemosiderosis in thalassemia and marrow infarction and fibrosis in sickle cell disease.

Etiology, Pathophysiology, and Clinical Presentation: Thalassemia is an inherited hemoglobinopathy characterized by ineffective erythropoiesis, intramedullary hemolysis, and anemia. With effective transfusion therapy, hematopoietic to fatty marrow conversion may proceed in the extremities by puberty. However, hematopoietic marrow hyperplasia in the skull, spine, and pelvis can remain pronounced. The paranasal sinuses often fail to develop, in part as a result of abrogation of the normal fatty conversion of marrow that precedes sinus pneumatization (see Chapter 8).³²

Thalassemia patients have an increased risk of pathologic fractures because of osteopenia, arthralgias that are believed to be related to iron overload or the use chelation therapy, and back pain as a result of the high incidence of scoliosis and early intervertebral disk degeneration.³⁵

Imaging: The radiographic changes of thalassemia are due to chronic hematopoietic marrow hyperplasia. These include diffuse osteopenia, undertubulization of bone, premature physeal fusion, “hair-on-end” appearance of the calvarium with widening of the diploic space (Fig. 142-10), decreased pneumatization of the paranasal sinuses, coarse trabeculation of appendicular bones, expansion of costochondral junctions, scoliosis, and soft tissue masses related to extramedullary hematopoiesis.³⁶

The appearance of the bone marrow on MRI is a reflection of the diffuse erythroid marrow hyperplasia, chronic transfusions, and iron chelation therapy.⁴ Iron deposition in the marrow results in lowered signal intensity on T1-, T2-, and especially T2*-weighted images because of T2 relaxation-time shortening and magnetic susceptibility effects.³⁷ However, excess iron deposition can still be identified by MRI in some of those whose chelation is thought to be clinically adequate on the basis of serum ferritin levels. In severe cases of thalassemia, bone marrow transplantation may be pursued as a potentially curative treatment, and MRI can reveal the consequent degree of marrow conversion. FDG-PET in patients with thalassemia shows diffuse increased marrow uptake.³⁸

Treatment: Transfusion therapy suppresses the marrow hyperplasia of thalassemia but at the cost of iron overload. The iron deposition occurs preferentially in areas of hematopoietic marrow and is reduced by chelation therapy.³⁷

Etiology, Pathophysiology, and Clinical Presentation: Sickle cell disease is an inherited hemoglobinopathy characterized by deformation of red blood cells and resultant hemolytic anemia and intravascular sludging of blood. The bone marrow imaging manifestations of sickle cell disease are primarily related to hematopoietic marrow hyperplasia, osteonecrosis, and perivascular fibrosis.³⁹ Because the circulation normally passes centrifugally from the medullary cavity to the cortex, sludging of blood flow in the nutrient artery branches places increased demands on the radially arranged periosteal-cortical system of anastomosing blood vessels to supply the marrow, particularly in the peripheral subcortical medullary cavity.^39,40 This, coupled with the increased oxygen needs from increased hematopoietic activity, accounts for the vulnerability of the bone marrow to osteonecrosis.

Imaging: Osseous abnormalities in sickle cell disease are secondary to osteonecrosis and, less often, osteomyelitis. Radiographic findings are similar to osteonecrosis from other causes, with geographic and patchy areas of sclerosis (Fig. 142-11) that most commonly involve the femoral heads. Other common radiographic abnormalities include dactylitis (e-Fig. 142-12), H-shaped, or “Lincoln log” vertebral bodies from central growth plate osteonecrosis (Fig. 142-13), diffuse osteosclerosis, and widespread periostitis.⁴¹

Findings on MRI include bone marrow changes that reflect red marrow reconversion secondary to chronic anemia, with decreased signal intensity on T1-weighted images and increased signal intensity on fat-suppressed T2-weighted and STIR sequences. Marrow changes secondary to osteonecrosis are more complex and often reflect a combination of acute and chronic processes. Foci of low, serpentine signal intensity on T1-weighted images and abnormal high signal intensity on fat-saturated T2-weighted or STIR images are classic imaging features of osteonecrosis. Chronic osteonecrosis may be suggested when these findings are confined to the marrow itself with absent joint effusions or juxtacortical soft tissue edema (Fig. 142-14). Acute osteonecrosis may be suggested when there is also infiltrative marrow edema, periostitis, and/or juxtacortical soft tissue edema (Fig. 142-15).

Acute osteomyelitis may be difficult to differentiate by MRI from the more common acute osteonecrosis and vasoocclusive crisis (see Fig. 142-15). Ultimately, aspiration and biopsy may be necessary to distinguish these entities, because both can manifest with marrow signal intensity abnormalities, periosteal reaction, and intraosseous, subperiosteal, or extraosseous soft tissue fluid collections. Gadolinium-enhanced MRI can be useful to distinguish vascularized inflammatory tissue from fluid collections and can guide the aspiration of fluid collections.⁴² Some evidence suggests the ability of unenhanced T1-weighted fat-saturated images to differentiate acute osteonecrosis from acute osteomyelitis based on the T1 shortening effects of sequestered red blood cell aggregrates in the marrow, but further studies are needed to validate this technique.⁴³

Treatment: The two most common acute musculoskeletal manifestations of sickle cell disease are vasoocclusive crisis followed by infection. Osteonecrosis is the most frequent presentation of vasoocclusive crisis and is treated with supportive care and pain management.⁴⁴ Infections include osteomyelitis and, less commonly, septic arthritis. Unlike the general pediatric population with musculoskeletal infections, Salmonella is most commonly identified by culture, followed by Staphylococcus aureus⁴¹; therefore antimicrobial therapy should be tailored accordingly in sickle cell patients.

Etiology, Pathophysiology, and Clinical Presentation: The musculoskeletal manifestations of GSD1b are related to a defect in myeloid maturation that leads to bone marrow hypercellularity with hyperplasia of the myelopoietic cells and a leftward shift. Patients with GSD1b have a propensity to develop bacterial infections related to chronic neutropenia.⁴⁵

Imaging: The most common radiographic changes in GSD1b during childhood are osteoporosis, undertubulization of appendicular bones, marked growth retardation, and delayed physeal closure.46,47 Epiphyseal ossification is often delayed, and open physes may persist into early adulthood. When epiphyseal ossification centers appear, they often have a spiculated and fragmented appearance but eventually normalize in shape as the skeleton matures.

On MRI, the signal intensity of the bone marrow varies and is dependent upon whether G-CSF has been administered for treatment. Without such treatment, patchy areas of low signal intensity on T1-weighted sequences and high signal intensity on T2-weighted FSE and STIR sequences reflect myeloid hyperplasia throughout the axial and appendicular skeleton. MR imaging features of glycogen storage disease may be indistinguishable from marrow hyperplasia related to profound anemia or leukemia. With G-CSF treatment, marrow hyperplasia becomes even more profound, and replacement of yellow marrow at the epiphyses may occur. In addition, G-CSF therapy often creates transverse bands of increased signal intensity on T2-weighted FSE and STIR sequences at the level of the juxtaphyseal metaphysis, likely related to hematopoeitic recruitment.25,45

Treatment: Medical nutrition therapy is the primary treatment. Additional therapies include allopurinol to prevent gout, lipid-lowering agents, citrate supplementation to help prevent urinary calculi, angiotensin-converting enzyme inhibitors to treat microalbuminuria, surgery or other treatments for hepatic adenomas, human G-CSF for recurrent infections, and organ transplantation for end-stage kidney and liver disease.⁴⁸

Etiology, Pathophysiology, and Clinical Presentation: Gaucher disease is the most prevalent heritable lysosomal storage disorder. Mutations that confer a deficient level of activity of β-glucocerebrosidase lead to accumulation of the lipid glucocerebroside in the lysosomes of macrophage-like Gaucher cells.⁴⁹ The symptoms and pathology of the type 1 form of Gaucher disease result from the accumulation of Gaucher cells in various organ systems, including the skeletal system. Those with Gaucher disease are subject to recurrent painful bone crises, like those with sickle cell disease.

Imaging: Skeletal imaging manifestations relate to marrow infiltration by Gaucher cells, leading to osteonecrosis, osteosclerosis, osteomyelitis, and predisposition to fractures.49,50 The most common radiographic abnormality in untreated Gaucher disease is undertubulation of the distal femoral metaphysis as a result of marrow infiltration, termed the Erlenmeyer flask deformity; this is less commonly identified in treated patients.

On MRI, replacement of fatty marrow by Gaucher cells results in low signal intensity of the marrow on T1- and T2-weighted sequences. Initially, marrow replacement is often diffuse and homogeneous and mimics a number of systemic metabolic disorders.⁵¹ Marrow replacement is usually confined to the metaphysis and diaphysis, and epiphyseal involvement is unusual except in severe cases.⁵² With time and treatment, marrow changes become patchy. Islands of Gaucher cells are separated by intervening fatty marrow throughout the skeleton with relative preservation of the epiphyseal fatty marrow (Fig. 142-16). The intramedullary presence of Gaucher cells is thought to hinder blood flow within the marrow, thereby predisposing these patients to osteonecrosis (e-Fig. 142-17).

Treatment: Enzyme-replacement therapy results in degradation of Gaucher cell deposits and reconversion to normal fatty marrow.⁵² MRI has been advocated to monitor for an increase in the fat fraction of the marrow as an indicator of response to therapy, and several MRI-based methods of quantifying the bone marrow infiltration in patients with Gaucher disease are under investigation, including Dixon chemical-shift imaging, T1 relaxation-time calculation, and MR spectroscopy.⁵¹

Etiology, Pathophysiology, and Clinical Presentation: Leukemia is the most common childhood malignancy, accounting for up to one third of childhood cancer.⁵³ Leukemia is classified by the morphology, immunophenotype, and cytogenetics of the leukemic cells. Acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML) account for three quarters and one fifth of childhood leukemia cases, respectively. The peak incidence of ALL among children is 2 to 3 years of age, and there is evidence that ALL can initiate in utero.⁵⁴ AML rates are highest in the first 2 years of life, decline to a nadir at 6 years of age, and slowly increase during the adolescent years.⁵⁵

An increased risk of leukemia is associated with certain genetic disorders, including trisomy 21, monosomy 7, neurofibromatosis type 1, and DNA repair disorders such as ataxia-telangiectasia.⁵³ Of special interest to the radiologist, and increasingly the public, is the reported increased risk of leukemia from prenatal or postnatal radiation exposure.^56,57

Imaging: Radiography of the appendicular and axial skeleton is most often normal when the patient is brought to medical attention. When radiographs are abnormal, the most common finding is osteopenia. Leukemic lines refers to juxtaphyseal radiolucent metaphyseal bands (Fig. 142-18), and has been attributed to various causes, including disruption of the zone of hypertrophy related to leukemic infiltration, insufficiency fractures of the metaphysis just above the physis, with rarefaction of the juxtaphyseal metaphysis; it has even been proposed to be a visual artifact related to profound osteopenia. Periostitis may also be seen at initial presentation related to the disease process itself or to insufficiency fractures as a result of osteopenia (e-Fig. 142-19).

On MRI, diffuse decreased signal intensity on Tl-weighted images58,59 and increased signal intensity on fat-suppressed T2-weighted and STIR sequences is seen in the setting of leukemic infiltration. The epiphysis is often involved (Fig. 142-20). Marrow involvement may also have a well-defined nodular appearance, especially in the setting of leukemic relapse.³⁴ The findings are less conspicuous in hematopoietic marrow than in fatty marrow, and consequently these are more difficult to appreciate in younger children, in whom marrow conversion has not yet occurred.³³ The MRI appearance of diffuse cellular infiltration of the marrow in children is not specific for acute leukemia and can also be seen in conditions associated with hematopoietic marrow hyperplasia, in myelodysplastic and myeloproliferative syndromes, and in lymphoma and solid tumor metastases. The findings of periostitis, geographic replacement of fatty marrow by hematopoietic marrow throughout the entirety of the long bones and axial skeleton, and juxtacortical soft tissue edema favor a diagnosis of leukemia over hematopoietic marrow hyperplasia.

During chemotherapy for leukemia, the bone marrow becomes hypocellular and edematous. Following chemotherapy, there is progressive regeneration of normal hematopoietic cells and fat (e-Fig. 142-21).⁶⁰ A marked increase in the marrow-fat fraction is observed by chemical-shift MRI in patients responding to chemotherapy, whereas a low marrow-fat fraction persists in the setting of unresponsive disease. In children with ALL who enter remission, marrow T1 relaxation time normalizes, whereas it remains prolonged in those who do not enter remission.^59,61 These findings imply that MRI could potentially allow earlier prediction of therapeutic response and identification of residual disease, and it could reduce the need for serial bone marrow biopsies.

Treatment and Follow-up: No standard guidelines exist for routine imaging follow-up in children with leukemia. PET imaging, bone densitometry, and whole-body MRI at presentation, treatment, or follow-up may be components of clinical trials or other research investigations but are not part of routine clinical practice.

Imaging is often obtained to evaluate musculoskeletal complications in children treated for leukemia. Musculoskeletal complaints are common at presentation and during follow-up.⁶² The most common complications amenable to imaging diagnosis are osteonecrosis and insufficiency fractures. Osteonecrosis is seen in up to 70% of children screened by MRI, and 15% to 20% have symptoms.⁶³ The reported incidence of fractures in leukemic patients is 18.5% with low lumbar spine bone mineral density and age greater than 10 years as independent predictors of fracture risk.⁶⁴

The most common site of ALL relapse is the marrow, followed by the central nervous system and testes.⁵⁵ MRI may be used to evaluate for relapse; however, its specificity is limited by difficulty in differentiating viable neoplasm from effects of therapy, including hematopoietic marrow regeneration (particularly with G-CSF or GM-CSF therapy), hematopoietic marrow reconstitution following stem cell transplantation, marrow iron overload from transfusional hemosiderosis, and marrow infarction and fibrosis.^65–67 Because of these limitations, MR imaging has not replaced marrow aspirate or biopsy for assessment of therapeutic response in leukemia.⁶⁸

Etiology, Pathophysiology, and Clinical Presentation: Marrow involvement by lymphoma in children usually occurs with Burkitt and lymphoblastic lymphomas and with lymphocyte-depleted Hodgkin disease. The pattern of involvement is usually multifocal, and there is a predilection for sites of predominantly hematopoietic marrow. When involvement is diffuse, an arbitrary threshold of neoplastic lymphoid cells constituting 25% or greater of the marrow cellularity differentiates the diagnosis of lymphoblastic leukemia from lymphoma.⁶⁹

Primary lymphoma of bone should be distinguished from Hodgkin and non-Hodgkin lymphoma with bone involvement. When isolated disease is confined to marrow without disease elsewhere for at least 6 months, a diagnosis of primary bone lymphoma can be made.⁷⁰

Imaging: Disseminated lymphoma with marrow involvement may be indistinguishable from leukemia by imaging. However, primary lymphoma of bone may be seen initially as isolated focal bone involvement that may mimic other primary bone neoplasms. The most common radiographic pattern is lytic-destructive (Fig. 142-22).⁷¹ The lytic pattern may be permeative, consisting of many uniform small lucencies or “moth-eaten” areas made of larger, poorly marginated lucencies.⁷⁰ Primary bone lymphoma often involves the flat bones of the pelvis and shoulder girdle as well as the diaphyses of long bones, similar to Ewing sarcoma.

Unlike leukemia, primary bone lymphoma usually presents as a well-defined intramedullary mass with variable cortical destruction and soft tissue extension. The involved sites typically show low signal intensity on T1-weighted sequences and high signal intensity on fat-suppressed T2-weighted and STIR sequences (see Fig. 142-8). When primary bone lymphoma is unifocal on MRI with masslike features, it cannot be distinguished from other aggressive lesions, such as Ewing sarcoma and Langerhans cell histiocytosis. When diffuse, involvement is often patchy, and up to one third of lymphoma patients with negative bone marrow biopsies have lymphoma involvement visible by MRI distant from the standard iliac crest biopsy sites.^67,72

Although MRI is a sensitive modality for detecting marrow infiltration, it is not routinely used for the staging of lymphoma. FDG-PET detects lymphoma on the basis of the increased glucose transporter activity and glycolysis in lymphoma, and it is emerging as the functional imaging study of choice for the evaluation of lymphoma. Like MRI, FDG-PET is more sensitive than standard bone scintigraphy for the detection of bone marrow involvement.⁷² However, assessment of lymphoma in the marrow after therapy may be limited, with false-positive findings as a result of hematopoietic growth factor effects.⁷²

Treatment and Follow-up: Systemic lymphoma with marrow involvement portends a worse prognosis and is considered stage IV disease.⁵³ Imaging follow-up is based on nonosseous disease, including nodal involvement. Patients with primary bone lymphoma have an excellent prognosis, and treatment is usually tailored to the site of involvement.⁷³

Etiology, Pathophysiology, and Clinical Presentation: Although often asymptomatic and rarely extensive enough to impair hematopoietic function, osteonecrosis can cause pain and disability. It can be separated into epiphyseal and nonepiphyseal osteonecrosis. The terms aseptic and avascular necrosis are alternative terms for epiphyseal osteonecrosis, and bone infarction is an alternative term for nonepiphyseal osteonecrosis.

Epiphyseal osteonecrosis portends a worse prognosis compared with nonepiphyseal osteonecrosis. This is because epiphyseal osteonecrosis and its complications—including cartilage injury, subchondral collapse, and subsequent early osteoarthritis—may compromise joint function.

Conditions associated with an increased risk of osteonecrosis include repetitive trauma, sickle cell disease, Gaucher disease, chronic renal failure, bone marrow transplantation, steroid therapy, pancreatitis, and highly active antiretroviral therapy for HIV infection.⁷⁴ Putative pathophysiologic mechanisms related to nontraumatic spontaneous osteonecrosis include vascular occlusion, elevated medullary cavity pressure, coagulopathy, and altered lipid metabolism. Osteonecrosis most commonly occurs in regions of fatty marrow and is unusual in hematopoietic marrow, except in patients with a hemoglobinopathy.⁷⁴

Legg-Calvé-Perthes (LCP) disease, also called idiopathic osteonecrosis of the capital femoral epiphysis, is one of the most common forms of epiphyseal osteonecrosis. LCP is a classic illustrative example of the clinical course and prognosis related to epiphyseal osteonecrosis at the femoral head and elsewhere. LCP affects boys between 2 and 14 years of age with peak incidence around 5 to 6 years (male/female ratio, 5 : 1). Bilateral involvement is seen in about 10% to 15% of cases, although it is almost always asynchronous.⁷⁵ Forms of idiopathic osteonecrosis that affect other bones include but are not limited to Kienböck disease (lunate; Fig. 142-23), Köhler disease (navicular; Fig. 142-24), and Freiberg infraction (second or third metatarsal head; Fig. 142-25).

For LCP, deformity of the femoral head, neck, and acetabulum results is a common cause of degenerative disease of the hip, particularly in men. Children who come to medical attention before 6 years of age generally have a benign course, whereas those who seek care after 8 years of age fare less well and often require surgery.⁷⁶

Children with LCP are typically seen after having had several weeks or months of limping, often without pain. Physical findings include spasm, limitation of abduction and internal rotation, and atrophy of the thigh and buttocks in a child who is otherwise normal.

Imaging: In acute osteonecrosis, radiographs are usually normal. With subacute and chronic osteonecrosis, radiographs will show geographic and patchy areas of sclerosis (see Fig. 142-11).

In the acute phase of osteonecrosis, marrow hemorrhage, edema, and liquefactive necrosis are evident. On MRI, this may be difficult to distinguish from infection and/or superimposed stress reaction. Acute-on-chronic osteonecrosis may occur, and the presence of edema and periostitis are helpful clues to identify the acute component (Fig. 142-26). Rarely, this may superficially mimic a neoplastic process. Unfortunately, early infection and superimposed stress reaction may be indistinguishable. During the reparative phase, central fatty marrow replacement usually occurs (Fig. 142-27). Osteonecrosis often has a geographic shape with a serpentine margin of low signal intensity on T1-weighted sequences and often does not respect epiphyseal, metaphyseal, and diaphyseal boundaries of the long bones (see Fig. 142-27). T2-weighted sequences may demonstrate a characteristic “double-line sign” consisting of an outer rim of low signal intensity corresponding to sclerotic bone and an inner rim of high signal intensity corresponding to vascularized granulation tissue (Fig. 142-28). Areas of low signal intensity on both T1- and T2-weighted sequences related to osteonecrosis represent fibrosis or calcification. Fatty marrow infiltration manifested by increased signal intensity on T1-weighted sequences within the area of osteonecrosis is believed to indicate central revascularization (see Fig. 142-27). Intense contrast enhancement is common at the periphery of evolving areas of osteonecrosis, and dynamic contrast-enhanced MRI has been advocated for early diagnosis but probably is not necessary in children with established osteonecrosis. The diagnosis can usually confidently be made on conventional MRI sequences without intravenous contrast.⁷⁴

Imaging findings of epiphyseal osteonecrosis related to other causes are indistinguishable from those of LCP, which progresses through four main stages: 1) avascularity, 2) revascularization, 3) healing, and 4) residual deformity (e-Fig. 142-29). Early radiographic diagnosis is based on the detection of periarticular osteoporosis, medial joint space widening and lateral displacement of the femoral head, and eventually a relatively smaller and denser appearance of the femoral head.^77,78

During the revascularization stage, which lasts from 1 to 4 years, surrounding tissues react to the dead bone. In this stage, the child experiences pain, and the most dramatic radiographic changes occur.⁷⁹ The ossific nucleus appears even denser because of new bone formation that occurs within the dead trabeculae; it is easily flattened and deformed. A pathologic subchondral fracture in the anterosuperior ossific nucleus creates the radiographic “crescent sign.” The epiphyseal ossification center undergoes varying degrees of fragmentation, primarily in its central portion (e-Fig. 142-30). The adjacent metaphysis sometimes develops cystic changes and broadens. Disturbed endochondral ossification from ischemia probably results in residual cartilage in the metaphysis.⁸⁰

During the healing phase, new bone slowly replaces granulation tissue in the ossific nucleus, and the epiphysis regains its height. The better the femoral head is contained by the acetabulum, the more spherically the femoral head will remodel. On MRI performed during the healing phase, the physeal cartilage may exhibit irregularity or bridging. Epiphyseal deformity is more pronounced anteriorly; hence, sagittal images are more useful for evaluation.

Residual deformity persists after healing, although the articular cartilage is reasonably preserved. As a result, joint function may be satisfactory for several years. Abnormal shape of the healed epiphysis is characterized by coxa magna, a residual enlargement of the femoral head and neck, and coxa breva, a short femoral neck as a result of premature physeal arrest. Subsequent varus or valgus hip deformity may occur, depending on the location of the physeal fusion. A deformed femoral head, especially with a large anterolateral segment that is uncovered, can impinge on the lateral acetabular lip and cause clinical symptoms of decreased range of motion, pain, and a “clunking” sensation. Joint incongruity leads to degenerative joint disease later in life.

Treatment: Treatment is aimed at managing the underlying condition or modifying administration of the causative agent, such as decreasing corticosteroid dosage. Resting the affected joint may help slow disease progression and may prevent complications, including pathologic fractures. Children with osteonecrosis are predisposed to pathologic fractures, and subchondral collapse occurs with epiphyseal osteonecrosis and pathologic long-bone fractures with nonepiphyseal involvement. For epiphyseal osteonecrosis, the most important predictors of patient outcome, aside from age of onset, are joint congruency and motion.

Treatment of LCP is individualized on the basis of clinical and radiographic findings, including age of onset, range of motion in the hip joint, extent of femoral head involvement, presence or absence of femoral deformity, and lateral subluxation of the femoral head. For many patients, a combination of traction treatment with an abduction cast, nonsteroidal antiinflammatory agents, and gentle range-of-motion exercises are used to enhance molding of the femoral head by the acetabulum. Surgery is indicated in children younger than 8 years who have femoral head deformity and in those older than 8 years even in the absence of deformity. Because LCP often affects the proximal femoral physeal vertex while preserving normal medial physeal physiology, a relative coxa valga may develop related to vertex physeal growth disturbance. As a consequence, surgical reduction to contain the femoral head is often indicated, including subtrochanteric varus osteotomies with or without congruent periacetabular shelf osteotomies, particularly in older patients (second decade). The main goals of surgical therapy are to preserve joint congruity during the active phase of the disease and to contain the femoral head within the acetabulum, preventing extrusion and subluxation. A femoral head that is contained by the acetabulum tends to heal more spherically than one that is partially subluxed. In severe cases, a vicious cycle ensues in which decreased containment leads to increased deformity that in turn leads to further subluxation. Surgery can prevent this progression and can be used to maintain a full range of motion at the hip joint.⁸¹

Etiology, Pathophysiology, and Clinical Presentation: Hematopoietic growth factors such as G-CSF, GM-CSF, and erythropoietin are cytokines that regulate the proliferation and differentiation of hematopoietic progenitor cells in the bone marrow.82,83 Recombinant human hematopoietic growth factors are commonly used to hasten the recovery of hematopoietic marrow from myelosuppressive chemotherapy or bone marrow failure syndromes and to stimulate more effective myelopoiesis in GSD1b.⁴⁵

Imaging: In those treated with G-CSF or GM-CSF, MRI will show signal intensity changes associated with reconversion of fatty marrow to hypercellular hematopoietic marrow that coincide temporally with increases in the absolute neutrophil count. These changes may be observed incidentally on routine imaging studies or on imaging studies prompted by bone pain accompanying G-CSF administration.²⁵ As in other instances of marrow hyperplasia, the imaging changes follow a signal intensity similar to that of red marrow. GRE out-of-phase sequences designed to detect altered proportions of fat and water may be the most sensitive for the effects of G-CSF therapy. The peak of hematopoietic marrow hyperplasia observed by MRI occurs about 2 weeks after discontinuation of G-CSF administration, and the bone marrow alterations normalize in most patients within 6 weeks after treatment. The marrow changes may be diffuse, patchy, or have a masslike appearance, and they can be asymmetric and simulate bone marrow involvement by leukemia, metastatic disease, or other infiltrative process.⁶⁶ The masslike appearance of red marrow on T1-weighted images may have well-defined margins rather than the wisplike, feathery appearance of residual red marrow (see Fig. 142-5). Marrow changes as a result of G-CSF may not follow the typical red marrow pattern of reconversion seen in patients who are anemic. For instance, diaphyseal red marrow islands may appear before metaphyseal red marrow changes are seen. The changes can also obscure underlying marrow lesions.⁸⁴ Masslike marrow replacement owing to G-CSF usually is not associated with juxtacortical soft tissue edema at the same level of the lesion that is more likely with metastatic marrow replacement. Consideration of the timing of the therapy and imaging is necessary to avoid misinterpretation of the marrow changes (see Fig. 142-7).

Similarly, FDG-PET, ⁹⁹mTc sulfur colloid scintigraphy, gallium-67 scintigraphy, and thallium-201 scintigraphy can show elevated radiopharmaceutical uptake by the stimulated hematopoietic marrow (Fig. 142-31). Large interindividual variability is seen in marrow FDG uptake induced by hematopoietic growth factors. FDG marrow uptake returns to normal within 1 month after the discontinuation of G-CSF treatment in most individuals.⁸⁴

The anemia of end-stage renal disease can be treated with recombinant human erythropoietin, which induces an increase in the erythropoeitic marrow that manifests as decreased signal intensity on T1-weighted sequences. Erythropoietin also increases the bone marrow uptake of FDG.

Etiology, Pathophysiology, and Clinical Presentation: In the acute phase, radiation therapy causes depression of the marrow cellularity and vascular sinusoid injury with edema and hemorrhage.⁸⁵ In the chronic phase, the vascular sinusoids are obliterated, and hematopoietic marrow is replaced by fat and fibrosis.⁸⁶

The process of fatty replacement of the marrow is largely irreversible for doses higher than 30 to 40 Gy, because destruction of the vascular sinusoids prevents migration of hematopoietic cells into the irradiated marrow.⁶ For doses less than 30 to 40 Gy, the fatty replacement of the marrow is less complete, because regeneration of the hematopoietic marrow may occur. This regeneration of hematopoietic marrow manifests as a mottled or band pattern in the vertebrae of children between 11 and 30 months after spinal irradiation of doses no greater than 40 Gy, and similar findings have been observed in the marrow of the long bones after irradiation.⁸⁷ Marrow regeneration is more likely to occur with larger volumes of irradiated marrow, suggesting that partial spinal irradiation may not provide enough stimuli for regeneration, because the nonirradiated marrow is sufficient to meet hematopoietic demands.⁸⁷

Imaging: STIR sequences are the most sensitive conventional MRI technique for depicting early postradiation changes in the marrow, and they can detect the effects of radiation therapy within a few days after initiation of treatment.⁶ Areas of increased signal intensity in the marrow on STIR sequences peak 9 days after therapy and reflect edema, hemorrhage, and early influx of nonirradiated cells.^4,25,84 T1-weighted sequences in the acute phase show a corresponding decrease in the signal intensity of the bone marrow and a corresponding increased signal intensity on STIR sequences indicative of marrow edema. From 2 to 6 weeks after irradiation, the signal intensity of the bone marrow begins to increase on T1-weighted sequences and decrease on STIR sequences (Fig. 142-32).⁸⁴ After 6 weeks, the marrow pattern either becomes more homogeneous, with high signal intensity on T1-weighted sequences and low signal intensity on STIR sequences, or it develops a band pattern of peripheral intermediate signal intensity surrounding central high signal intensity on T1-weighted sequences and produces reciprocal changes on STIR sequences.^84,87 In the chronic phase, the marrow also shows a marked decrease in contrast enhancement, reflecting vascular obliteration.^25,85

Marrow regeneration after radiation usually occurs when less than 40 Gy is administered.⁸⁷ Marrow fibrosis is more likely to occur when more than 40 Gy is administered. Additional late changes after radiation therapy include osteonecrosis and radiation-induced osteochondromas.²⁵

Etiology, Pathophysiology, and Clinical Presentation: Bone marrow transplantation is used in the treatment of numerous pediatric malignancies, as well as severe aplastic anemia, hereditary hemoglobinopathies, inborn errors of metabolism, and certain immunodeficiency and autoimmune syndromes. In preparation for bone marrow transplantation, the host marrow is ablated and conditioned by chemotherapy and possibly also by fractionated total-body irradiation to induce immune suppression and eliminate residual malignant cells, if applicable. Stem cells are then infused intravenously and engraft in the marrow within 2 to 4 weeks. Typical fatty and hematopoietic bone marrow is usually seen within 90 days after transplantation.⁴

Imaging: In the interval following bone marrow transplantation, until the recovery of hematopoiesis, hematopoietic growth factors and multiple blood transfusions may be given, and the appearance of the marrow on MRI during this time will reflect the combined effects of marrow necrosis, early hematopoietic reconstitution, and possibly iron overload.88,89

Within 40 to 90 days after transplantation, a band pattern in the vertebral bodies may develop, consisting of a peripheral zone of intermediate signal intensity and a central zone of high signal intensity on T1-weighted sequences with reciprocal signal intensities on STIR sequences.⁸⁸ This pattern corresponds to regenerating hematopoietic marrow peripherally and fatty marrow centrally. The band pattern may gradually evolve into a homogeneous appearance of the marrow.⁶ However, the hematopoietic marrow recovery after bone marrow transplantation may not achieve full reconstitution, corroborated by the observation that the fat fraction in the vertebral and pelvic marrow determined by chemical-shift MRI is higher in transplanted patients for several years after transplantation.²⁵