Deep Learning Based Shear Wave Detection and Segmentation Tool for Use in Point-of-Care for Chronic Liver Disease...

Area of Science:

• Hepatology
• Medical Imaging
• Artificial Intelligence

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) prevalence is increasing globally.
• Accurate, point-of-care liver assessment tools are needed for MASLD diagnosis and management.
• Current methods for liver assessment can be invasive or lack precision.

Purpose of the Study:

• To evaluate the performance of a novel deep learning-based software tool for liver shear wave detection and segmentation.
• To assess the software's utility in improving liver tissue characterization for MASLD patients.
• To determine the effectiveness of the software in a real-world, point-of-care setting using the Velacur device.

Main Methods:

• A U-Net architecture deep learning algorithm was trained on 15,045 expert-segmented images from 103 patients.
• The algorithm was tested on 4,429 images from 36 volunteers and MASLD patients across different clinics.
• Performance was evaluated using sensitivity, specificity, and Dice coefficient, with a prototype tested on the Velacur system.

Main Results:

• The shear wave detection algorithm achieved 81% sensitivity and 84% specificity.
• Dice coefficients for image-based and patient-based averages were 0.74 and 0.75, respectively.
• Software implementation as a B-Mode ultrasound overlay enhanced the quality of liver assessments by operators.

Conclusions:

• The deep learning shear wave algorithm demonstrated strong performance in a diverse test set of volunteers and MASLD patients.
• Integrating this software into the Velacur system significantly improved the quality of liver assessments at the point of care.
• The tool shows promise for enhancing diagnostic accuracy and efficiency in MASLD management.