This is the best radiology-specific vision language model launched to date

Haymarket, New South Wales, Australia-based Harrison.ai has unveiled its newest innovation: Harrison.rad.1, a radiology-specific vision language model — which is poised to transform how AI is applied in the healthcare sector.

Unlike traditional AI models, Harrison.rad.1 has been designed to handle complex, real-world clinical data and is being shared with select healthcare professionals, industry partners, and regulators in an effort to foster dialogue around the safe, ethical, and responsible use of AI in healthcare.

How does it work?

Harrison.rad.1 is a specialized vision language model created to assist radiologists and other medical professionals by performing a wide range of tasks related to X-ray images. It is capable of open-ended dialogue, detecting and localizing radiological findings, generating detailed reports, and providing longitudinal reasoning based on clinical history and patient context.

The model has been trained on a massive dataset that includes millions of images, radiology studies, and reports. These datasets were further annotated by a team of medical specialists, ensuring that Harrison.rad.1 receives clinically accurate training signals.

The model's standout feature is its superior performance in standardized exams. Specifically, Harrison.rad.1 outperformed major language models (LLMs) on the Royal College of Radiologists' 2B examination — a benchmark test for radiologists — doubling the average scores of its competitors. This capability makes it the most advanced and specialized vision language model in radiology to date.

Why does it matter?

The launch of Harrison.rad.1 represents a significant advancement in the use of AI for healthcare, addressing one of the sector's most pressing challenges: the shortage of qualified radiologists and the increasing demand for diagnostic imaging.

As Robyn Denholm, Harrison.ai Board Director, notes, "Harrison.rad.1 is transformative and an exciting next step for the company. Harrison.ai is delivering on the promise of helping solve real-world problems more effectively and reliably and helping to save lives."

By making the model accessible to healthcare professionals and regulators, Harrison.ai aims to initiate a broader conversation about the ethical use of AI in healthcare. Its model aims to set new standards for clinical safety and accuracy, critical factors in any healthcare innovation. Furthermore, Harrison.rad.1's ability to outperform human radiologists in rigorous exams underscores its potential to complement and enhance human expertise in medical imaging.

The context

AI in healthcare is a rapidly evolving field, but its application has often been limited by concerns about safety, reliability, and regulatory compliance. Most existing generative AI models are generic, trained on open-source data that may lack clinical relevance. In contrast, Harrison.rad.1 has been developed with a focus on real-world medical applications, using proprietary clinical data and annotations from medical specialists. This approach has led to a model that meets and exceeds the benchmarks set by human radiologists.

Dr. Aengus Tran, co-founder and CEO of Harrison.ai, emphasizes the importance of quality and ethics in AI development: "AI's promise rests on its foundations - the quality of the data, rigor of its modeling, and its ethical development and use. Based on these parameters, the Harrison.rad.1 model is groundbreaking."

By engaging in further evaluations and inviting global professionals to assess its model, Harrison.ai is demonstrating its commitment to responsible AI development and contributing to the ongoing global conversation about the future of AI in healthcare.