Algorithm Development for Fascicle-Aponeurosis Detection from Ultrasound Images

Mentor 1

Xuefeng Bao

Start Date

10-5-2022 10:00 AM

Description

The amount of torque produced by a human muscle is highly correlated with the angle between fascicles and the aponeurosis of that muscle. The torque produced by a human muscle can then be calculated in a standard way using visual processing software. We obtained an ultrasound image stream of the muscle fibers from a human subject during functional actions. Then, we invited a human expert to label the two target muscle fibers as our reference, or the ground truth. The goal was to then accurately calculate the angle between muscle fibers by matching the calculation to the ground truth. Using Python’s OpenCV library, it was possible to calculate the angle between the fascicle and aponeurosis with acceptable accuracy. Utilizing Microsoft Azure, Numba, and CUDA enabled us to use parallel computing on selected GPUs to decrease the time complexity of the algorithm without sacrificing the accuracy of the results. The calculation of human muscle torque performed in this research has applications in human robotics, degenerative muscle disease tracking, and physical therapy. The decrease in program run time increases the potential for real-time computing for these applications in clinical settings.

This document is currently not available here.

Share

COinS
 
May 10th, 10:00 AM

Algorithm Development for Fascicle-Aponeurosis Detection from Ultrasound Images

The amount of torque produced by a human muscle is highly correlated with the angle between fascicles and the aponeurosis of that muscle. The torque produced by a human muscle can then be calculated in a standard way using visual processing software. We obtained an ultrasound image stream of the muscle fibers from a human subject during functional actions. Then, we invited a human expert to label the two target muscle fibers as our reference, or the ground truth. The goal was to then accurately calculate the angle between muscle fibers by matching the calculation to the ground truth. Using Python’s OpenCV library, it was possible to calculate the angle between the fascicle and aponeurosis with acceptable accuracy. Utilizing Microsoft Azure, Numba, and CUDA enabled us to use parallel computing on selected GPUs to decrease the time complexity of the algorithm without sacrificing the accuracy of the results. The calculation of human muscle torque performed in this research has applications in human robotics, degenerative muscle disease tracking, and physical therapy. The decrease in program run time increases the potential for real-time computing for these applications in clinical settings.