On September 4, Harrison.ai announced the launch of a new AI model called Harrison.rad.1, designed specifically for radiology. This model can chat about imaging, spot and highlight findings on X-rays, write reports, and explain its reasoning using patient history and clinical information. Unlike other AI models, Harrison.rad.1 was trained using millions of real-world radiology images, studies, and reports.
Robyn Denholm, a board director at Harrison.ai, called the model a “transformative and exciting next step” for the company. She emphasized that Harrison.ai aims to solve real-world problems more effectively and save lives with this new technology.
When tested, Harrison.rad.1 performed well on the Fellowship of the Royal College of Radiologists (FRCR) 2B Rapids exam, scoring 51.4 out of 60 on its first attempt—better than the average score of radiologists taking the exam for the second time. In contrast, other AI models like GPT-4, Gemini-1.5 Pro, and LLaVA-Med scored below 30 on average.
Aengus Tran, MD, co-founder and CEO of Harrison.ai, described the model as “groundbreaking” for radiology. “We are excited about its performance so far. It significantly outperforms other major AI models in the radiology exam,” Tran said. He also mentioned the company’s commitment to responsible AI development, highlighting plans for further evaluation by healthcare professionals.
Currently, Harrison.rad.1 is available to select industry partners, healthcare professionals, and regulators to promote discussions on the responsible use of AI in medicine. Tran added, “Harrison.ai is dedicated to being a leading voice in shaping the future of AI in healthcare.”
Source: healthimaging
