Male pelvic multi‐organ segmentation on transrectal ultrasound using anchor‐free mask CNN

Abstract

Purpose Current prostate brachytherapy uses transrectal ultrasound images for implant guidance, where contours of the prostate and organs‐at‐risk are necessary for treatment planning and dose evaluation. This work aims to develop a deep learning‐based method for male pelvic multi‐organ segmentation on transrectal ultrasound images. Methods We developed an anchor‐free mask convolutional neural network (CNN) that consists of three subnetworks, that is, a backbone, a fully convolutional one‐state object detector (FCOS), and a mask head. The backbone extracts multi‐level and multi‐scale features from an ultrasound (US) image. The FOCS utilizes these features to detect and label (classify) the volume‐of‐interests (VOIs) of organs. In contrast to the design of a previously investigated mask regional CNN (Mask R‐CNN), the FCOS is anchor‐free, which can capture the spatial correlation of multiple organs. The mask head performs segmentation on each detected VOI, where a spatial attention strategy is integrated into the mask head to focus on informative feature elements and suppress noise. For evaluation, we retrospectively investigated 83 prostate cancer patients by fivefold cross‐validation and a hold‐out test. The prostate, bladder, rectum, and urethra were segmented and compared with manual contours using the Dice similarity coefficient (DSC), 95% Hausdorff distance (HD95), mean surface distance (MSD), center of mass distance (CMD), and volume difference (VD). Results The proposed method visually outperforms two competing methods, showing better agreement with manual contours and fewer misidentified speckles. In the cross‐validation study, the respective DSC and HD95 results were as follows for each organ: bladder 0.75 ± 0.12, 2.58 ± 0.7 mm; prostate 0.93 ± 0.03, 2.28 ± 0.64 mm; rectum 0.90 ± 0.07, 1.65 ± 0.52 mm; and urethra 0.86 ± 0.07, 1.85 ± 1.71 mm. For the hold‐out tests, the DSC and HD95 results were as follows: bladder 0.76 ± 0.13, 2.93 ± 1.29 mm; prostate 0.94 ± 0.03, 2.27 ± 0.79 mm; rectum 0.92 ± 0.03, 1.90 ± 0.28 mm; and urethra 0.85 ± 0.06, 1.81 ± 0.72 mm. Segmentation was performed in under 5 seconds. Conclusion The proposed method demonstrated fast and accurate multi‐organ segmentation performance. It can expedite the contouring step of prostate brachytherapy and potentially enable auto‐planning and auto‐evaluation.

Deep learning-based two-step organs at risk auto-segmentation model for brachytherapy planning in parotid gland carcinoma

“…Current gold standard manual segmentation is, however, laborious and time-consuming [ 26 ]. Increasing attention has been placed on the use of deep learning-assisted models for automated segmentation in brachytherapy [ 27 , 28 ]; however, to our knowledge, such models have not been utilized in the head and neck region. Precise delineation and prevention of severe radiotherapy-related complications are particularly crucial in head and neck cancers, given the complexity of anatomical structures and clinical emphasis on maintaining aesthetics [ 29 ].…”

Automatic segmentation of neurovascular bundle on mri using deep learning based topological modulated network

“…The purpose of focal modulation is to locate the left and right NVBs from the explored features of input MR image I Img . In recent medical imaging studies, OAR locations were estimated via deep learning‐based object detection networks, such as a cascaded network composed of regional proposal network and regional convolutional neural network, 26 and a fully convolutional one‐state object detector 27 . However, a potential problem of using these convolutional layer‐based networks is that a small OAR's global and spatial information may not be captured.…”

“…The activation map is denoted by $${\{ {F_{\rm{A}}^{\rm{i}} \in {R^{{w^i} \times {h^i} \times {d^i} \times {n^i}}}} \}_{i\; = {\rm{\;}}1,2}}$$. A description of hierarchical block implementation is detailed in our previous study 27 …”

“…In recent medical imaging studies, OAR locations were estimated via deep learning-based object detection networks, such as a cascaded network composed of regional proposal network and regional convolutional neural network, 26 and a fully convolutional one-state object detector. 27 However, a potential problem of using these convolutional layer-based networks is that a small OAR's global and spatial information may not be captured. This limitation is attributable to the feature extraction method of sliding limited sized kernels (limited feature receptive field) from the previous layers' feature map and down-sampling via pooling or setting of kernel's stride size larger than one.…”

Effect of dataset size, image quality, and image type on deep learning-based automatic prostate segmentation in 3D ultrasound

“…With the increasing prevalence of deep learning in medicine (Piccialli et al 2021), and specifically of convolutional neural networks (CNNs) for medical imaging tasks, many deep learning-based automatic prostate segmentation approaches have been proposed for TRUS imaging, promising reduced procedure time and similar performance compared to manual approaches (Anas et al 2018, Ghavami et al 2018, Karimi et al 2019, Lei et al 2019, Wang et al 2019. Recently, Lei et al proposed an anchor-free mask CNN for multi-organ segmentation in 3D TRUS volumes, trained using data from 83 PCa patients with five-fold cross-validation (Lei et al 2021). They reported prostate segmentation accuracy with a Dice similarity coefficient (DSC) of 0.93±0.03 and 95% Hausdorff distance (HD 95 ) of 2.27±0.79 mm.…”