Innovative AI Tool Advances Nerve Disorder Screening

Introduction

Researchers at the Indian Institute of Science (IISc) in Bengaluru have achieved a significant breakthrough with the development of an artificial intelligence (AI) tool designed to detect carpal tunnel syndrome (CTS) using ultrasound videos. Collaborating closely with Aster-CMI Hospital in Bengaluru, the team focused on identifying the median nerve, a key indicator of CTS, within these videos.

Carpal tunnel syndrome arises when the median nerve, which extends from the forearm into the hand, becomes compressed at the wrist's carpal tunnel. This compression often leads to symptoms such as numbness, tingling, or pain, particularly affecting individuals engaged in repetitive hand movements, such as office workers, assembly line personnel, and athletes.

Traditionally, doctors utilize ultrasound imaging to examine the median nerve, assessing its dimensions, shape, and any potential abnormalities. However, interpreting ultrasound images and videos presents challenges compared to other imaging techniques like X-rays or MRI scans.

Karan R Gujarati, formerly an MTech student at IISc and lead author of the study published in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, noted the difficulty in discerning nerve boundaries in ultrasound images, particularly in areas beyond the wrist where multiple structures obscure clarity. This challenge prompted the team to explore AI as a solution.

Drawing inspiration from transformer architecture—similar to the technology behind ChatGPT—the researchers adapted a machine learning model originally developed for detecting multiple objects in YouTube videos. They streamlined the model, focusing its capabilities exclusively on tracking the median nerve. To train the model, ultrasound videos from both healthy participants and individuals diagnosed with CTS were meticulously annotated in collaboration with Lokesh Bathala, Lead Consultant Neurologist at Aster-CMI Hospital.

Once trained, the AI model demonstrated proficiency in segmenting the median nerve across frames of ultrasound videos, akin to tracking objects in a moving video. Professor Phaneendra K Yalavarthy, corresponding author of the study and faculty at IISc's Center for Data Science (CDS), emphasized the model's capability to measure the nerve's cross-sectional area automatically—a crucial diagnostic parameter for CTS, conventionally assessed manually by sonographers.

"The tool automates this process, providing real-time measurements with more than 95% accuracy at the wrist region," highlighted Bathala, underscoring the tool's potential to enhance diagnostic efficiency and accuracy in clinical settings. By automating the measurement of the nerve's cross-sectional area, the AI tool not only expedites the diagnostic process but also standardizes measurements, reducing variability inherent in manual assessments.

This development marks a significant stride towards leveraging AI in healthcare for precise diagnostic applications, transforming how nerve-related disorders like CTS are diagnosed and managed. As technology continues to evolve, such innovations promise to make healthcare more accessible, efficient, and patient-centric, ultimately improving outcomes for individuals worldwide affected by such conditions.