Radiologists from all practice sizes, locations and settings are involved in every aspect of ACR DSI’s work. They participate on use case development panels and create annotated image datasets to help developers train and test algorithms for clinical use. Radiologists also lead valuable educational efforts for radiologists and developers, and provide expert perspectives through the DSI’s AI Assistant blog, the RFS AI Journal Club , and the JACR .
We value our relationship with the AI developer community and are working together with them to lay a foundation for successful implementation of AI in radiological practice. With our panels of specialty experts, we are developing and publishing freely available use cases for AI development in our TOUCH-AI Directory. With detailed narrative descriptions, these structured use cases provide guidance on clinical context with attention to the needs of the radiological community and the problems that are most amenable to AI solutions. We’re also working with a large variety of vendors (EHRs, PACS, RIS, VRs, VNAs and more) to define standards on how to encode and communicate information generated by AI so that its effects can be as far reaching as possible.
To create these “end-to-end” AI use cases for developers of all sizes, we’re leveraging two ACR open frameworks. These include tools and common data elements radiologists can use to create annotated datasets for algorithm training and testing, statistical metrics for validating AI algorithms, and pathways to integrate algorithm output into the clinical workflow and monitor the algorithm’s performance in clinical practice.