AI algorithm measures shoulder kinematics on DDR

Wednesday, November 29 | 9:40 a.m.-9:50 a.m. | W3-SSMK08 | Room 5E450A

In this session on musculoskeletal imaging, a deep-learning AI algorithm will be presented for measuring shoulder kinematics using dynamic digital radiography (DDR) images.

John Sabol, PhD, a scientist at DDR developer Konica Minolta Healthcare Americas, will discuss a prototype deep-learning model developed to measure the scapulohumeral rhythm (SHR), the motion between the scapula and the humerus.

“Although dynamic visualization of anatomy is useful to detect numerous pathologies, quantitative assessment of joint motion may improve sensitivity and specificity of DRR in diagnosis and evaluation of therapy response,” Sobol noted.

The researchers trained the algorithm on 447 images from 267 cases to recognize the humerus and scapula positions, calculate the scapulothoracic and glenohumeral angles, and thus determine the SHR across the complete range of abduction. AI and human measurements were compared using intraclass correlations (ICC) and interreader reliability was excellent between the two manual measurements (ICC 0.87), according to the findings.

Ultimately, the prototype algorithm has the potential to diagnose shoulder conditions by automated analysis of scapular and humeral kinematics, according to the group.

Check out this session on Tuesday morning for the details.