Diffusion-weighted imaging (DWI) is a key component in multi-parametric MRI (mp-MRI) characterization of prostate cancer. However, geometric distortions, image shifts, and image artifacts are significant problems for conventional DWI using single-shot echo planar imaging (ssEPI¹, Fig.1a). The distortion and artifacts may not only degrade the diffusion information, but also obstruct direct alignment of DWI with complementary T2-weighted (T2w) MRI and dynamic contrast-enhanced MRI. Multi-shot EPI sequences, e.g., readout-segmented EPI (RESOLVE², Fig.1b), have shown promise in reducing geometric distortions in prostate DWI^3,4, but are limited in clinical practice due to comparatively long scan times. In this work, we develop and evaluate a new prostate DWI technique that combines RESOLVE with reduced field-of-view imaging (rFOV RESOLVE) to reduce geometric distortion and improve scan time efficiency.

Sequence: To acquire rFOV RESOLVE (Fig.1c) while eliminating aliasing artifacts, two outer volume spatial saturation (OVSS) bands were placed coronally, anterior and posterior to the FOV in the axial plane. The rFOV allowed data acquisition with fewer segments, achieving higher scan-time efficiency than RESOLVE.

Experiments: Eleven healthy male volunteers (age = 32±8 years) were scanned on a 3T system (MAGNETOM Skyra, Siemens) with three DWI protocols: standard clinical single-shot EPI (1 segment), RESOLVE (7 segments) and rFOV RESOLVE (5 segments). The majority of imaging parameters were matched (Table 1). The readout durations and total scan durations for ssEPI/RESOLVE/rFOV RESOLVE were 42/39/23 ms and 5:52/7:03/4:59 min, respectively. Phase-encoding (PE) direction was anterior-posterior (A-P). High-resolution T2w turbo spin-echo (TSE) images were acquired with matched slice locations for anatomical reference.

Quantitative Analysis: To assess geometric distortion and shift, all three DWI protocols were acquired with forward (A-P) and reverse (P-A) PE directions. The forward and reverse PE images were each manually segmented on a central slice of the prostate using a free-form polygonal region of interest (MATLAB, The MathWorks) to comprise only the prostate. Geometric distortion/shift was measured by the overlap between the prostate segmentation boundary from forward and reverse PE images using the Dice similarity coefficient⁵:

$$DSC = \frac{2(A \cap B)}{|A|+|B|}$$

where A and B denote the forward and reverse PE images, respectively. DSC is an index between 0 and 1, with higher values indicating greater overlap between two images. Mean and standard deviations of DSC were calculated from n=11 volunteers, and pairwise t-tests (p=0.05 to indicate significance) were performed to determine statistically significant DSC differences between protocols.

Visual inspection of all images showed differences in the shape of the prostate using the three prostate DWI protocols (Fig.2), and within the same protocol in forward and reverse PE directions (Fig.3). Figure 2 shows images from a representative volunteer at the full FOV (a-d) and zoomed in to the rFOV (e-h), with anatomical reference (Fig.2a,e) for the diffusion images. ssEPI (Fig.2b,f) shows characteristic blurring, distortion and artifacts in the PE direction. In general, the prostate is better delineated using either RESOLVE protocol (Fig.2c,d,g,h), although some stretching and compression was observed on RESOLVE images (Fig.2c,g). The prostate outline in rFOV RESOLVE (Fig.2d,h) closely matched the ROI from the T2w anatomical reference.

The forward (Fig.3a-c) and reverse (Fig.3d-f) PE images were used in DSC quantification (Fig.3g-i). Mean DSC increased from 0.74±0.18 in ssEPI to 0.77±0.18 in RESOLVE and 0.91±0.05 in rFOV RESOLVE. The differences between ssEPI and rFOV RESOLVE were statistically significant (p<0.02). No significant differences were found between RESOLVE and either ssEPI or rFOV RESOLVE, although the DSC difference between RESOLVE and rFOV RESOLVE was also comparatively large (p=0.08).

Geometric distortions, image shifts and blurring are extremely common in the PE direction in conventional prostate DWI using ssEPI due to long readout durations. RESOLVE can reduce spatial distortion effects and image shift; however, RESOLVE cannot always be acquired within a clinically feasible scan duration. Our results in healthy volunteers (n=11) indicate that rFOV RESOLVE can improve the geometric fidelity of DWI in the prostate (DSC = 0.91) compared with ssEPI (DSC = 0.74). This improvement in image quality is accompanied by higher scan time efficiency than RESOLVE due to fewer acquisition segments (7 vs. 5 in the current design), with less than 5 minutes scan duration for this specific protocol. Future work aims to explore the relative advantages of alternative rFOV strategies to OVSS, such as inner-volume imaging and two-dimensional RF excitation pulses. In conclusion, rFOV RESOLVE was shown to have significantly lower distortion than ssEPI and higher time efficiency than RESOLVE, and may be a promising new technique for DWI in the prostate.