Healthcare technology company Harrison.ai has announced the launch of its cutting-edge radiology-specific vision language model, Harrison.rad.1. The model is now being made available to selected industry partners, healthcare professionals, and regulators globally, aiming to ignite discussions about the safe, ethical, and responsible use of AI to transform healthcare access and capabilities, and ultimately, improve patient outcomes.
Robyn Denholm, Board Director at Harrison.ai, expressed her enthusiasm, stating, “The Harrison.rad.1 model is transformative and marks an exciting next step for the company. Harrison.ai is delivering on its promise to solve real-world problems more effectively and reliably, ultimately saving lives.”
Designed as a dialogue-based model, Harrison.rad.1 performs a range of tasks including open-ended conversations related to X-ray images, detecting and localizing radiological findings, generating detailed reports, and providing longitudinal reasoning based on clinical history and patient context. According to the company, clinical safety and accuracy remain top priorities.
Harrison.ai has previously introduced its radiology solution, Annalise.ai, which is currently deployed in over 40 countries, benefiting millions of patients each year. In the same spirit of dedication, Harrison.rad.1 will undergo rigorous and competitive evaluations by world-renowned professionals.
Dr. Aengus Tran, co-founder and CEO of Harrison.ai, emphasized the importance of data quality, modeling rigour, and ethical use in AI development, saying, “Based on these parameters, the Harrison.rad.1 model is groundbreaking.”
The model has been trained on a massive, real-world clinical dataset consisting of millions of images, radiology studies, and reports. It benefits from extensive annotations provided by medical specialists, ensuring clinically accurate training signals. As a result, Harrison.rad.1 is positioned as the most capable specialized vision language model in radiology to date.
Notably, the model has outperformed major large language models (LLMs) in the Royal College of Radiologists’ FRCR 2B exam by an impressive margin of approximately 2x. Tran reiterated the company’s commitment to responsible AI development, emphasizing the importance of open and competitive evaluations by professionals.
“Harrison.ai is committed to being a leading voice in the global conversation on the future of AI in healthcare. By making Harrison.rad.1 accessible to researchers, industry partners, regulators, and others, we aim to advance this conversation,” Tran concluded.
