LARS
Description
A deep learning model, Lymphoma Artificial Reader System (LARS), was developed to classify [18F]FDG-PET-CT scans of lymphoma patients based on the presence of hypermetabolic tumor sites. Trained on over 16,000 scans, LARS demonstrated high accuracy, sensitivity, and specificity in both internal and external test cohorts. This model has the potential to assist imaging specialists in managing large scan volumes by accurately identifying metabolically active disease, potentially serving as a second reader or decision support tool.
Creator
Memorial Sloan Kettering
Information
Pediatrics or Adult
Adult
Speciality
Oncology
Modality
[18F]FDG-PET-CT
Training
16,583 [18F]FDG-PET-CTs of 5,072 patients with lymphoma, 1,000 additional scans from a second center for external training
Github
Publication
FDA