In vivo measurement of the apparent diffusion coefficient in normal and malignant prostatic tissues using echo-planar imaging

Authors

  • Bashar Issa PhD

    Corresponding author
    1. Centre for MRI, Hull Royal Infirmary, Anlaby Road, Hull HU3 2JZ, England
    • Department of Physics, Faculty of Science, UAE University, P.O. Box 17551, Al-Ain, United Arab Emirates
    Search for more papers by this author

Abstract

Purpose

To measure for the first time the apparent diffusion coefficient (ADC) values in anatomical regions of the prostate for normal and patient groups, and to investigate its use as a differentiating parameter between healthy and malignant tissue within the patient group.

Materials and Methods

Single-shot diffusion-weighted echo-planar imaging (DW-EPI) was used to measure the ADC in the prostate in normal (N = 7) and patient (N = 19) groups. The spin-echo images comprised 96 × 96 pixels (field of view of 16 cm, TR/TE = 4000/120 msec) with six b-factor values ranging from 64 to 786 seconds/mm2.

Results

The ADC values averaged over all patients in non-cancerous and malignant peripheral zone (PZ) tissues were 1.82 ± 0.53 × 10−3 (mean ± SD) and 1.38 ± 0.52 × 10−3 mm2/second, respectively (P = 0.00045, N = 17, paired t-test). The ADC values were found to be higher in the non-cancerous PZ (1.88 ± 0.48 × 10−3) than in healthy or benign prostatic hyperplasia central gland (BPH-CG) region (1.62 ± 0.41 × 10−3). For the normal group, the mean values were 1.91 ± 0.46 × 10−3 and 1.63 ± 0.30 × 10−3 mm2/second for the PZ and CG, respectively (P = 0.011, N = 7). Significant overlap exists between individual values among all tissue types. Furthermore, ADC values for the same tissue type showed no statistically significant difference between the two subject groups.

Conclusion

ADC is quantified in the prostate using DW-EPI. Values are lower in cancerous than in healthy PZ in patients, and in BPH-CG than PZ in volunteers. J. Magn. Reson. Imaging 2002;16:196–200. © 2002 Wiley-Liss, Inc.

THE EFFECT OF RANDOM motion of diffusing molecules on the nuclear magentic resonance (NMR) signal has been known for a long time (1). Early applications of diffusion-weighted (DW) magnetic resonance imaging (DW-MRI) on humans concentrated on imaging the brain. The mapping of apparent diffusion coefficient (ADC) on the brain (2, 3) has gained from recent advances in hardware improvement and the application of fast imaging methods (4, 5). DW-MRI has proved capable of following early changes in cerebral ischemia in both humans and animals (6–8).

In cancer, DW-MRI has been used to study recurrence in malignant and atypical meningiomas, and to differentiate between different types of meningiomas using ADC values (9). ADC values were also used to distinguish high-grade glioma from normal tissue (10). Recent hardware advances have allowed the application of DW-MRI to anatomical regions other than the brain. The ability of high speed echo-planar imaging (EPI) to minimize artifacts due to respiratory motion has extended the application of DW-MRI to the breast region (11), aiming to differentiate between tumor and normal tissues. DW-EPI was also used to study the contents of cystic ovarian lesions (12), and ADC values may be useful in evaluating small to medium cystic ovarian lesions.

Tests used to screen for prostate cancer include digital rectal exam (DRE) and monitoring of prostate specific antigen (PSA), which is elevated in nodular hyperplasia, prostatitis, and prostate adenocarcinoma. Diagnosis of prostate malignant disease relies on histological examination of a transrectal ultrasound (TRUS)-guided needle biopsy sample. However, the above techniques may miss the area of abnormality, and therefore, new techniques are being sought to provide spatially related information in vivo. MRI is currently unable to differentiate consistently between benign prostatic hypertrophy (BPH), which is prevalent in the central gland (CG) region (especially at advanced age), and early malignancy, or between those cancers that will progress and metastasize and those that will remain latent (13). Liney et al (14) have shown that high spatial resolution T2 mapping and the quantification of citrate concentration using NMR spectroscopy may aid in the discrimination of BPH, tumor, and normal tissue (14). The combination of NMR spectroscopy and dynamic contrast-enhanced MRI is also investigated to aid in differentiating prostate adenocarcinoma from BPH (15).

This study extended the application of MRI in imaging the prostate (16) to study the incoherent motion of water in the different anatomical regions of the prostate. The aim of this work was to measure for the first time the ADC values in the CG and the peripheral zone (PZ) for both normal and patient groups. The use of the ADC as a differentiating parameter between healthy and malignant tissue within the patient group has been investigated.

MATERIALS AND METHODS

Subjects and MRI

Seven healthy volunteers with mean age 29 years (range between 22 and 39 years) and nineteen patients with mean age 58 years (range between 48 and 67 years) were studied using a General Electric 1.5-Tesla MR scanner. The volunteers did not have any symptoms of prostatic disease and they had normal appearing MR images. Initial scout images were obtained in the coronal plane followed by axial T2-weighted fast spin-echo (FSE) images (TR/TE = 2500/168 msec). Each slice was 4-mm thick and had a 1-mm intersection gap covering the entire gland. These images were inspected initially to define slice locations for the DW imaging sequence. Spin-echo echo-planar (EP) images were then acquired with each image comprising 96 × 96 pixels and covering a field of view (FOV) of 16 cm. The repetition time (TR) and echo time (TE) were 4000 and 120 msec, respectively. The slice thickness was 7 mm, allowing coverage of two to four locations in the prostate according to its size. Smooth shaped regions of interests (ROIs) were drawn on the non-DW-EPI images (b-factor = 0) using accepted criteria (focal areas of hypointensity relative to surrounding parenchyma) by an experienced radiologist. All patients examined in this study had high PSA scores, along with other supporting evidence for prostate carcinoma for some patients, such as results of DRE or biopsy. Subsequent histological confirmation of carcinoma was made available. ROIs that sampled cancer in the PZ, or benign tissue in the CG and PZ, contained, on average, eight pixels. All individual images were inspected on a cine display and rejected if they showed bulk motion. This occurred in four data points only. Statistical analysis was performed on the data by applying either paired or unpaired t-test as detailed in the results section. Signal detection was achieved using a pelvic phased-array (PPA) multi-coil (14) firmly attached to the patient to reduce effects of motion. An endorectal (ER) coil was not used to reduce discomfort of the volunteers and patients.

Diffusion Weighting

Single shot DW-EPI was used to reduce motion effects with the diffusion gradient applied along the phase (blipped) encoding direction (maximum value of 21 mT/m). Four images were averaged for each diffusion gradient step, resulting in a total scanning time for the diffusion study of four minutes.

Due to the high vascular periphery of the prostate, and as was shown previously (16), the DW NMR signal S(b) follows the bi-exponential relation (3, 17):

equation image(1)

S0 is the NMR signal due to T2 relaxation only (i.e., when the diffusion gradient amplitude is set to zero), D is the diffusion coefficient, and D* is the pseudo-diffusion coefficient, which accounts for flow. The fraction of the total water content moving with D* is f. The diffusion exponent b depends on the sequence details used for the diffusion sensitization. For a typical pulsed-gradient spin-echo (PGSE) sequence employing a pair of rectangular-shaped unipolar gradient pulses (18), the gradient b-factor is given by:

equation image(2)

γ is the proton gyromagnetic ratio and G is the diffusion gradient strength. The time between the commencement of the two pulses (Δ) was 37 msec, and the pulse duration (δ) was 30 msec. If the rise and fall time (300 μsec) for the gradient pulse is taken into consideration, then a modified form of Eq. 2 must be used for calculating b (18, 19). The difference in the value of the b-factor between the two forms is only 1.6% for the parameters mentioned above. Including the rise time in the calculation, the six values of the b-factor used in this study were 64, 144, 257, 401, 578, and 786 seconds/mm2. The minimum value was chosen in order to render the effects of D* on the signal negligible (Eq. 1) (20). The values of D, in this case the ADC, were then calculated for all subjects by fitting the data points to a single exponential function.

RESULTS

Two pairs of spin-echo EP images, obtained from a healthy volunteer and a patient, are shown in Figures 1 and 2, respectively. The figures show good contrast between the different anatomical regions of the prostatic gland, namely the PZ, which exhibits higher signal intensity, and the CG. Some residual signal intensity is also present, originating from the bladder due to radiofrequency inhomogeneity; however, this did not contaminate pixels within the prostate. FSE image of the same patient is shown in Figure 3.

Figure 1.

A pair of DW-EP images obtained from a volunteer with b-factor = 0 and 401 seconds/mm2. The images show good contrast between the different anatomical regions of the gland and cover a region 10 × 10 cm2, as the outer area was full of noise only.

Figure 2.

A pair of DW-EP images obtained from a patient with b-factor = 0 and 401 seconds/mm2. Enlarged CG region has compressed PZ, with the two halves of the PZ exhibiting different signal intensity due to cancer in one side, as labeled.

Figure 3.

FSE image obtained from the same patient shown in Fig. 2. Anatomical regions possess similar image contrast to that obtained with EPI. This image has higher spatial resolution (in all three directions) and signal-to-noise ratio, and suffers less geometric distortion than shown in the EP images of Figures 1 and 2.

Signal intensity values sampled from tissues in two sides of a patient's PZ are plotted against the b-factor as shown in Figure 4. The values of ADC are 2.30 × 10−3 and 1.20 × 10−3 mm2/second for the non-cancerous and malignant tissues, respectively, as obtained from the slope of the straight lines fitted to six data points only. Although 14 diffusion sensitization steps were used for this patient to demonstrate the bi-exponential nature of signal decay, all values of ADC reported below were generated using only six data points (b > 60 seconds/mm2). Three ROIs were sampled from each patient (BPH-CG, non-cancerous PZ, and malignant PZ) and two from each volunteer (CG and PZ). Two groups of comparisons are reported below. The first is paired comparisons between tissues from the same prostate (intra-volunteer or intra-patient group), and the second unpaired group comparing the mean ADC value averaged over the volunteer group to that of the patient group (inter-group).

Figure 4.

Signal intensity decay originating from a ROI containing nine pixels in a non-cancerous PZ region (open and closed circles) and malignant PZ tissue (closed triangles). The ADC values obtained from the slope of the straight line fitted to the data points acquired with b > 60 seconds/mm2 (closed circles and triangles only) are 2.3 × 10−3 and 1.2 × 10−3 mm2/second for the benign and cancerous tissues, respectively. Fast non-linear signal decay, due to microcirculation, is exhibited at low b values (open circles).

Examining ADC values in the patient group (intra-patient group), the mean value of the ADC was lower in the malignant PZ tissue (1.38 ± 0.52 × 10−3 mm2/second) than in the non-cancerous PZ tissue (1.82 ± 0.53 × 10−3 mm2/second). The difference was significant with a P value 0.00045 using a paired t-test. Two ROIs representing malignant PZ were excluded from this comparison due to large patient motion (N = 17). In addition to the mean value, the ADC was greater in non-cancerous PZ than in malignant PZ for every individual patient in this group, as reflected by the line plot shown in Figure 5. When the mean values averaged over 17 patients were compared between the two non-cancerous regions of the gland (BPH-CG and PZ), it was significantly higher (P = 0.0028, paired t-test) in PZ (1.88 ± 0.48 × 10−3 mm2/second) than in CG (1.62 ± 0.41 × 10−3 mm2/second). A small difference exists in the ADC values measured in the PZ for the two paired comparisons mentioned above (1.82 and 1.88 × 10−3 mm2/second) due to excluding different pairs of ROIs because of large patient motion. Further comparisons of ADC values within the volunteer group or between the two subject groups are summarized in Table 1.

Figure 5.

Higher ADC values of non-cancerous PZ tissues over malignant PZ tissues exist for every patient (N = 17). The dotted line represents the mean value averaged over 17 patients (non-cancerous mean ADC = 1.82 × 10−3 mm2/second and tumor mean ADC = 1.38 × 10−3).

Table 1. Mean ± SD (× 10−3 mm2/s) of the ADC and P-values for Intra- and Intergroup Comparisons Discussed in the Text*
VolunteerPatient
PZCGNoncancerous PZBPH-CGMalignant PZP-value
  • *

    Higher values of ADC exist in noncancerous PZ over both malignant PZ and BPH-CG tissues. No significant differences exist in the ADC values of noncancerous tissues in the patients group and normal tissues in the volunteer group (either CG or PZ).

  1.82 ± 0.53 1.38 ± 0.520.00045
  1.88 ± 0.481.62 ± 0.41 0.0028
1.91 ± 0.461.63 ± 0.30   0.011
 1.63 ± 0.30 1.62 ± 0.41 >0.05
1.91 ± 0.46 1.82 ± 0.51  >0.05

Finally, although significant differences exist in the mean ADC values, large overlap exists between individual values in all tissue types. This is evident, for example, in Figure 5 for the patient group.

DISCUSSION

This paper presents the first application of DW-EPI to measure the values of ADC in both CG and PZ prostatic tissues for normal and patient groups. The tumor sampled in all patients was located in the PZ, although this does not rule out its infiltration into the CG. ROIs were, therefore, carefully positioned in opposite sides of the prostate (tumor PZ and non-cancerous PZ), or as far separated as possible (tumor PZ and CG). Further reduction in the contamination between CG and PZ voxels can be achieved if improved spatial resolution was available. The required higher signal-to-noise ratio can be made available by the use of an ER coil in combination with the PPA multi-coil. The ER coil would have to be inflated with liquid rather than air to overcome susceptibility artifacts associated with EPI.

MR has been used to discriminate between benign and malignant tissue in the prostate. In addition to 1H spectroscopy, the spatial mapping of T2 (14) offers diagnostic and structural information. As mentioned in the results section, the mean ADC value was lower in the malignant PZ tissue from that in benign or healthy PZ tissue. The significant reduction of the water mobility observed in tumors located in the PZ may be due to the replacement of water-rich acinar structures due to the presence of carcinoma. Prostate cancers commonly have numerous small, closely packed glands with little stroma between them. Normally, the prostatic acini are distributed in a linear pattern radiating from the urethra; however, in cancer, this pattern may be disrupted, and masses of malignant epithelial cells and glands may be irregularly distributed, and the internal architecture of the malignant gland itself may be deranged. Decreased motility may be due to the increased nuclear-to-cytoplasmic ratio, with pleomorphic nuclei where solid sheets of tumor cells are evident (21).

The ADC is significantly higher in non-cancerous PZ tissue than in the CG of the prostate. This is true for the patients, as well as the volunteer sets of subjects, studied in this work, as comparison was performed in paired fashion (intra-group). Higher ADC values reflect the higher mobile water content in the prostatic luminal space and correlates with higher T2 (14) values measured in the PZ. Kurhanewicz et al (22) confirmed the correlation between luminal space and T2 values from histological sections.

Inter-group comparisons were also made between the mean ADC values for the same tissue type using unpaired tests. No statistically significant difference was found for the ADC values in either of the CG or non-cancerous PZ tissues between the patient and the normal group. This is important in view of the unmatched age of the two groups, and suggests that the ADC values are independent of age for the non-malignant tissue. Liney et al (14) showed no correlation between T2 in either region and age.

As a result of the present study, DW-MRI seems to offer a good tool to discriminate between cancerous PZ and PZ in the patients, or between PZ and CG in volunteers; however, large overlap exists between the different groups. ADC values may predict malignancy in the prostate and may help reduce the need for biopsy. Work is being carried out to investigate anisotropic properties of ADC and correlation with histology and grading.

Acknowledgements

The author thanks Yorkshire Cancer Research for supporting this work. The author especially thanks Dr. Gary P. Liney for substantial help in this study and Professor Lindsay W. Turnbull for drawing the ROIs.

Ancillary