Automatic Region of Interest Detection as a Complement for Reverberant Shear Wave Elastography Assessment in Foot

Abstract

Plantar soft tissue has a critical role in foot biomechanics. The disorders associated with the status of this tissue can influence an individual’s mobility. In this regard, previous studies show the relevance of stiffness quantification through elastography approaches. In particular, the Reverberant Shear Wave Elastography (RSWE) was used to differentiate elasticity values through the Shear Wave Speed (SWS) calculation from different groups using the 1st, the 3rd metatarsal head (MH) and the heel pad as anatomical landmarks. In the acquisition, anatomical landmarks are normally distinguished by expert health personnel whereas the selection of the region of interest (ROI) requires human intervention with the same expertise in foot ultrasound. In this study, an average intensity curve is created along the lateral axis for the automatic identification of the ROI. A median filter is applied to the curve to reduce noise while preserving important peak values. Subsequently, along the axial axis, the Otsu thresholding method is utilized for the segmentation of bone. The results showed minimal difference in the ROI selection between automatic and manual selection by comparing the mean SWS using their respective ROI. The variations were observed for each specific foot region and the kernel used in the median filter. The best performance in ROI selection was achieved with a kernel size of 19 for the 1st metatarsal (mAP 78.64%), 16 for the 3rd metatarsal (mAP 100%), and 10 for the heel (mAP 96.65%). The presented methodology has the potential for the automatic detection of ROI in foot ultrasound as a complement to RSWE. Finally, this approach holds the potential to facilitate and enhance subsequent image acquisitions utilizing the RSWE technique.