Automatic Segmentation for Favourable Delineation of Ten Wrist Bones on Wrist Radiographs Using Convolutional Neural Network

Abstract

Purpose: This study aimed to develop and validate an automatic segmentation algorithm for the boundary delineation of ten wrist bones, consisting of eight carpal and two distal forearm bones, using a convolutional neural network (CNN). Methods: We performed a retrospective study using adult wrist radiographs. We labeled the ground truth masking of wrist bones, and propose that the Fine Mask R-CNN consisted of wrist regions of interest (ROI) using a Single-Shot Multibox Detector (SSD) and segmentation via Mask R-CNN, plus the extended mask head. The primary outcome was an improvement in the prediction of delineation via the network combined with ground truth masking, and this was compared between two networks through five-fold validations. Results: In total, 702 images were labeled for the segmentation of ten wrist bones. The overall performance (mean (SD] of Dice coefficient) of the auto-segmentation of the ten wrist bones improved from 0.93 (0.01) using Mask R-CNN to 0.95 (0.01) using Fine Mask R-CNN (p < 0.001). The values of each wrist bone were higher when using the Fine Mask R-CNN than when using the alternative (all p < 0.001). The value derived for the distal radius was the highest, and that for the trapezoid was the lowest in both networks. Conclusion: Our proposed Fine Mask R-CNN model achieved good performance in the automatic segmentation of ten overlapping wrist bones derived from adult wrist radiographs.

Keywords: CNN; carpal bone; deep learning; segmentation; wrist.

Figure 1

Anatomy and labeling of ten wrist bones on a wrist radiograph. (a) The anatomy of ten wrist bones, consisting of eight carpal bones and two distal forearm bones, on an anteroposterior radiograph. (b) Labeling process for the ground truth masking of wrist bones using a self-made customized tool. (1) The classification as one of ten wrist bones and the delineation of each bone’s boundary; (2) Labeling and extraction of each bone.

Figure 2

The Fine Mask R-CNN architecture. Our proposed network operates on a 2-stage method. (a) Detection of the regions of interest in the input wrist radiographs using SSD (blue) in the first stage and the delineation of 10 wrist bones using Mask R-CNN with the extended mask head in the second stage (yellow). (b) The structure of the extended mask head. This is an encoder–decoder structure, which can use previous information for the prediction of a specific part. ROI, region of interest; CNN, convolutional neural networks; SSD, Single-Shot Multibox Detector.

Figure 3

Visualization of Fine Mask R-CNN and Mask R-CNN network for the segmentation of ten wrist bones. (a) Original image of each wrist bone on the radiograph, (b) Delineation of segmented bone by physicians manually, (c) Delineation of segmented bone by Mask R-CNN, (d) Delineation of segmented bone by Fine Mask R-CNN with an extended mask head. Black lines indicate the ground truth masking segmented by physicians and yellow lines indicate the predicted masking segmented by CNN. CNN; convolutional neural networks.