Abstract #1139

Multimodal qMRI Framework for Knee Imaging Biomarker Fusion and Osteoarthritis Prediction

Alejandro Morales Martinez^1,2, Francesco Caliva¹, Claudia Iriondo^1,2, Sarthak Kamat¹, Sharmila Majumdar¹, and Valentina Pedoia^1,2,3

¹Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States, ²Graduate Program in Bioengineering, University of California, Berkeley, Berkeley, CA, United States, ³Center for Digital Health Innovation (CDHI), University of California San Francisco, San Francisco, CA, United States

Bone and cartilage segmentation models were trained and validated with a segmented dataset of 40 and 176 3D DESS MRI volumes respectively. The trained models were used to run inference on 20,989 3D DESS MRI volumes from the Osteoarthritis Initiative dataset. Biomarkers such as femoral bone shape, cartilage thickness and cartilage T2 average values were extracted from the segmentations. Point clouds representing each biomarker were transformed into spherical coordinates and merged using different fusion strategies. The spherical maps were used to train an OA diagnosis model with a test specificity, sensitivity and AUC was 84.1%, 78.7%, and 89.7% respectively.

This abstract and the presentation materials are available to members only; a login is required.

Join Here