A deep learning approach to median nerve evaluation in ultrasound images of carpal tunnel inlet

Abstract

Ultrasound (US) imaging is recognized as a useful support for Carpal Tunnel Syndrome (CTS) assessment through the evaluation of median nerve morphology. However, US is still far to be systematically adopted to evaluate this common entrapment neuropathy, due to US intrinsic challenges, such as its operator dependency and the lack of standard protocols. To support sonographers, the present study proposes a fully-automatic deep learning approach to median nerve segmentation from US images. We collected and annotated a dataset of 246 images acquired in clinical practice involving 103 rheumatic patients, regardless of anatomical variants (bifid nerve, closed vessels). We developed a Mask R-CNN with two additional transposed layers at segmentation head to accurately segment the median nerve directly on transverse US images. We calculated the cross-sectional area (CSA) of the predicted median nerve. Proposed model achieved good performances both in median nerve detection and segmentation: Precision (Prec), Recall (Rec), Mean Average Precision (mAP) and Dice Similarity Coefficient (DSC) values are 0.916 ± 0.245, 0.938 ± 0.233, 0.936 ± 0.235 and 0.868 ± 0.201, respectively. The CSA values measured on true positive predictions were comparable with the sonographer manual measurements with a mean absolute error (MAE) of 0.918 mm². Experimental results showed the potential of proposed model, which identified and segmented the median nerve section in normal anatomy images, while still struggling when dealing with infrequent anatomical variants. Future research will expand the dataset including a wider spectrum of normal anatomy and pathology to support sonographers in daily practice.

Keywords: Carpal tunnel syndrome; Deep learning; Median nerve; Segmentation; Ultrasound imaging.

Fig. 1

US transverse scan sample acquired at the proximal carpal tunnel inlet. A red box includes the median nerve section; asterisks of different colors mark other relevant structures: pisiform bone profile in blue, semilunar bone profile in purple, ulnar artery in green, digital flexor tendons in orange

Fig. 2

Schematic representation of model architecture, composed by a backbone, Region Proposal Network (RPN), and the three heads for classification, bounding box regression and segmentation, all fed from the ROIAlign with 100 ROI candidates. The segmentation head is represented more in details as it was provided with two additional transposed layers compared with original Mask-RCNN [13]

Fig. 3

Four visual samples of the median nerve section. From top to bottom row: original US image, ground truth mask, U-Net trained with BCE loss prediction, U-Net trained with BCE − DSC loss prediction, Lightweight U-Net trained with BCE loss prediction, Lightweight U-Net trained with BCE − DSC loss prediction, proposed model prediction. For displaying purpose, only the upper part of the US images, which contains the median nerve section, is shown