Abstract #0239

A BART machine learning model to classify lipomatous tumors in MR images

Felipe Godinez^1,2, Nimu Yuan³, Yasser G Abdelhafez⁴, Anik Roy⁵, Hande Nalbant¹, Cyrus P Bateni¹, Jinyi Qi³, Michelle Zhang⁶, Sonia Lee⁷, Ahmed Moawad⁸, Michele Guindani⁹, and Lorenzo Nardo¹

¹Radiology, University of California Davis, Sacramento, CA, United States, ²Comprehensive Cancer Center, University of California Davis, Sacramento, CA, United States, ³Biomedical Engineering, University of California Davis, Davis, CA, United States, ⁴University of California Davis, Sacramento, CA, United States, ⁵Indian Statistical Institute, Kolkata, India, ⁶Radiology, McGill University Health Center, Montreal, ON, Canada, ⁷Radiological Sciences, University of California Irvine, Irvine, CA, United States, ⁸Diagnostic Imaging, MD Anderson Cancer Center, Houston, TX, United States, ⁹Biostatistics, University of California Los Angeles, Los Angeles, CA, United States

Synopsis

Keywords: Diagnosis/Prediction, AI/ML Software

Motivation: Atypical lipomatous tumors (ALTs) neoplasms are distinguished from benign lipomas primarily through histopathology. MRI alone has variable effectiveness in distinguishing ALTs from lipomas. Machine learning has shown promise in this differentiation.

Goal(s): This study introduces the Bayesian additive regression trees (BART) model for non-invasive ALT classification, leveraging Bayesian priors to enhance model robustness.

Approach: Retrospective data was collected from 5 medical institutions, 437 patients where used to train the model.

Results: A Bayesian additive regression trees model, built from MRI radiomic features, performed comparable to an experienced radiologist for classifying lipomatous tumors.

Impact: This method can assist radiologists in screening for ALT in regions with low incidence rates. By using machine learning less biopsies’ will be needed. We would like to implement this model in other detection tasks such as brain tumor classification.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.