Reliability of Robotic Ultrasound Scanning for Scoliosis Assessment in Comparison with Manual Scanning

TL;DR

This study demonstrates that robotic ultrasound scanning provides scoliosis measurements comparable to manual methods, with high reliability and consistency, potentially improving ultrasound assessment for scoliosis by non-expert operators.

Contribution

The paper introduces a robotic ultrasound scanning system that achieves reliable scoliosis assessment comparable to manual scanning, addressing operator dependency issues.

Findings

High intra- and inter-rater reliability (ICC > 0.77)

No significant difference between robotic and manual measurements (p < 0.05)

Acceptable measurement error with MAD within 0-5 degrees for most angles

Abstract

Background: Ultrasound (US) imaging for scoliosis assessment is challenging for a non-experienced operator. The robotic scanning was developed to follow a spinal curvature with deep learning and apply consistent forces to the patient' back. Methods: 23 scoliosis patients were scanned with US devices both, robotically and manually. Two human raters measured each subject's spinous process angles (SPA) on robotic and manual coronal images. Results: The robotic method showed high intra- (ICC > 0.85) and inter-rater (ICC > 0.77) reliabilities. Compared with the manual method, the robotic approach showed no significant difference (p < 0.05) when measuring coronal deformity angles. The MAD for intra-rater analysis lies within an acceptable range from 0 deg to 5 deg for a minimum of 86% and a maximum 97% of a total number of the measured angles. Conclusions: This study demonstrated that scoliosis deformity angles measured on ultrasound images obtained with robotic scanning are comparable to those obtained by manual scanning.