Foundational models for radiology are gaining significant traction. Companies like Harrison (Annalise) and Aidoc have recently introduced the first versions of their foundational AI models for radiology (CARE1 model; Aidoc, Harrison.rad.1; Annalise). These large-scale models, trained on extensive unlabeled and diverse datasets, can tackle multiple radiology tasks without requiring specialized training for each individual application. Their capabilities include enhancing image interpretation and automating report generation.
Despite their promise, foundational models still face challenges, such as ensuring data diversity, mitigating biases, and managing high computational demands. Researchers from Stanford have published a review aiming to establish standardized terminology for foundational models, focusing on the requirements for training data, training paradigms, model capabilities, and evaluation strategies. Discussion around these topics is critical to the eventual responsible adoption of these models into clinical practice. Definitely worth a read for anyone interested in the future of radiology AI!
Read full study
Foundation Models in Radiology: What, How, When, Why and Why Not
Arxiv, 2024
Abstract
Recent advances in artificial intelligence have witnessed the emergence of large-scale deep learning models capable of interpreting and generating both textual and imaging data. Such models, typically referred to as foundation models, are trained on extensive corpora of unlabeled data and demonstrate high performance across various tasks. Foundation models have recently received extensive attention from academic, industry, and regulatory bodies. Given the potentially transformative impact that foundation models can have on the field of radiology, this review aims to establish a standardized terminology concerning foundation models, with a specific focus on the requirements of training data, model training paradigms, model capabilities, and evaluation strategies. We further outline potential pathways to facilitate the training of radiology-specific foundation models, with a critical emphasis on elucidating both the benefits and challenges associated with such models. Overall, we envision that this review can unify technical advances and clinical needs in the training of foundation models for radiology in a safe and responsible manner, for ultimately benefiting patients, providers, and radiologists.