Diffusion-weighted MRI (DWI) in the oncology patient: Value of breathhold DWI compared to unenhanced and gadolinium-enhanced MRI
Abstract
Purpose
To evaluate the feasibility and added value of single breathhold diffusion-weighted (DW) imaging (DWI) in oncology patients undergoing abdominal MRI.
Materials and Methods
A total of 169 patients with malignancy underwent abdominal MRI at 1.5T, including T1-weighted (T1W), T2-weighted (T2W), and dynamic gadolinium-enhanced imaging. Axial DWI was performed with a single-shot spin-echo (SE) echo-planar imaging (EPI) sequence using a b-value of 500 seconds/mm2. A total of 24 slices were obtained during a 20-second breathhold. Two observers reviewed the conventional MR images for tumor. Next, the DW images were reviewed for additional tumor not depicted on conventional MR images
Results
For the 169 patients, additional tumors were noted on the DW images in 77 (0.46) for observer 1 and 67 (0.40) for observer 2. For observer 1 the additional tumor included lymphadenopathy (47), peritoneal metastases (15), renal (1), liver (12), and osseous (2), while for observer 2 the corresponding values were lymphadenopathy (40), peritoneal (12), renal (1), liver (6), osseous (4), and gastrointestinal (1). The DW images resolved as benign findings noted on the conventional MR images in three patients for observer 1 and four patients for observer 2. The conventional MR exam was entirely normal while the DW images showed tumor in 12 (0.07) patients for observer 1 and 10 (0.06) patients for observer 2.
Conclusion
DWI is feasible in a single breathhold and provides additional clinically important information in oncology patients when added to routine abdominal MR sequences. J. Magn. Reson. Imaging 2007. © 2007 Wiley-Liss, Inc.
DIFFUSION IS A PHYSICAL PROPERTY that describes the microscopic random movement of molecules in response to thermal energy. Also known as Brownian motion, diffusion may be affected by the biophysical properties of tissues such as cell organization and density, microstructure, and microcirculation. Diffusion-weighted (DW) imaging (DWI) utilizes pulse sequences and techniques that are sensitive to very small-scale motion of water protons at the microscopic level. Single-shot echo-planar imaging (EPI) DWI is utilized to provide very rapid imaging sensitive to subtle small-scale alternations in diffusion. Areas of restricted water diffusion are displayed as areas of high signal intensity (1-4).
The application of DWI for the evaluation of acute cerebral infarcts is well established. However, the challenges posed by DWI of the abdomen and pelvis have limited its application for body MRI. On DWI, artifacts related to physiologic motion, susceptibility, and chemical shift are compounded by inherent limitations in signal-to-noise ratio and image resolution. The larger fields of view used for abdominal imaging accentuate many of the artifacts inherent in DWI and single-shot EPI (1-4).
Hardware and technical advances in MRI have overcome many of these limitations, making DWI feasible for abdominal and pelvic MRI. Recent works have described the use of single-shot EPI DWI for evaluation of lymphadenopathy (5), liver tumors (6-8), renal masses (9-15), prostate cancer (16-18), and colorectal cancer (19). The routine use of DWI for abdominal and pelvic oncologic MRI has not been evaluated. We undertook this study to determine the feasibility and value of breathhold DWI when added to routine unenhanced and gadolinium-enhanced MRI in the oncology patient.
MATERIALS AND METHODS
Beginning in March 2004, DWI was added to the routine abdominal and pelvic MR exam for all patients imaged at our institution. Between March and October, 2004, 169 consecutive patients (71 men and 98 women; mean age 62.3 years.) with known malignancy were referred for abdominal or abdominal and pelvic MRI. This study represents the retrospective review of these 169 MR examinations. This retrospective study was approved by our institution's Investigational Review Board, which waived requirements for informed consent. Health Insurance Portability and Accountability Act (HIPAA) compliance was established. All records had identifying information removed to protect patient confidentiality.
Malignancies included colorectal cancer (27), lymphoma (27), ovarian cancer (24), hepatocellular cancer (14), breast cancer (11), renal cell cancer (10), lung cancer (8), pseudomyxoma peritonei (7), gastric cancer (4), bladder cancer (4), sarcoma (5), pancreatic cancer (4), melanoma (4), gastrointestinal (GI) stromal tumors (GIST) (3), prostate cancer (3), neuroendocrine tumor (4) testicular cancer (2), and one case each of primary tumors of the esophagus, urothelium, leukemia, islet cells, gallbladder, and endometrium, as well as a mesothelioma and a schwannoma.
MRI
All 169 patients underwent abdominal MRI at 1.5T (Philips Intera, Best, The Netherlands) (30 MT/m, 150 MT/m/second) with a phased-array surface coil or quadrature body coil. Parallel imaging was utilized for all surface coil acquisitions with a sensitivity encoding (SENSE) factor of 2. Breathhold T1 gradient-echo, and fat-suppressed T2-weighted (T2W) imaging was performed followed by dynamic gadolinium-enhanced three-dimensional (3D) T1-weighted (T1W) high-resolution isotropic volume exam (THRIVE) imaging, and delayed high-resolution 2D gradient-echo imaging. Specific imaging parameters for surface coil and body coil imaging are shown in Table 1.
| Sequence | TR | TE | Matrix size | NEX | Thickness | Gap | Fat saturation | Plane | Sense | # Slices | Time |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Surface coil | |||||||||||
| SSTSE | Shortest | 80 | 512 × 192 | 0.65 | 8 | 2 | No | Coronal | 2 | 25 | 0:11 |
| T1 FFE | 172 | 4.6/2.3 | 512 × 192 | 1 | 7 | 3 | No | Axial | 2 | 24 | 0:11 |
| T2 SSTSE | 650 | 80 | 512 × 192 | 1 | 7 | 3 | SPIR | Axial | 2 | 24 | 0:16 |
| SS EPI DWI | 2650 | 58 | 256 × 115 | 0.6 | 7 | 3 | SPIR | Axial | 2 | 24 | 0:20 |
| Gadolinium injection | |||||||||||
| 3D THRIVE | 3.5 | 1.7 | 512 × 192 | 1 | 4 | 2-mm overlap | SPIR | Axial | 2 | 100 | 0:21 |
| 2D FFE | 144 | 5 | 512 × 228 | 1 | 10 | 0 | Proset | Axial | 2 | 12 × 2 | 0:17 |
| Body coil | |||||||||||
| SSTSE | Shortest | 80 | 256 × 204 | 0.65 | 8 | 2 | No | Coronal | na | 25 | 0:20 |
| T1 FFE | 172 | 4.6/2.3 | 512 × 192 | 1 | 7 | 3 | No | Axial | na | 24 | 0:22 |
| RT TSE | 1600 | 70 | 512 × 204 | 2 | 7 | 3 | SPIR | Axial | na | 30 | 3:28 |
| SS EPI DWI | 2650 | 58 | 256 × 115 | 0.6 | 7 | 3 | SPIR | Axial | na | 24 | 0:20 |
| Gadolinium injection | |||||||||||
| 3D THRIVE | 3.5 | 1.7 | 512 × 192 | 1 | 6 | 3-mm overlap | SPIR | Axial | na | 50 | 0:21 |
| 2D FFE | 144 | 5 | 512 × 228 | 1 | 10 | 0 | Proset | Axial | na | 8 × 3 | 0:17 |
- NEX = number of excitations, SSTSE = single shot turbo spin echo, FFE = fast field echo, THRIVE = T1W high-resolution isotropic volume exam, RT = respiration triggered, SPIR = selective partial inversion recovery, Proset = selective water excitation, na = not applicable.
DW images were obtained prior to gadolinium injection in the axial plane using a single-shot spin-echo (SE) EPI sequence with SPIR fat suppression (TR = 2650, TE = 58, flip angle = 90, number of excitations (NEX) = 2, matrix size = 256 × 115, half-scan factor 0.6), using a b-value of 500 second/mm2. A total of 24 slices were obtained during a 20-second breathhold with slice thickness of 7 mm and interslice gap of 1 mm. A SENSE factor of 2 was applied when scanning with a phased-array surface coil. Gradient overplus was utilized, combining the three gradients (x, y, z) into a single vector producing a single stronger gradient with shorter TEs and less susceptibility. The final DW image was thus a composite image of three separate images with b = 500 in three orthogonal planes.
MR Image Analysis
All MR images were reviewed on a Picture ArChiving System (PACS) workstation independently and without conference by two experienced body MR radiologists with 16 years and eight years of experience in interpreting body MR examinations. First the conventional MR examination including T1W and T2W images and gadolinium-enhanced MR images, were reviewed and interpreted in the standard fashion. All sites of abdominal or pelvic tumor were recorded. Multiple anatomic sites were evaluated including the liver, spleen, kidneys, pancreas, biliary tree and gallbladder, adrenal glands, lymphadenopathy, peritoneum and omentum, mesentery, gastrointestinal tract, uterus, bladder, ovaries, osseous structures, and lung bases. Lymph nodes were evaluated strictly on the criteria of size and number of identified lymph nodes.
Next, each observer independently reviewed the DW images in conjunction with the conventional MR examination. Additional sites of tumor depicted on the DW images were recorded by each observer.
The quality of the DW images was categorized as excellent if there were no significant artifacts degrading image quality; adequate if artifacts were present without interfering with image interpretation; and poor if artifacts significantly degraded image quality.
Establishing Truth on DW Images
Proof of diagnosis was obtained by surgical exploration in 24 patients, biopsy and imaging studies in 25 patients, concurrent imaging examinations in 15 patients (helical computed tomography [CT]: nine patients; ultrasound: one patient; and positron emission tomography [PET]: five patients), and follow-up cross-sectional imaging studies in 92 patients. The mean time to follow-up imaging examination was 12 months (range 3–23 months). In 13 patients there was clinical evidence of tumor presence or absence established by serum tumor markers and physical examination. For these 13 patients, the conventional MR and DW images were in agreement. In 10 of the 13 patients, both the conventional MR images and DW images were without evidence of tumor. In the remaining three patients both exams showed identical tumors.
Additional tumor noted on the DW images was accepted as true if proven by surgical and histopathologic findings, concurrent imaging examinations, or by follow-up MR examinations confirming development of tumor at the same anatomic site predicted on the original DW images. Other tumors depicted on DW images was accepted as true if both observers concurred that tumor was in retrospect visible on the unenhanced T1W, T2W, or gadolinium-enhanced images. If proof of tumor was not established by any of these means, findings were considered falsely positive.
DW images depict both benign and malignant lymphadenopathy. Lymph nodes were accepted as malignant if proven by surgery and histopathologic evaluation, biopsy, or if follow imaging studies showed progressive nodal tumor with increasing size and number of lymph nodes. Obvious nodal masses (>3 cm) in patients with a history of malignancy were also accepted as evidence of tumor.
Statistical Analysis
The McNemar test of correlated proportions was used to assess the depiction of tumor at the anatomic sites on conventional MR images and the DW images. Two-tailed P-values are reported. The null hypothesis was rejected for P < 0.05. Interobserver agreement was assessed with a Kappa analysis.
RESULTS
DW image quality was rated as excellent in 144 patients for observer 1 and 122 patients for observer 2, acceptable in 22 and 44, and poor in three patients for both observers. The three examinations with poor DW image quality were degraded by susceptibility artifact. Artifacts on the DWI were related to susceptibility, which was most noticeable at the lung bases and adjacent to air-filled segments of bowel. The image quality was diagnostic in all except the three patients with poor DW image quality. There was no significant difference between the image quality for the DW images obtained with the SENSE body surface coil and the large body coil.
All Patients
A total of 114 of the 169 examinations showed evidence of tumor on the conventional MR images and/or the DW images. The remaining 55 examinations showed no tumor on either the conventional MR images or on the DW images.
Table 2 illustrates the depiction of tumor at different anatomic locations on DWI compared to conventional MRI. This analysis does not consider the number of tumors depicted at each location. Rather, the depiction of any tumor at each location was considered positive. For instance the depiction of a single liver tumor was scored the same as the depiction of multiple liver tumors. The DW images were superior for depicting tumor involving lymph nodes (P < 0.01) (Fig. 1). For the other anatomic locations there was no difference in per patient tumor depiction on the DW images and the conventional MR images.
| Tumor location | Observer 1 | Observer 2 | ||||||
|---|---|---|---|---|---|---|---|---|
| MRI | DWI | MRI | DWI | |||||
| TP | FN | TP | FN | TP | FN | TP | FN | |
| Liver | 27 | 4 | 29 | 2 | 28 | 3 | 30 | 1 |
| Lymph nodes | 36 | 25 | 56 | 5 | 41 | 20 | 53 | 8 |
| Pancreas | 6 | 0 | 6 | 0 | 5 | 1 | 5 | 1 |
| Gallbladder | 1 | 0 | 1 | 0 | 1 | 0 | 1 | 0 |
| Kidneys | 3 | 1 | 4 | 0 | 3 | 1 | 4 | 0 |
| Peritoneum | 24 | 3 | 25 | 2 | 27 | 0 | 26 | 1 |
| GI tract | 7 | 0 | 7 | 0 | 4 | 3 | 4 | 3 |
| Bones | 7 | 4 | 9 | 2 | 9 | 2 | 11 | 0 |
| Adrenal | 2 | 0 | 2 | 0 | 2 | 0 | 2 | 0 |
| Pelvis | 3 | 1 | 3 | 1 | 4 | 0 | 4 | 0 |
| Lungs | 2 | 1 | 2 | 1 | 2 | 1 | 2 | 1 |
| Totals | 118 | 39 | 144 | 13 | 126 | 31 | 142 | 15 |
| Normal exams | 51 | 52 | 46 | 49 | ||||
| False + | 4 | 3 | 9 | 6 | ||||
- * Numbers represent the number of patients in whom tumor at each anatomic location was depicted on conventional MRI alone compared to DWI plus conventional MRI.
- TP = true positive, FN = false negative, GI = gastrointestinal.
A 64-year-old woman with lung cancer. T2W image (a), arterial (b), and delayed (c) gadolinium-enhanced 3D THRIVE images, and DW image (d) show a 2.5-cm retroperitoneal nodal metastasis (arrows). The marked conspicuity of lymphadenopathy on the DW image improved observer confidence for the presence of nodal metastases. e: DW image from a follow-up MR examination obtained three months later shows marked progression of bulky retroperitoneal nodal metastases and new diffuse liver metastases.
Table 3 illustrates the additional tumor noted on the DW images compared to the conventional MR images alone. In this analysis, depiction of individual tumors was compared. Additional tumors were noted on the DW images 77 (0.46) patients for observer 1 and 67 (0.40) patients for observer 2 (P < 0.001) (Fig. 2). For observer 1 the additional sites of tumor included lymphadenopathy (47), peritoneal metastases (15), renal (1), liver (12), and osseous (2), while for observer 2 the corresponding values were lymphadenopathy (40), peritoneal (12), renal (1), liver (6), osseous (4), and gastrointestinal (1) (Fig. 3). In three patients for observer 1 and four patients for observer two the DW images resolved as benign findings noted on the conventional MR images. Clinical and imaging follow-up confirmed absence of tumor in these patients.
| Primary tumor | Obs | DWI added tumor | Lymph nodes | Peritoneal | Renal | Liver | GI | Bone | Resol FP | |
|---|---|---|---|---|---|---|---|---|---|---|
| Yes | No | |||||||||
| Lymphoma | 1 | 17 | 10 | 17 | ||||||
| 2 | 13 | 14 | 12 | 1 | ||||||
| Ovarian CA | 1 | 12 | 12 | 5 | 6 | 1 | ||||
| 2 | 11 | 13 | 4 | 5 | 1 | 1 | ||||
| Pseudomyxoma | 1 | 5 | 2 | 1 | 4 | |||||
| Peritonei | 2 | 3 | 4 | 1 | 2 | |||||
| Hepatocellular | 1 | 10 | 4 | 5 | 5 | |||||
| Cancer | 2 | 8 | 6 | 6 | 2 | |||||
| Lung CA | 1 | 2 | 6 | 1 | 1 | |||||
| 2 | 2 | 6 | 1 | 1 | ||||||
| Renal cell | 1 | 1 | 9 | 1 | ||||||
| 2 | 1 | 9 | 1 | |||||||
| Bladder CA | 1 | 1 | 3 | 1 | ||||||
| 2 | 1 | 3 | 1 | |||||||
| Pancreatic CA | 1 | 4 | 0 | 3 | 1 | |||||
| 2 | 2 | 2 | 1 | 1 | ||||||
| Gastric CA | 1 | 3 | 1 | 1 | 2 | |||||
| 2 | 3 | 1 | 3 | 1 | ||||||
| Melanoma | 1 | 1 | 3 | 1 | ||||||
| 2 | 2 | 2 | 1 | 1 | ||||||
| GIST | 1 | 1 | 2 | 1 | ||||||
| 2 | 2 | 1 | 1 | 1 | ||||||
| Breast CA | 1 | 2 | 9 | 1 | 1 | |||||
| 2 | 3 | 8 | 1 | 1 | 2 | |||||
| Colorectal CA | 1 | 8 | 19 | 6 | 2 | 1 | ||||
| 2 | 6 | 21 | 3 | 1 | 1 | 1 | ||||
| Sarcoma | 1 | 4 | 1 | 2 | 1 | 1 | ||||
| 2 | 3 | 2 | 1 | 1 | 1 | |||||
| Prostate CA | 1 | 1 | 1 | 1 | ||||||
| 2 | 1 | 1 | 1 | |||||||
| Neuroendocrine | 1 | 1 | 3 | 1 | ||||||
| 2 | 1 | 3 | 1 | |||||||
| Assorted CA | 1 | 4 | 7 | 4 | ||||||
| 2 | 5 | 6 | 5 | |||||||
| Totals | 1 | 77 | 92 | 47 | 14 | 1 | 12 | 0 | 2 | 3 |
| 0.46 | 0.28 | 0.09 | 0.01 | 0.07 | 0 | 0.01 | 0.02 | |||
| Totals | 2 | 67 | 102 | 40 | 12 | 1 | 6 | 1 | 4 | 4 |
| 0.41 | 0.24 | 0.07 | 0.01 | 0.04 | 0.02 | 0.02 | 0.02 | |||
- * Numbers represent the number of patients for each observer in which the DW images showed additional tumor when added to the conventional MR examination for different primary malignancies or cases in which the DW images resolved as benign findings suspected as tumor on the conventional MR images. The location of additional tumor is categorized as lymph nodes, peritoneal, renal, liver, gastrointestinal, and bone. In some patients the DW images depicted additional tumor in more than one category.
- Obs = observer number, Resol FP = false positive, CA = cancer, GI = gastrointestinal, GIST = gastrointestinal stromal tumors.
A 48-year-old woman with a neuroendocrine tumor. T1W (a) and T2W (b) images show a solitary metastasis (arrow) in the right hepatic lobe. Arterial phase (c) gadolinium-enhanced 3D gradient echo (GE) images show the hypervascular mass (arrow) with washout on delayed gadolinium-enhanced 3D GE image (arrow) (d). e: DW image shows multiple liver metastases (arrows) and an osseous metastasis (long arrow) in the thoracic spine.
A 57-year-old man with a gastrointestinal stromal tumor. a: T2W MR image shows a few anterior abdominal peritoneal tumors (arrows). Gadolinium-enhanced 3D GE (b) and delayed 2D spoiled gradient echo (SGE) (c) MRI images show confluent anterior abdominal soft tissue (arrows) that is difficult to distinguish from adjacent collapsed bowel. d: DWI image more clearly depicts multiple anterior abdominal peritoneal metastases (arrows) with marked tumor conspicuity.
The conventional MR exam was entirely normal while the DW images showed tumor in 12 (0.07) patients for observer 1 and 10 (0.06) patients for observer 2. In these patients, tumor was proved by concurrent or follow-up cross-sectional imaging, surgery, and biopsy, or by retrospective review of the conventional MR examination confirming tumor presence.
Table 3 also shows the added information provided by the DW images for each of the types of malignancy included in this study. For patients with lymphoma, ovarian cancer, pseudomyxoma peritonei, gastric cancer, pancreatic cancer, hepatocellular cancer, and sarcomas; the DW images showed additional tumors in more than 40% of patients. The most common sites of additional tumor included lymphadenopathy, peritoneum, and liver.
Additional Tumor on DW Images at Specific Anatomic Locations
Tumor Involving Lymph Nodes
The marked conspicuity of lymph nodes on the B500 DW images facilitated depiction of lymph nodes of all sizes (Figs. 1, 4, 5, and 6). The lymph nodes that were prospectively missed on conventional MR images, but depicted on the DW images varied in size from 1 to 3 cm. The percentage of exams in which the DW images depicted more sites of nodal metastases for patients with lymphoma was 63% for observer 1 and 44% for observer 2. For other malignancies the corresponding percentages of increased nodal tumor depiction on DW images were as follows: ovarian cancer, 21% for observer 1 and 17% for observer 2; hepatocellular cancer, 36% for observer 1 and 43% for observer 2; pancreatic cancer, 75% for observer 1 and 25% for observer 2; and colorectal cancer, 22% for observer 1 and 11% for observer 2 (Table 3).
A 55-year-old man with esophageal cancer. a: Coronal single shot turbo spin echo (SSTSE) image shows a mass (arrow) at the gastroesophageal junction. T1W (b) and T2W (c) images through the upper abdomen are normal without evidence of lymphadenopathy. Arterial phase (d) and portal venous phase (e) gadolinium-enhanced 3D GE images were prospectively interpreted as not showing lymphadenopathy. f: DW image depicts a 1.5-cm celiac lymph node. In retrospect, this node may be visible on (e; arrow) but is difficult to separate from adjacent structures. Restricted diffusion of the spleen is noted and is likely physiologic. Surgery and histopathologic exam confirmed a nodal metastasis from primary esophageal cancer.
A 64-year-old man with upper abdominal pain. T1W (a) and T2W (b) images show collapsed loops of small bowel without a definite mass. Arterial (c) and delayed (d) gadolinium-enhanced 3D THRIVE images show a questionable hypointense mass (arrows). e: DW image depicts a 4-cm pancreatic cancer. f: DW image caudal to (e) shows peripancreatic nodal metastases (long arrows) and peritoneal metastases (short arrows). The size of the pancreatic cancer and metastatic tumor was best demonstrated on the DW images. Findings were confirmed by biopsy.
A 55-year-old man with hematochezia. a: T1W image shows incomplete distention of the rectosigmoid colon. It is difficult to distinguish collapsed bowel, from retained stool and mural thickening. b: Gadolinium-enhanced 3D GE MR image shows enhancement of the thickened rectosigmoid colon (arrow); c: DW image confirms rectal cancer with mural thickening. The marked high signal intensity of the cancer is characteristic on DW images. d: A contiguous DW image shows perirectal lymph nodes (arrows). Surgery and histopathologic examination confirmed a T3N1 rectal cancer.
Peritoneal Tumor
On the DW images, ascitic fluid was low intensity and fluid within bowel showed partial suppression, increasing the conspicuity of adjacent peritoneal and bowel serosal tumor (Fig. 3). The DW images depicted more sites of peritoneal tumor for patients with ovarian cancer in 29% of patients for observer 1 and 21% for observer 2. For patients with pseudomyxoma peritonei the DW images showed more peritoneal tumors in 57% of patients for observer 1 and 29% for observer 2. Both observers noted additional peritoneal tumor in a single patient with pancreatic cancer, a GIST, a sarcoma, and breast cancer. Observer 2 only noted additional peritoneal tumor in a single patient with colorectal cancer (Table 3).
Liver Tumors
The number of additional liver metastases depicted on the DW images varied from one lesion to 30 lesions (Fig. 2). The DW images depicted more hepatic tumor for patients with hepatocellular cancer in 36% of patients for observer 1 and 14% for observer 2. Both observers noted a single case in which the DW images showed additional liver metastases in patients with melanoma, and neuroendocrine tumor. Observer 1 only noted additional liver metastases in two patients with colorectal cancer, two with gastric cancer, and a single patient with lung cancer. Observer 2 also noted additional liver metastases in a patient with gastric cancer, and another patient with ovarian cancer (Table 3).
GI Tract Tumors
Primary tumors of the GI tract involving the colon, rectum, esophagus, and stomach were very conspicuous on DW images, facilitating distinction of the tumor from collapsed bowel or retained stool. In all cases the primary tumor was depicted on the conventional MR images but was typically much more conspicuous on the DW images (Fig. 6).
Patients With Surgical or Biopsy Proof of Diagnosis
For the 24 patients with surgical proof of tumor, the DW images depicted additional tumor in 12 patients (0.50) for both observers, including seven patients each with additional tumor involving lymph nodes and five with additional peritoneal tumor. In one patient with primary esophageal cancer, only the DW images correctly showed 1.5-cm celiac nodal metastases, representing the only site of metastatic tumor (Fig. 4).
For 25 patients with proof by biopsy and follow-up imaging studies, the DW images depicted additional tumor in 18 patients (0.72) for observer 1 and 14 patients (0.56) for observer 2. Additional tumor included nodal metastases in 10 patients for observer 1 and nine patients for observer 2, liver metastases in five patients for observer 1 and two patients for observer 2; peritoneal tumor in one patient for both observers; and osseous metastases in one patient for both observers. For observer 2 only the DW images depicted an additional renal tumor in one patient.
For both observers in two patients with primary gastrointestinal cancer, nodal metastases were only depicted on the DW images. In one patient with biliary obstruction, the obstructing portal mass was only depicted on the DW images. In one patient with focal liver metastasis on the conventional MRI, the DW images showed diffuse liver metastatic tumor. For observer 1 in one patient with a normal conventional MRI, the DW images demonstrated multiple liver metastases. In the final patient, the conventional MRI demonstrated a heterogeneous liver suspicious for liver metastases while the DW images were normal. Biopsy confirmed absence of liver tumor.
Interobserver Agreement
Kappa analysis showed good interobserver agreement for determining the presence or absence of tumor at the individual anatomic sites. For the conventional MR images the two observers agreed in 1814 of 1859 observations, kappa = 0.80. For the combined conventional MR images and DW images the two observers agreed in 1828 of 1859 observations, kappa = 0.86.
DISCUSSION
Our study confirms that DWI is feasible for abdominal and pelvic MR imaging and can be implemented as a breathhold single-shot acquisition (3). Images obtained with a surface coil or with the quadrature body coil were of diagnostic image quality. Some artifact from susceptibility effect was present on many examinations near the lung base. However, this artifact did not interfere with overall interpretation of the DW images.
We found that the DW images were difficult to interpret by themselves. It was much easier to establish anatomic landmarks when the DW images were interpreted in conjunction with the conventional T1W, T2W, and gadolinium-enhanced images. This allowed the observers to dismiss areas of T2 “shine-through” from bowel contents and other long-T2 substances. The anatomic images also provided landmarks to identify high-signal peritoneal, serosal, mesenteric, and nodal tumor.
It is important to emphasize that our study did not look at DWI alone, but rather at the addition of DW images to the conventional MR examination. However, the conspicuity of many types of tumor on the DW images can assist the radiologist in body MR interpretation. Current abdominal and pelvic MR exams now have almost 1000 images to review. The sheer volume of images presented to the radiologist can be challenging. The DW images help direct the attention of the radiologist to findings that may otherwise be overlooked. In retrospect, these sites of tumor depicted on the DW images are often seen on the conventional MR images.
The addition of the DWI to the conventional MR images clearly improved the depiction of lymphadenopathy (5). It is likely that inflammatory and malignant lymph nodes will show restricted water migration and high signal on the DW images. However, the sensitivity of the DW images for nodal metastases has made DWI the primary imaging sequence for depicting lymphadenopathy at our institution. While other authors have advocated postprocessing the DW data to generate video-inverted maximum image projection (MIP) models (5), we believe that an analysis of the source DW images is preferable.
The DW images were also useful to depict additional tumor involving the peritoneum, bowel serosa, omentum, and mesentery. In patients with small-volume diffuse peritoneal and serosal tumor, interpretation of gadolinium-enhanced images can be challenging. On DW images, serosal and mesenteric tumor will show high signal intensity while most bowel contents will be of low signal intensity. When comparing the B0 and B500 images, one will see a reversal of signal intensity. On the B0 images the bowel contents are bright, while adjacent tumor is dark, while on the B500 images the tumor is bright and the bowel contents are dark. Corresponding enhancement of the thickened bowel wall and mesentery on the fat-suppressed gadolinium-enhanced images is present. The DW images are thus used to direct the radiologist's attention on the conventional MR images to areas of suspected tumor. Areas that are entirely normal in appearance on the conventional MR images are dismissed as artifact or T2 shine-through on the DW images.
Tumors involving the solid organs such as the liver, kidneys, and pancreas were also depicted as areas of high signal on the DW images (6-15). In selected cases the DW images showed additional tumor involving these solid abdominal organs. For hepatic DWI some authors have advocated using a lower B value of 20 seconds/mm2 to create a black-blood DWI technique (20). This lower B value will suppress the high signal from intrahepatic vessels while combining the effects of DW weighting and prolonged T2 relaxation of liver tumors. A lower B value will also improve the image signal-to-noise ratio on the DW images. We currently use a B value of 20 seconds/mm2 for DWI of the liver and a B value of 500 seconds/mm2 for extrahepatic DWI.
False-negative DW images occurred in 13 patients for observer 1 and 15 patients for observer 2. The false-negative DW exams included cases in which the conventional MR images depicted tumor that was not seen on the DW images. False-positive DW images occurred in three patients for observer 1 and five patients for observer 2. In these cases the DW images depicted findings that were believed to represent tumor. However, follow-up MR exams showed interval stability over an extended period of time, indicating a benign etiology.
Limitations of our feasibility study should be acknowledged. In this retrospective study we did not have pathologic proof of tumor for all anatomic sites in every patient. However, we only accepted the positive findings of tumor on the DW images if they were proven by surgery or biopsy, concurrent or follow-up cross-sectional imaging studies, or if in retrospect the conventional MR images confirmed tumor at these locations. A prospective study limited to patients with surgical proof would be useful for confirmation, but was beyond the scope of this early study of the value of DWI in the oncology patient.
In conclusion, DWI provides additional information when added to conventional MRI in the oncology patient. The DW images helped the radiologist direct his attention to suspicious areas when combined with conventional MR imaging. While our study did not evaluate DW images alone, we found that the combination of DW images and conventional MR images depicted additional sites of abdominal and pelvic tumor and improved the radiologist's confidence in image interpretation. At our institution DWI is now routinely performed for all oncology patients referred for abdominal and pelvic MR imaging.
