Radiology genAI announced
Harrison.ai recently introduced a visual language model focused on radiology.
The interactive model, called Harrison.rad.1, enables free-form chat about x-ray images, detects and identifies radiological findings, and generates reports, with a focus on clinical safety and accuracy.
Unlike existing generative AI models that are trained on generic open-source data, Harrison.ai’s models are trained on diverse, unique, real-world data sets that are annotated at scale by a team of medical experts.
The company further claims that its genAI model outperformed leading LLMs such as OpenAI’s GPT-4o and Microsoft LLaVA-Med in the highly competitive British College of Radiologists’ 2B Rapids exam, scoring an average of 50.88 out of 60 points, on par with experienced radiologists.
Additionally, Harrison.rad.1 demonstrated 82% accuracy on the VQA-Rad benchmark dataset of clinical x-ray questions and answers, and 73% accuracy on Harrison.ai’s in-house open-source RadBench dataset.
See-Mode Receives 510(k) Grant for Thyroid Ultrasound Analysis AI
Victoria-based See-Mode Technologies has received the first 510(k) clearance from the U.S. Food and Drug Administration for its AI-powered solution that detects and diagnoses thyroid problems through ultrasound scans.
The thyroid ultrasound analysis software detects single or multiple nodules and automatically classifies each one based on the American College of Radiology’s TI-RADS grading system, automatically generates a complete worksheet for immediate review and correction by the radiologist or provides a preliminary impression after clinician review and approval, and streamlines reporting for thyroid follow-up exams.
New medical AI centre opens in Victoria
A new centre for innovation in healthcare AI has been established at La Trobe University in Victoria.
The Australian Centre for Medical Innovation and Artificial Intelligence, funded by A$10 million (US$6.8 million) from the state government, aims to develop AI to find new treatments, vaccines and immunotherapies for cancer, diseases and viruses.
Ongoing projects include an AI-powered color map to track and predict breast cancer progression, and a biosensor to detect cancer cells.
The media release also stated that the centre will be the first in Australia to have access to Nvidia’s DGXH200 supercomputer.
New Zealand seeks national medical AI implementation guide
A recent survey found that New Zealand general practices are calling for official national guidance on the adoption and use of AI in healthcare.
AI in Primary Care, a group of primary care organisations, surveyed GPs over the past two months, asking them how AI is being used in primary care across the country.
It was revealed that about a quarter of the approximately 300 respondents use AI on a daily basis in their work.
Meanwhile, Dr Janine Bycroft, founder and CEO of the Health Navigator Charitable Trust, said the survey results also showed organisations were seeking “some sort of recognition or framework from national authorities that could be a game changer in terms of public trust and acceptance”.
A comprehensive AI implementation guide by the WellSouth Primary Health Network is currently being provided to GP practices across the country. Hosting webinars and conducting privacy impact assessments are other ways primary care organisations are driving the adoption of AI within the sector.
