A comparison of modern imaging modalities in osteonecrosis of the femoral head

Clinical Radiology (1990) 42, 427-432 A Comparison of Modern Imaging Modalities in Osteonecrosis of the Femoral Head M. J. LEE, J. CORRIGAN*, J. P. STACK and J.T. ENNIS Institute of Radiological Sciences, Dublin and *Department of Orthopaedic Surgery, Mater Misericordiae Hospital Dublin Eleven symptomatic and seven asymptomatic patients, considered to be at high risk of osteonecrosis of the hip, were studied using plain radiography, scintigraphy, computed tomography (CT) with multiplanar reconstruction (MPR), and magnetic resonance imaging (MRI), to determine the precise roles of the various imaging modalities in detecting and staging femoral head osteonecrosis, and to evaluate the incidence in an asymptomatic high-risk patient group. Osteonecrosis was best detected by MRI and was particularly useful for the diagnosis of early disease when other imaging modalities were negative. It is recommended that MRI be performed for the detection of early osteonecrosis, while CT/MPR be used for accurate staging and treatment planning in the established disease. Lee, M.J., Corrigan, J., Stack, J.P. & Ennis, J.T. (l 990) Clinical Radiology 42, 427-432. A Comparison of Modern Imaging Modalities in Osteonecrosis of the Femoral Head Osteonecrosis of the femoral head may be either idio- pathic or secondary to interference with the blood supply to the bone. The latter may occur as a consequence of: extra- or intra-osseous vessel occlusion by embolism, thrombosis or external compression; vessel wall disease (vasculitis); or traumatic disruption of vessels. Primary osteonecrosis is spontaneous in onset and has no known cause although associated conditions, such as steroid therapy and chronic alcoholism, have been implicated (Marcus et al., 1973). Osteonecrosis of the femoral head is a disorder typi- cally found involving individuals of 20 to 50 years of age. Clinically, it is silent with an insidious onset. Pain may only occur when associated synovitis secondary to dege- nerative joint disease is present. Complications include fractures through the site of bone infarction and infection complicating a bone infarct, particularly in sickle cell anaemia. As the disease progresses, secondary osteoarth- ritis is the rule, causing considerable pain and decreased mobility. With the advent of new treatment modalities, including core decompression and transtrochanteric rotational osteotomy, early detection and accurate staging of osteo- necrosis is mandatory. Traditionally, plain radiography was used for staging but it is now well known that radiographic changes in steroid induced osteonecrosis generally lag 6 months behind initial steroid administra- tion (Marcus et al., 1973). It is not surprising, therefore, that radiographic sensitivity in the detection of the early stages may be as low as 41% (Conkin et al., 1983). Scintigraphy is more sensitive than plain radiography in the early detection of femoral head osteonecrosis; however, findings may be difficult to interpret during the early phase, and comparison with the opposite hip is often inaccurate as the disease may be bilateral in a high percentage of patients (Mitchell et al., 1986a). Precise anatomic location of the avascular segment is also unavailable with scintigraphy. Correspondence to: Dr M. J. Lee, Institute of Radiological Sciences, 52 Eccles Street, Dublin 7, Ireland. Newer imaging modalities, such as computed tomogra- phy and magnetic resonance imaging, offer the advan- tages of high spatial and contrast resolution. Precise anatomic localization of avascular segments can be performed with these modalities, as bony architecture is defined without overlap. Osteonecrosis can be discrimi- nated from other hip diseases by assessment of the structural and signal features on MRI, allowing a specific diagnosis to be made (Glickstein et al., 1988). In the recent literature, sensitivities ranging between 90 100% have been attributed to magnetic resonance imaging (MRI) in the early detection of femoral head osteonecro- sis (Jergensen et al., 1985; Mitchell et al., 1986a; Mitchell et al. 1986b; Thickman et al., 1986). The high sensitivity of MRI is due to its ability to depict the cellular changes that occur before fracture, collapse and repair of the femoral head. It has also been reported that MRI can detect femoral head osteonecrosis in asymptomatic but high- risk patients (Mitchell et al., 1987; Coleman et al., 1988). In one such study, early stage focal lesions were seen on MRI in 14 of 16 asymptomatic patients (88%) (Coleman et al., 1988). Little published material is available on the use of computed tomography (CT) in femoral head osteonecro- sis. Dihlmann (1982) has described irregular sclerosis and lucencies of the femoral head formed by the interdigitat- ing tensile and compressive trabeculate. Magid et al. (1985) found CT with multiplanar reconstruction (CT/ MPR) to be superior to plain radiography in the early detection of femoral head osteonecrosis. CT/MPR has not, to our knowledge, been compared with MRI in this regard. The aims of this study were to determine the efficacy and define the precise roles of plain radiography, scinti- graphy, CT/MPR and MRI in osteonecrosis of the femoral head. MATERIALS AND METHODS Two groups of patients (Groups 1 and 2) were examined. Group 1 comprised 11 patients (5 males and 6 428 CL IN ICAL RADIOLOGY females) with 13-painful hips, and an average age of 40 years, some of whom were high-risk patients. Three patients had had a traumatic incident to the hip or acetabulum. Three patients had been treated with steroids (two post cardiac transplant and one with connective tissue disease); one patient had sickle cell anaemia and one patient had chronic renal failure. In three, there were no predisposing factors (idiopathic). Histological confir- mation was available in seven of these patients. Diagnosis of osteonecrosis was established in the remainder by characteristic clinical and serial radiographic changes. Group 2 consisted of seven asymptomatic but high-risk patients (6 males, 1 female) with an average age of 32 years. These patients had all received high-dose steroid therapy post cardiac transplant. The average time of imaging post transplantation was 15 months. AP and lateral radiographs, scintigraphy, CT/MPR and MRI were performed on all patients within a 2 week period. Osteonecrosis on AP and lateral radiographs was graded according to the criteria of Cruess (1978). With this classification, a normal appearing hip is Stage 1; Table l Patient Causative factor Radiographic Scintigraphy Pathology stage I S tero id therapy I + ON 2 (Rt ) S tero id therapy I I - ON 2 (Lt ) S tero id therapy 11 - ON 3 Pos t t raumat ic I I + - 4 Pos t t raumat ic I1 + - 5 Id iopath ic I I + 6 S ick le cell anaemia IV + ON 7 Post t raumat ic IV + ON 8 Id iopath ic IV + ON 9 (Rt ) Chron ic rena l fa i lu re IV + - 9 (Lt ) Chron ic rena l fa i lure IV + 10 Id iopath ic V + ON 11 Stero id therapy V + ON 12 (Rt ) * S tero id therapy 1 12 (Lt)* S tero id therapy I + * Pat ient f rom Group 2. osteolysis or sclerosis in the femoral head is Stage 2; a subchondral lucent crescent is seen in "stage 3; the subchondral bone is collapsed in Stage 4~ and acetabular involvement with joint space narrowing and secondary degenerative disease is present in Stage 5. Bone scintigraphy was performed using 99Tc m methy- lene diphosphonate. Anterior and posterior views of the pelvis, as well as pinhole views of both hips, were obtained at 2 hours post-intravenous injection of 20 mCi (740 MBq) of the radiopharmaceutical. Regions of increased or decreased activity were noted on isotope scans. Non-contrast CT scans were obtained using a Siemens DRH scanner. Patients were placed supine within the gantry, with the feet in the neutral anatomic position. The knees and feet were taped together to prevent inadvertent motion. A scannogram of the pelvis and hips was first obtainedl Scan lines were entered on this scannogram 2 cm above the dome of the acetabulum and at 2 cm below the acetabulum. Contiguous 2 mm thick slices at 2 mm intervals were obtained between these scan lines. The technical scanning factors used were 125 kVp and 330 Mas, with a 7 s scan time. Approximately 30-40 sections were thus acquired per patient. Following acquisition of the transaxial images, the Somatom multiplanar recon- struction and display (MPR/MPD) software package were used for data manipulation. Coronal, sagittal and oblique two-dimensional reconstructions were obtained on each patient. The desired reconstructions were pro- duced by adding identical rows from each transaxial image in the plane of reconstruction initially defined by the operator. The average reconstruction time was two minutes in this study. Magnetic resonance imaging was performed using a 1.5 Tesla super-conducting unit (Sie- mens Magnetom). All patients were examined in the body coil (transmit/receive) using T1 weighted (TR = 500 ms, TE = 17 ms), T2 weighted (TR = 2000 ms, TE = 80 ms) and density weighted (TR = 2000 ms, TE = 28 ms) spin- echo sequences, with a 256 x 256 matrix. Coronal and axial TI weighted, contiguous 5 mm sections, with two excitations were obtained on all patients. T2 and density weighted sequences were performed if characteristic features were not evident on T1 weighted sections. MRI could not be performed on one patient because of an indwelling cardiac pace-maker. Table 2 Patient Radiographic stage CT upgrade MRI class Fractures Joint fluid Joint space narrowing MRI CT MRI CT MRI CT 1 1 111 A 2 (Rt ) 11 I I No MRI 2 (Lt ) 11 11 No MRI 3 11 I I A 4 I I IV C 5 I I IV C 6 IV IV D 7 IV IV C 8 IV V D 9 (Rt ) 17 IV D 9 (Lt ) IV IV D 10 V V D 11 V V D 12" (Rt ) I I A 12" (Lt ) I I I A + + - _ _ - + + - _ _ - + + - _ _ - + + -- _ _ - + + -- _ + + + + -- _ _ - + + -- _ _ - + + + - _ + + + + -- + + * Pat ient f rom Group 2. OSTEONECROSIS OF THE FEMORAL HEAD 429 (a) (b) (c) (d) (e) (f) Fig. 1 - Plain radiograph, scintigram, CT/MPR and MRI of Group 2 asymptomatic patient with bilateral early osteonecrosis demonstrated by MRI. (a) Normal antero-posterior radiograph of the pelvis and hips (Stage 1). (b) Two hour delayed film from bone scan (99Tcm MDP) showing generalized increased uptake in the left femoral head. The right hip appears normal. (c) There is intense sclerosis in the left femoral head. The right femoral head is normal. (d) CT/MPR with oblique reconstruction along the line of the right femoral neck and head. The line of reconstruction is displayed on the scannogram below the reconstructed image. Good definition of the femoral head and joint space is seen which appear normal. A number of the transaxial cuts contained artefactual data which was incorporated into the reconstruction leading to the broad band of artefact across the reconstructed image. (e) Similar CT/MPR oblique reconstruction on the left femoral head as in (d). On the left side there is intense sclerosis of the femoral head and thickening of stress trabeculae (arrows) compared to the right. These changes are consistent with osteonecrosis. (f) Coronal T 1 weighted MRI image, again showing bilateral focal wedge-shaped areas of fatty signal intensity with surrounding 10w intensity rims (arrows), thought to represent the reparative front. A classical distribution for osteonecrosis in the antero-superior margin of the femoral head is also seen. k 430 CLINICAL RADIOLOGY RESULTS Characteristic radiographic changes of osteonecrosis were seen in 12 of the 13 symptomatic hips in the 11 patients in Group 1. Five had Stage 2 disease, 5 had Stage 4 disease, and 2 had Stage 5 disease. No patients with Stage 3 were present in this study group (Table 1). One symptomatic hip appeared normal radiographically (Stage 1). Scintigraphy was abnormal in 11 of these 13 hips. The patient with Stage 1 disease radiographically had a photopenic area in the femoral head, while increased uptake of the radiopharmaceutical was noted in 10 hips (Table 1). Scintigraphy on one patient with bilateral Stage 2 radiographically did not show any abnormality. Abnormalities seen on CT/MPR of the femoral head included focal sclerotic and cystic areas along with thickening and sclerosis of primary stress trabeculae. Fractures and joint space narrowing were clearly depicted on CT/MPR. Radiographic staging was upgraded in four patients by CT/MPR (Table 2). One patient with Stage 1 disease was upgraded to Stage 3. Two patients with Stage 2 disease were upgraded to Stage 4, and one patient with Stage 4 disease was upgraded to Stage 5. No joint effusions were seen with CT/MPR. MRI abnormalities consisted mainly of focal, wedge- shaped areas of altered signal in the femoral head or a more generalized, irregular area of altered signal in the marrow cavity. MRI was positive in all 13 painful hips of the Group 1 patients. MR! findings were classified according to the method of Mitchell et al. (1987), into Classes A, B, C, D. Fractures, joint space narrowing and the presence of joint effusions were also assessed on MRI (Table 2). Fractures were present in nine patients, joint space narrowing in two, and joint space fluid in two patients. A good correlation was seen between MRI class and the CT/MPR upgraded radiographic staging (Table 2). Class A was equivalent to Stage 1/2 while Classes C and D were equivalent to more advanced stages. No Class B lesions were seen on MRI. The remaining hips in these Group 1 patients were normal on all imaging modalities. Of the Group 2 asymptomatic patients, two of 14 hips were abnormal on MRI. These were bilateral Class A lesions on MRI in one patient (Fig. 1). In this patient, scintigraphy, CT and CT MPR were positive in one hip. All investigations in the other six patients were negative. DISCUSSION New therapeutic techniques, such as core decompres- sion and transtrochanteric rotational osteotomy, have become available in recent times for the treatment of osteonecrosis of the hip. Whilst insertion of a prosthetic hip joint or arthroplasty are required for Stage 5, the newer techniques can be used in earlier disease stages. Some authors suggest that Stages 1 and 2 should have intramedullary pressure measurements performed and, if these are elevated, then core decompression should be carried out (Solomon et al., 1981; Hungerford and Lennox, 1985). This technique involves coring a segment of bone from the femoral head. This core should traverse the osteonecrotic segment not only to decompress it, but also to provide a channel for revascularization (Camp and Colwell, 1986). Not all authors agree with this approach (Petty, 1986); however, most agree that early diagnosis is essential. Transtrochanteric rotational osteo- tomy involves identification of a structurally normal portion of the femoral head and articular cartilage, which is then surgically rotated to become the new weight- bearing surface. Accurate staging and high sensitivity in the early diagnosis of osteonecrosis are absolute prequi- sites for performing these new therapeutic techniques. In addition, precise spatial localization of osteonecrotie areas in the femoral head is essential in choosing and performing a particular surgical technique. It is clear that plain radiography and scintigraphy cannot provide accurate spatial localization of osteonec- rotic segments due to the superimposition of bony elements with both these modalities. In addition, plain radiography is an insensitive method for detecting early changes. Scintigraphy is thought to be more sensitive in detecting osteonecrosis; however, false negative radio- nuclide bone scans were present in three of five patients with Stage 1 or 2 disease (Table 1) in this study. This emphasizes the low sensitivity of scintigraphy in the diagnosis of early osteonecrosis which has also been reported by other authors. Coleman et al. (1988) reported a sensitivity of 50% for radionuclide bone scans for Class A lesions seen on MRI, while Bieber et al. (1985) found a false negative rate for scintigraphy of 18% in biopsy proven cases. CT/MPR changes included focal, cystic and sclerotic areas in conjunction with thickening of the primary stress trabeculae. The trabecular thickening was better appre- ciated on coronal and oblique reconstructions than on transaxiat images in this study (Fig. le). As with Magid et al. (1985), we also found that CT/MPR had advantages over transaxial CT in imaging the superior pole of the femur, the superior joint space and the dome of the acetabulum, because of the absence of partial volume averaging seen with transaxial images. CT/MPR was more sensitive than MRI in depicting femoral head fractures and joint space narrowing (Table 2). As a consequence, radiographic staging was upgraded in four patients by CT/MPR. One patient, with Stage 1 disease, demonstrated a CT crescent sign consistent with subchondral fracture which is not evident radiographi- cally or on MRI. Two patients with radiographic Stage 2 disease were upgraded to Stage 4 by the CT/MPR demonstration of fractures and undermining of the femoral head. A further patient with Stage 4 disease was upgraded to Stage 5 by the CT/MPR demonstration of joint space narrowing. This had important therapeutic implications in these patients, as differing surgical 4reat= ment modalities were then chosen. CT/MPR was positive overall in 14 of 15 hips but was normal in 1 of 5 patients with early osteonecrosis (Stages 1 and 2, Table 2)i MRI was positive in the 13 painful hips of Group 1 patients and in two asymptomatic hips of Group 2 patients. Abnormalities found on MRI in patients in this study included diffuse, irregular, low signal areas in the femoral head on T1 weighted sequences and/or focal, wedged-shaped areas of altered signal in the superior margin of the femoral head. These abnormalities were classified according to Classes A, B, C and D according to the method described by Mitchell et al. (1987). Early in the course of osteonecrosis repair and mecha- nical failure have not extended into the necrotic segment of bone. The fat signal is thus preserved in the avascular area except at the sclerotic reactive margin, which forms at the junction between necrotic and viable bone. This OSTEONECROSIS OF THE FEMORAL HEAD 431 constitutes a Class A lesion which has a high signal on T1 weighted sequences and an intermediate signal on T2 weighted sequences, surrounded by a low intensity ring (Fig. ~lf). Lang et al. (1988) in their radiological/patho- logical correlation study, showed that this low intensity ring represented the repaired tissue interface of mesen- chymal and fibrous tissue, amorphous cellular debris, and thickened trabecular bone adjacent to the necrotic area. As the reparative process begins to invade the necrotic area, the high central fat signal disappears. When there is sufficient inflammation or vascular engorgement, or a subacute haemorrhage occurs, a Class B (blood-like) MR pattern may result, with high signal intensity on both T1W and T2W sequences. As more inflammation, hyper- aemia and fibrosis occur, a Class C (fluid-like) appearance is observed. This has a low signal intensity on T1W sequences and high signal intensity on T2W sequences. Finally, as fibrosis and sclerosis predominate, a Class D (fibrous-like) lesion is seen with low signal intensity on T 1W and T2W sequences (Fig. 2e,d). There were no Class B lesions in this study. A good correlation was observed between the MRI class and CT upgrading of radiographic staging (Table 2). Generally, Class A correlated with Stages 1 and 2, and Classes C and D correlated with later stages. MRI was not as sensitive as CT/MPR in detecting femoral head fracture or joint space narrowing (Table 2). Fractures were observed in 10 patients on CT/MPR and nine patients on MRI. The one patient with a fracture not discernible on MRI had a Class A lesion (Patient 1, Table (a) (b) (0 (d) Fig. 2 - CT and MRI images from a patient with sickle cell anaemia and radiographic Stage 4 osteonecrosis. (a) CT/MPR: coronal reconstruction showing focal sclerotic area in the left femoral head with encystification (arrow) of an ischaemic area in the left femoral neck. Subchondrat fracture not well seen on this image. (b) CT/MPR: Oblique reconstruction demonstrating the subchondral fracture (arrow). (c) T1 weighted coronal MRI demonstrating a focal wedge-shaped area of low signal corresponding with the sclerotic area seen on coronal CT reconstruction. The cyst is present below this in the femoral neck with heterogeneous signal. Subchondral collapse is not well seen but disruption of the femoral head outline is visible supero-laterally (arrow). (d) T2 weighted coronal MRI showing persisting low signal in the wedge-shaped infarcted area, indicating a Class D (fibrous type) lesion. The cyst is of higher signal intensity on this T2 weighted image, confirming the presence of fluid within it. Subchondral collapse is more clearly seen (arrow). 432 CLINICAL RADIOLOGY 2). CT, however, upgraded this to a Stage 2 lesion by a demonstrat ion of a CT crescent sign. This had important therapeutic implications. MRI was the best modal i ty for detecting jo int fluid. Because o f the signal characteristics of water, jo int effusions were elegantly displayed. Two patients in this study had jo int effusions observed on MRI but not on CT. Whilst increased fluid can occur in the early stages of osteonecrosis and is often responsible for the develop- ment of symptoms (Coleman et al., 1988), jo int effusions were present in advanced disease in this study. Of the asymptomat ic high-risk patients (Group 2), only one patient had osteonecrosis. Bilateral focal changes were noted on MRI (Fig. 1) in the antero-super ior port ions of the femoral head. Scintigraphy, CT and CT/ MPR were positive in one hip (Fig. 1). Our study contrasts with that of Coleman et al. (1988) where an incidence o f 88% was observed in 16 asymptomat ic patients. A l though our group is small and statistical inferences cannot be drawn, Genez et al. (1988) have reported low sensitivity of MR! (46%) in detection of osteonecrosis compared with biopsy in high risk patients who were symptomatic. This is possibly due to the macroscopic anatomic response to cell death lagging behind the initial microscopic insult, with resultant delayed change in signal characteristics. CONCLUSION In this study, MRI proved to be the most sensitive technique in the diagnosis of osteonecrosis, being part icu- larly useful for the detection of early disease. Scint igraphy and CT/MPR were not as sensitive in detecting early changes but, CT /MPR was superior to MRI in the detection of fractures and jo int space narrowing; thus, in Stage 2 lesions and higher, the roles of MRI and CT are complementary. MRI is therefore recommended in the diagnosis of early osteonecrosis and CT/MPR can be performed on documented cases for staging prior to therapy. CT /MPR can also be performed where MRI is contra- indicated (as in one pat ient with a cardiac pace- maker in this study). REFERENCES Bieber, E, Hungerford, DS & Lennox, DW (1985). Factors in diagnosis of avascular necrosis of the femoral head. Advanced Orthopaedic Surgery, 9, 93-96. Camp, JF & Colwell, CW. (1986). Core decompressign of the femoral head for osteonecrosis. American Journal of Bone and Joint Surgery, 68A, 1313-1319. Coleman, BG, Kressel, HY, Dalinka, MK, Scheibler, ML, Burk, DL & Cohen, EK (1988). Radiographically negative avascular necrosis: detection with MR imaging. Radiology, 168, 525-528. Conkin, JJ, Alderson, PO & Zizic, TM. (1983). Comparison of bone scan and radiographic sensitivity in the detection of steroid induced ischaemic necrosis of bone. Radiology, 147, 221-226. Crness, RL (1978). Experience with steroid-induced avascular necrosis of the shoulder and aetiologic considerations regarding osteonecro. sis of the hip. Clinical Orthopaedics, 130, 86 93. Dihlmann, W (1982). CT analyses of the upper end of the femur: the asterisk sign ofischaemic bone necrosis of the femoral head. Skeletal Radiology, 8, 251-257. Genez, BM, Wilson, MR, Houk, RW, Weiland, FL, Unger, HR, Jr., Shields, NN et al. (1988). Early osteonecrosis of the femoral head: detection in high-risk patients with MR imaging. Radiology, 168, 521-524. Glickstein, MF, Burk, DL, Jr., Schiebler, ML, Cohen, EK, Murray, KD, Steinberg, ME et al. (1988). Avascular necrosis versus other diseases of the hip: sensitivity of MR imaging. Radiology, 169, 213- 215. Hungerford, DS & Lennox, DW (1985). The importance of increased intraosseous pressure in the development of osteonecrosis of the femoral head: implications for treatment. Orthopedic Clinics of North America, 16, 635-654. Jergensen, HE, HelleL M & Genant, HK (1985). Magnetic resonance imaging in osteonecrosis of the femoral head. Orthopedic Clinics of North America, 16, 705-715. Lang, BA, Jergensen, HW, Moseley, ME, Block, JE, Chafetz, NI & Genant, HK (1988). Avascular necrosis of the femoral head: high- field strength MR imaging with histologic correlation. Radiology, 169, 517-524. Magid, D, Fishman, EK, Scott, WW, Jr., Brooker, AF, Arnold, WP, Lennox DW, et al. (1985). Femoral head avascular necrosis: CT assessment with multiplanar reconstruction. Radiology, 157, 751- 756. Marcus, ND, Enneking, WF & Massam, RA. (1973). The silent hip in idiopathic septic necrosis. Journal of Bone and Joint Surgery, 55A, 1351 1366. Mitchell, DG, Kressel, HY, Arger, PH, Dalinka, M, Spritzer, CE & Steinberg, ME (1986a). Avascular necrosis of the femoral head: morphologic assessment by MR imaging, with CT correlation. Radiology, 161, 739-742. Mitchell, DG, Rao, VM, Dalinka, MK, Spritzer, CE, Alavi, A, Steinberg, ME et al. (1987). Femoral head avascular necrosis: correlation of MR imaging, radiographic staging, radionuclide imaging and clinical findings. Radiology, 162, 709-715. Mitchell, MD, Kundel, HL, Steinberg, ME, Kressel, HY, Alavi, A & Axel, L. (1986b). Avascular necrosis of the hip: comparison of MR, CT and scintigraphy. American Journal of Roentgenology, 147, 67- 71. Petty, W (1986). Editorial. Osteonecrosis. American Journal of Bone and Joint S~rgery, 68A, 1311-1312. Solomon, L (198 I). Idiopathic necrosis of the femoral head: pathogene- sis and treatment. Cancer Journal of Surgery, 24, 573-578. Thickman, D, Axel, L, Kressel, HY, Steinberg, M. Chen, H, Velchick, M et al. (1986). Magnetic resonance imaging of avascular necrosis of the femoral head. Skeletal Radiology, 15, 133 140.