Magnetic resonance imaging (MRi) provides detailed anatomical images of the prostate and its zones. it has a crucial role for many diagnostic applications. Automatic segmentation such as that of the prostate and prostate zones from MR images facilitates many diagnostic and therapeutic applications. However, the lack of a clear prostate boundary, prostate tissue heterogeneity, and the wide interindividual variety of prostate shapes make this a very challenging task. to address this problem, we propose a new neural network to automatically segment the prostate and its zones. We term this algorithm Dense U-net as it is inspired by the two existing state-of-the-art tools-Densenet and U-net. We trained the algorithm on 141 patient datasets and tested it on 47 patient datasets using axial T2-weighted images in a four-fold cross-validation fashion. The networks were trained and tested on weakly and accurately annotated masks separately to test the hypothesis that the network can learn even when the labels are not accurate. the network successfully detects the prostate region and segments the gland and its zones. compared with U-net, the second version of our algorithm, Dense-2 U-net, achieved an average Dice score for the whole prostate of 92.1± 0.8% vs. 90.7 ± 2%, for the central zone of 89.5 ± 2 % vs. 89.1 ± 2.2 %, and for the peripheral zone of 78.1± 2.5% vs. 75 ± 3 %. Our initial results show Dense-2 U-net to be more accurate than state-of-the-art U-net for automatic segmentation of the prostate and prostate zones.

No abstract

Objectives Water diffusion, tissue stiffness, and viscosity characterize the biophysical behavior of tumors. However, little is known about how these parameters correlate in prostate cancer (PCa). Therefore, we paired tomoelastography of the prostate with diffusion-sensitive magnetic resonance imaging for the quantitative mapping of biophysical parameters in benign prostatic hyperplasia (BPH) and PCa. Materials and Methods Multifrequency magnetic resonance imaging elastography with tomoelastography processing was performed at 60, 70, and 80 Hz using externally placed compressed-air drivers. Shear-wave speed (SWS) and loss angle (φ) were analyzed as surrogate markers of stiffness and viscosity-related fluidity in the normal peripheral zone (PZ), hyperplastic transition zone (TZ), which is consistent with BPH, and PCa lesions. The SWS and φ were correlated with the normalized apparent diffusion coefficient (nADC). Results Thirty-nine men (median age/range, 67/49–88 years), 25 with BPH and 14 with biopsy-proven PCa, were prospectively enrolled in this institutional review board–approved study. The SWS in PCa (3.1 ± 0.6 m/s) was higher than in TZ (2.8 ± 0.3 m/s, P = 0.004) or tended to be higher than in PZ (2.8 ± 0.4 m/s, P = 0.025). Similarly, φ in PCa (1.1 ± 0.1 rad) was higher than in TZ (0.9 ± 0.2 m/s, P < 0.001) and PZ (0.9 ± 0.1 rad, P < 0.001), whereas nADC in PCa (1.3 ± 0.3) was lower than in TZ (2.2 ± 0.4, P < 0.001) and PZ (3.1 ± 0.7, P < 0.001). Pooled nADC was inversely correlated with φ (R = −0.6, P < 0.001) but not with SWS. TZ and PZ only differed in nADC (P < 0.001) but not in viscoelastic properties. Diagnostic differentiation of PCa from normal prostate tissues, as assessed by area under the curve greater than 0.9, was feasible using nADC and φ but not SWS. Conclusions Tomoelastography provides quantitative maps of tissue mechanical parameters of the prostate. Prostate cancer is characterized by stiff tissue properties and reduced water diffusion, whereas, at the same time, tissue fluidity is increased, suggesting greater mechanical friction inside the lesion. This biophysical signature correlates with known histopathological features including increased cell density and fibrous protein accumulation.