Genitourinary Imaging
Original Research
An Intravoxel Incoherent Motion Diffusion-Weighted Imaging Study of Prostate Cancer
Audio Available | 
Share
OBJECTIVE. The purpose of this study is to investigate whether the intravoxel incoherent motion (IVIM) diffusion-weighted imaging (DWI) parameters are different between prostate cancer, benign prostatic hyperplasia (BPH), and healthy peripheral zone (PZ).
MATERIALS AND METHODS. Detailed diffusion measurements of 26 patients with histologically proven prostate cancer have been made in this retrospective study. Trace IVIM DWI was performed using 10 b values (0, 10, 20, 30, 50, 80, 100, 200, 400, and 1000 s/mm2). Biexponential fits were applied to diffusion decay curves to calculate molecular diffusion coefficient, perfusion-related diffusion coefficient, and perfusion fraction on the basis of the IVIM model. Decay curves were also fit with monoexponential decay functions, and a statistical comparison between mono- and biexponential fits was performed. Paired t tests were performed to evaluate the statistical significance of the parameters of IVIM DWI and apparent diffusion coefficient (ADC) between prostate cancer, BPH, and PZ.
RESULTS. The chi-square values of biexponential fits were smaller than those from monoexponential fits in all cases. Biexponential functions provided statistically improved fits over monoexponential functions in 81% of cases. The ADC, molecular diffusion coefficient, and perfusion fraction in prostate cancer were significantly lower than those found in the PZ; however, perfusion fractions in prostate cancer and BPH were not significantly different. There were no significant differences in the prostate cancer, BPH, and PZ for the perfusion-related diffusion coefficient, which had large SDs.
CONCLUSION. IVIM DWI parameters are significantly different between prostate cancer and PZ. IVIM DWI may offer additional information for tissue characterization in the prostate gland.
Keywords: biexponential fitting, diffusion, intravoxel incoherent motion, MRI, prostate cancer
Prostate cancer is the second most common cancer among men in the world, and the incidence of prostate cancer is still rising in some Asian and Eastern European countries [1]. The common diagnostic methods of prostate cancer are serum prostate-specific antigen (PSA), digital rectal examination, and transrectal ultrasound–guided biopsy; however, the cancer detection rate still remains unsatisfactory [2, 3]. MRI of the prostate provides excellent anatomic information and has been considered sensitive for prostate cancer detection. However, the specificity for prostate cancer on MRI is generally considered insufficient because of the presence of other diseases, including prostatitis or benign prostatic hyperplasia (BPH), which can mimic prostate cancer on T2-weighted imaging (T2WI).
Diffusion-weighted imaging (DWI) has been proven to improve prostate cancer detection. Apparent diffusion coefficient (ADC) values of prostate cancer are generally lower than those of normal prostate tissues, particularly in the peripheral zone (PZ). To date, most DWI studies for the prostate have used b factors within the 0–2000 s/mm2 range, with ADC values calculated assuming monoexponential decay functions of signal with b value [4–9]. High b values have been generally preferred for prostate cancer detection to minimize T2 shine-through and perfusion effects within the capillary networks. Intravoxel incoherent motion (IVIM) DWI, as originally described by Le Bihan et al. [10], utilizes low b values to sensitize DWI to motions of water within the capillary bed. According to the IVIM DWI model, both pure extravascular molecular diffusion and microcirculation of blood within the capillaries (perfusion) can be separated using a biexponential decay function, providing additional parameters for tissue characterization [10].
In this study, detailed diffusion measurements of normal prostate tissues, BPH, and prostate cancer using multiple b values have been made. Biexponential fits were applied to the diffusion decay curves to calculate the molecular diffusion coefficient, perfusion-related diffusion coefficient, and perfusion fraction on the basis of the IVIM DWI model. These parameters may offer additional information to characterize normal prostate tissues and prostate cancer.
| Materials and Methods |   |
|---|
Our institutional review board approved this retrospective study and deemed that patient informed consent was not required. Between September 2010 and July 2011, 155 consecutive patients with suspected prostate cancer with elevated PSA levels (> 4.0 ng/mL) underwent MRI, including IVIM DWI. The inclusion criteria of this study were obvious region of cancer on MRI (the diagnostic criteria are stated in detail later, in the Image Interpretation and Data Analysis subsection) and histopathology-revealed cancer in the same region; normal-appearing contralateral PZ on MRI and no cancerous tissue revealed by histopathology in the corresponding PZ area; and no hemorrhage seen on T1-weighted images. In addition, five patients were excluded from this study because MRI of these patients was considerably degraded as a result of large amounts of air and peristalsis of the rectum. The final study population comprised 26 patients (age range, 60–78 years; mean age, 67.3 years). The mean PSA level in these 26 patients was 13.7 ng/mL (range, 4.1–130.4 ng/mL). Prostate cancer was proven in 14 patients with transrectal ultrasound–guided biopsies and in 12 patients with radical prostatectomy. Gleason scores were 3 + 3 in five patients, 3 + 4 in 14 patients, 4 + 3 in six patients, and 4 + 5 in one patient.
 View larger version (161K) | Fig. 1A—70-year-old man with prostate cancer (prostate-specific antigen level, 5.4 ng/mL). A, Typical regions of interest (ROIs) were set within prostate cancer and peripheral zone (PZ) on diffusion-weighted image (b = 1000 s/mm2) (A) and corresponding T2-weighted image (B) and apparent diffusion coefficient (ADC) map (C). Prostate cancer is seen as hypointense lesion on T2-weighted image and ADC map (arrows, B and C), and hyperintense lesion is seen on diffusion-weighted image (A) in anterior part of left lobe. Normal hyperintense PZ on T2-weighted image is seen in right lobe. ROI within benign prostatic hyperplasia is not shown here. |
 View larger version (249K) | Fig. 1B—70-year-old man with prostate cancer (prostate-specific antigen level, 5.4 ng/mL). B, Typical regions of interest (ROIs) were set within prostate cancer and peripheral zone (PZ) on diffusion-weighted image (b = 1000 s/mm2) (A) and corresponding T2-weighted image (B) and apparent diffusion coefficient (ADC) map (C). Prostate cancer is seen as hypointense lesion on T2-weighted image and ADC map (arrows, B and C), and hyperintense lesion is seen on diffusion-weighted image (A) in anterior part of left lobe. Normal hyperintense PZ on T2-weighted image is seen in right lobe. ROI within benign prostatic hyperplasia is not shown here. |
 View larger version (54K) | Fig. 1C—70-year-old man with prostate cancer (prostate-specific antigen level, 5.4 ng/mL). C, Typical regions of interest (ROIs) were set within prostate cancer and peripheral zone (PZ) on diffusion-weighted image (b = 1000 s/mm2) (A) and corresponding T2-weighted image (B) and apparent diffusion coefficient (ADC) map (C). Prostate cancer is seen as hypointense lesion on T2-weighted image and ADC map (arrows, B and C), and hyperintense lesion is seen on diffusion-weighted image (A) in anterior part of left lobe. Normal hyperintense PZ on T2-weighted image is seen in right lobe. ROI within benign prostatic hyperplasia is not shown here. |
All examinations were performed with a 3-T MRI scanner (Achieva, Philips Healthcare) using a six-channel phased-array coil. The examinations included axial and coronal T2WI, IVIM DWI, and gadolinium-enhanced dynamic MRI. To prevent artifacts from bowel peristalsis, 1 mg of intramuscular glucagon (Glucagon G Novo, Eisai) was administered before imaging. T2WI was performed with TR/TE of 4238/70, 3.5-mm slice thickness with 0.1-mm gap, 10 echo-train length, 512 × 260 (zerofilled interpolation [ZIP] 1024) matrices, and two excitations. Dynamic MRI (enhanced T1-weighted high-resolution isotropic volume excitation) was performed with TR/TE of 3.8/1.9, flip angle of 15°, 3.0-mm (ZIP 1.5 mm) slice thickness, and 240 × 194 (ZIP 512) matrices. Unenhanced baselines and 25, 60, and 180 seconds after bolus injection of 0.1 mmol/kg of gadodiamide hydrate (Omniscan, Daiichi Sankyo) were sequentially obtained in dynamic MRI. IVIM DWI was performed before dynamic MRI with TR/TE of 5132/40, 3.5-mm slice thickness with 0.1-mm gap, 80 × 80 matrices, and two excitations. The diffusion weighting was performed along three orthogonal directions at 10 b values of 0, 10, 20, 30, 50, 80, 100, 200, 400, and 1000 s/mm2. A parallel imaging technique, sensitivity encoding, was used to reduce gradient-echo train lengths by a factor of 2. The acquisition time of IVIM DWI was 5 minutes 54 seconds.
MRI examinations were interpreted and analyzed by consensus of two experienced radiologists who were blinded to the patient’s clinical history and data. The diagnostic criteria of prostate cancer and noncancerous PZ (n = 26 each) were determined as follows. Prostate cancer was diagnosed when a mass lesion showed hypointensity on T2WI and early enhancement on the gadolinium-enhanced dynamic study in the region where biopsies or step-section pathologic maps revealed the existence of cancerous tissues. Noncancerous PZ was diagnosed when PZ showed hyperintensity on T2WI and biopsies or step-section pathology did not reveal cancerous tissue in the corresponding PZ area. In addition, BPH (n = 12) was selected from 12 patients who underwent radical prostatectomy. The diagnostic criteria of BPH were a round mass with distinct T2-hypointense capsule and step-section maps revealing BPH without cancerous tissue in the corresponding lesion.
Regions of interest (ROIs) were placed within proven prostate cancer, BPH, and PZ on the basis of previously discussed MRI and pathologic findings to calculate the parameters of IVIM DWI (Figs. 1A, 1B, and 1C). The ROIs were chosen to be as large as possible, consistent with minimal contamination from unintended tissues, and signal intensities were measured for each b value with a copy-paste operation. The acquired datasets were transferred to a PC, and data analysis was performed with inhouse software. The decay curves were fit with biexponential decay functions as follows:
 |
| Results | |
|---|
IVIM DWI datasets were successfully acquired from 26 patients. The chi-square values in the biexponential fits were smaller than those from monoexponential fits in all cases (all cases of cancerous tissue, n = 26; BPH, n = 12; and healthy PZ, n = 26); however, biexponential functions provided statistically improved fits over monoexponential functions in 81% (52/64) of the cases, as assessed by F tests. Figure 2 shows semilog plots of typical signal decays with b values for the healthy PZ, BPH, and the cancerous tissue of a patient. Figure 3 shows both monoexponential and biexponential curve fits from the cancerous tissue of a patient. These signal decay curves clearly indicate that the decay curves are well characterized with biexponential functions.
Table 1 shows a summary of IVIM DWI parameters obtained from prostate cancer, BPH, and PZ. The ADC and molecular diffusion coefficient of the cancer were significantly lower than those found in the BPH and PZ. The perfusion fraction in the prostate cancer was significantly lower than that in the PZ, but not in the BPH. There were no significant differences in the PZ, BPH, and prostate cancer for the perfusion-related diffusion coefficient. No significant correlation was found between the IVIM DWI parameters and Gleason scores.
| Discussion | |
|---|
Applying DWI to the body has become widespread because of technologic innovations such as parallel imaging techniques and improved MRI gradient performance. For oncologic applications, ADC value has emerged as a useful tool for discriminating malignant from benign lesions. Because malignant lesions typically have dense cellularity, they show lower ADC values compared with benign lesions [13, 14]. This could also be applied to prostate cancer, as several studies have indicated [4–9]. In this context, pure molecular diffusion is crucial, whereas perfusion-related diffusion is cumbersome and its elimination from contaminating the pure diffusion coefficient measurement should be considered.
 View larger version (29K) | Fig. 2 —67-year-old man with prostate cancer. Graph shows semilog plots of normalized signal intensity versus b value in patient’s healthy peripheral zone (PZ), benign prostatic hyperplasia (BPH), and cancer tissue. |
In 1988, Le Bihan et al. [10] described the IVIM model and indicated that pure molecular diffusion and perfusion-related diffusion might be separately evaluated using a biexponential fitting function. More recently, Luciani et al. [15] showed that restricted diffusion observed in cirrhotic livers could be related not only to restricted pure molecular diffusion but also to reduced perfusion by means of IVIM DWI. In the present study, we successfully performed IVIM DWI and found that signal decays from PZ, BPH, and prostate cancer showed biexponential behaviors, offering IVIM parameters of molecular diffusion coefficient, perfusion-related diffusion coefficient, and perfusion fraction. We also found that the IVIM parameters of molecular diffusion coefficient and perfusion fraction were significantly smaller in prostate cancer when compared with noncancerous PZ. These results are consistent with a previous report by Döpfert et al. [16], though they used only four b values for IVIM DWI characterization. We have presented here more-detailed measurements of IVIM DWI using 10 b values under 1000 s/mm2. Our results imply that a decreased perfusion fraction may play a role in observed reductions of ADC in prostate cancer, in addition to an actual decrease in molecular diffusivity, as Döpfert et al. also suggested.
An IVIM study of pancreatic cancer by Lemke et al. [17] found that perfusion fraction was the best parameter for differentiation between healthy pancreas and pancreatic cancer, with no significant difference in molecular diffusion coefficient between healthy tissue and pancreatic cancer. In the present study, both perfusion fraction and molecular diffusion coefficient were significantly smaller in prostate cancer than in noncancerous tissue. This difference in molecular diffusion coefficient may be due to the presence of abundant fibrotic components or other microstructural differences in pancreatic cancer compared with prostate cancer.
Regarding the physiologic basis of the IVIM parameters, Le Bihan and Turner [18] suggested that the molecular diffusion coefficient, perfusion-related diffusion coefficient, and perfusion fraction correspond with tissue diffusivity, capillary blood velocity, and fractional volume of capillary blood flowing, respectively, and that perfusion might be estimated from the product of perfusion fraction multiplied by perfusion-related diffusion coefficient. In contrast, Henkelman [19] argued that IVIM parameters do not measure classic perfusion, but rather blood volume transit. Franiel et al. [20] used dynamic contrast-enhanced MRI to show that high-grade prostate cancer had greater perfusion and blood volume compared with normal prostate tissue. If perfusion fraction is associated with blood volume, then their report is inconsistent with our results, in which perfusion fraction was found to be significantly smaller in prostate cancer than in noncancerous PZ. A possible reason for this discrepancy is that the perfusion fraction is not specific to perfusion only but also may be sensitive to other bulky flow phenomenon, such as glandular secretion and fluid flow in the ducts. Although the relationship between perfusion parameters measured with IVIM DWI and dynamic contrast-enhanced MRI still remains unclear, further studies are clearly required to elucidate the mechanisms underlying the “perfusion” parameters associated with either technique.
 View larger version (25K) | Fig. 3 —78-year-old man with prostate cancer. Graph shows monoexponential and biexponential curve fits from patient’s cancerous tissue. Noise values from air outside pelvis are also plotted for each b value. |
TABLE 1: Results of Intravoxel Incoherent Motion Parameters and Apparent Diffusion Coefficient (ADC)
Recently, it has been shown that biexponential diffusion decay functions were required to adequately describe healthy prostate gland and prostate cancer diffusion signal decay curves over an extended b value range up to 3000 s/mm2 [11, 21]. Shinmoto et al. [21] described that the fast and slow ADCs of prostate cancer over this b value range are significantly lower than those of noncancerous tissue and that the fraction of the fast diffusion component is significantly smaller in cancer than in PZ. Care should be taken to not confuse these previous reports with the current study, which used the small end of the b value range from 0 to 1000 s/mm2 using the IVIM type of biologic interpretation of fast (perfusion-related diffusion coefficient) and slow (molecular diffusion coefficient) components as being associated with perfusion and pure molecular diffusion, respectively. Meanwhile, the biologic interpretation of fast and slow ADCs calculated from diffusion studies in the high b value range remains unclear, though recent ultra–high-field studies have shed some light on the potential water compartments responsible for very slow diffusion coefficients observed above 2000 s/mm2 [22].
In our study, the perfusion-related diffusion coefficient was not significantly different between PZ, BPH, and prostate cancer, and there were large SDs in the estimation of perfusion-related diffusion coefficients. In addition, although chi-square values of biexponential fits were smaller than those of monoexponential fits in all cases, biexponential functions did not provide statistically improved fits over monoexponential functions in approximately 20% of cases. These results may be partially attributed to signal measurement errors at low b values, though we selected ROIs of the targets as large as possible to reduce such errors. Koh et al. [23] reported poor reproducibility of the perfusion-sensitive low ADC range (0–100 s/mm2). Another reason for large perfusion-related diffusion coefficient variations may be biexponential data fitting errors. In general, biexponential fitting is considered to be difficult to perform reliably [24], and further improvements in mathematic modeling of DWI signal decay behavior might provide improved curve fits and a better understanding of the physiologic basis of DWI.
Some limitations exist in our study. First, only 12 patients underwent radical prostatectomy, whereas the remaining patients, who were considered to have prostate cancer, were diagnosed by systemic biopsy alone. It is impossible to fully exclude the possibility that the patients who were not considered to have cancer by MRI and systemic biopsy may actually have had prostate cancer. In addition, because we selected BPH from the patients who underwent radical prostatectomy, only a small number of BPHs (n = 12) were included in our study. Second, MRI did not dovetail perfectly with the histologic sections provided by prostatectomy or did not correspond directly to the biopsy results. Third, in this study, we used 10 b values, though the appropriate number and choice of b values suitable for the prostate gland are still unknown. Döpfert et al. [16] used only four fixed b values in their IVIM DWI study; however, most multiexponential studies in abdominal organs used more low b values [15, 17, 25]. We think that, when the number of decay components is unknown, it is prudent to sample as dense and wide a range of b values as clinically feasible. Reduction in the number of b values and their selection would then be a secondary optimization procedure beyond the scope of the present work.
In conclusion, we have shown that the signal decay curves from PZ, BPH, and prostate cancer are biexponential within a b value range of 0–1000 s/mm2. In addition, the IVIM parameters of molecular diffusion coefficient and perfusion fraction are significantly lower in prostate cancer than that in PZ. The reduction of perfusion fraction value may imply that a decreased perfusion fraction may play a certain role in observed reductions of ADC values in prostate cancer, as well as decreased tissue diffusivity. IVIM DWI may offer additional information for tissue characterization in the prostate gland.
WEB
This is a Web exclusive article.
| References | |
|---|
