UltraMN: Advancing Real-Time Median Nerve Ultrasound Monitoring With a Multitask Deep Learning Framework.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Deep Learning

Background:

• Real-time ultrasound monitoring of the median nerve presents significant challenges.
• Accurate identification and segmentation are crucial for diagnosing median nerve disorders.

Purpose of the Study:

• To develop an advanced deep learning framework, UltraMN, for real-time median nerve ultrasound.
• To improve the accuracy and efficiency of median nerve assessment using AI.

Main Methods:

• Proposed UltraMN, a novel multitask learning model integrating standard plane classification (UltraCLS) and tissue segmentation (UltraSEG).
• Utilized a dataset of 3568 videos (249,985 images) across four standardized imaging planes (4-SIP).
• Compared UltraCLS performance against MedMamba and FPT models; evaluated UltraSEG using precision, recall, F1 scores, and mean Intersection over Union (mIoU).

Main Results:

• UltraMN significantly outperformed existing models (MedMamba, FPT).
• UltraCLS achieved 95.6% classification accuracy with >95.0% precision, recall, and F1 scores.
• UltraSEG demonstrated superior segmentation with an mIoU of 97.6% across all imaging planes.

Conclusions:

• UltraMN provides a robust and efficient solution for real-time median nerve assessment.
• Achieved high classification accuracy and precise segmentation, enhancing diagnostic potential.
• This feasibility study on healthy subjects requires further validation for pathological conditions like carpal tunnel syndrome.