SBIR TOPIC 454: A SOFTWARE TOOL FOR STANDARDIZATION OF RADIOTHERAPY TREATMENT PLANS FOR AI VALIDATION

The goal of this proposal is to create a medical device development tool (MDDT) for use during performance evaluation of artificial intelligence (AI)-based radiotherapy software. The MDDT is intended to be used by manufacturers during external validation, whereby AI models are tested for generalizability. Test results can be used for pre-market submissions to the Food and Drug Administation (FDA) and post-market surveillance. The role of the MDDT is to standardize test datasets consisting of tens or hundreds of radiotherapy treatment plans. Functionality includes: making contoured structure names comply with American Association of Physicists in Medicine (AAPM) Task Group 263 (TG-263) nomenclature, making plan information searchable, and creation of test cohorts. The MDDT will be cloud-based, employing Microsoft Azure Health Data services. Uploaded treatment plans, and derived information such as standardized names, will be stored using HL-7 Fast Healthcare Interperability Resources (FHIR) and Digital Imaging and Communications in Medicine (DICOM). The specific aims of this Phase I project are to design and implement the data communication, data storage, and user interface elements of the MDDT, to carry out a pilot usability study, and to create a qualification plan detailing steps to qualify the tool within the FDA MDDT program.