Deep Attentive Features for Prostate Segmentation in 3D Transrectal Ultrasound

TL;DR

The paper tackles automatic prostate segmentation in 3D TRUS, a task hampered by boundary ambiguity and intensity inhomogeneity. It introduces a 3D attention-guided network that generates deep attentive features (DAF) by refining per-layer features with a learned attention over multi-level features, integrated with a 3D-FPN backbone, dilated convolutions, and 3D ASPP. A hybrid loss combining Dice and binary cross-entropy with deep supervision is used, and the model is trained end-to-end in four-fold cross-validation. Results on 40 TRUS volumes show state-of-the-art Dice/Jaccard/Conformity metrics and favorable surface distances, with significant gains over baselines and fast inference (~0.3 seconds per volume). The work provides a generalizable strategy for multi-level feature refinement in 3D medical segmentation and releases code for broader use.

Abstract

Automatic prostate segmentation in transrectal ultrasound (TRUS) images is of essential importance for image-guided prostate interventions and treatment planning. However, developing such automatic solutions remains very challenging due to the missing/ambiguous boundary and inhomogeneous intensity distribution of the prostate in TRUS, as well as the large variability in prostate shapes. This paper develops a novel 3D deep neural network equipped with attention modules for better prostate segmentation in TRUS by fully exploiting the complementary information encoded in different layers of the convolutional neural network (CNN). Our attention module utilizes the attention mechanism to selectively leverage the multilevel features integrated from different layers to refine the features at each individual layer, suppressing the non-prostate noise at shallow layers of the CNN and increasing more prostate details into features at deep layers. Experimental results on challenging 3D TRUS volumes show that our method attains satisfactory segmentation performance. The proposed attention mechanism is a general strategy to aggregate multi-level deep features and has the potential to be used for other medical image segmentation tasks. The code is publicly available at https://github.com/wulalago/DAF3D.

Figures (10)

• Figure 1: Example TRUS images. Red contour denotes the prostate boundary. There are large prostate shape variations, and the prostate tissues present inhomogeneous intensity distributions. Orange arrows indicate missing/ambiguous boundaries.
• Figure 2: The visual comparisons of TRUS segmentations using conventional multi-level features (rows 1 and 3) and proposed attentive features (rows 2 and 4). (a) is the input TRUS images; (b)-(e) show the output feature maps from layer 1 (shallow layer) to layer 4 (deep layer) of the convolutional networks; (f) is the segmentation results predicted by corresponding features; (g) is the ground truths. We can observe that directly applying multi-level features without distinction for TRUS segmentation may suffer from poor localization of prostate boundaries. In contrast, our proposed attentive features are more powerful for the better representation of prostate characteristics.
• Figure 3: The schematic illustration of our prostate segmentation network equipped with attention modules. FPN: feature pyramid network; SLF: single-layer features; MLF: multi-layer features; AM: attention module; ASPP: atrous spatial pyramid pooling.
• Figure 4: The schematic illustration of the atrous spatial pyramid pooling (ASPP) with dilated convolution and group normalization (GN).
• Figure 5: The schematic illustration of the proposed attention module.