Hong Kong university to test four genAI models in hospitals

You May Be Interested In:Pharmacists Can Address Gaps in Long-Acting PrEP Care for HIV



The Hong Kong University of Science and Technology has recently announced its development of four large language models for healthcare.

WHAT IT’S ABOUT

Developed out of HKUST SuperPOD, the university’s AI supercomputing facility, the LLM-based tools include the following:

  • MOME: An AI model for identifying breast cancer pathologies in MRI images.

  • mSTAR: A pathology assistant tool that directly models whole slide images.

  • MedMr: A chatbot multimodal language model. 

  • XAIM: An explainable AI model that provides visual and textual explanations of the AI-generated analyses.

In a statement to Healthcare IT News, the HKUST research team further shared details and findings from developing these models. 

Their breast cancer AI model has achieved 87% diagnostic accuracy in multicentre tests. MOME can also predict patients’ responses to neoadjuvant chemotherapy. 

mSTAR, the team said, addresses “a wide array of pathological clinical issues, including seven types of clinical diagnostic and prognostic applications such as cancer subtyping and staging, metastasis detection, molecular prediction, survival analysis, and report generation.” 

The medical chatbot MedMR, which answers questions, generates medical reports, and provides initial diagnoses based on medical images, has achieved 93% accuracy in identifying tumours and non-tumours using the publicly available PCam200 dataset from Patch Camelyon.

Moreover, they shared that XAIM has attained 98.67% accuracy in diagnosing skin lesions on the PH² dataset from Portugal. 

The research team is now preparing to put these models to test with various hospitals in the city. “We are in discussions with several hospitals in Hong Kong regarding potential trials and implementations.” 

“Before the AI models are deployed in clinical practices, we will also need to collaborate with hospitals for large-scale, multicentre model validation to ensure their generalisability and reliability,” they added. 

THE LARGER TREND

Early this year, the Centre for Artificial Intelligence and Robotics in Hong Kong, a research centre under one of China’s national research institutes, the Chinese Academy of Sciences, also introduced a doctors’ chatbot called CARES. Built based on Meta’s Llama 2 LLM, the chatbot is being tested in seven unnamed hospitals in Beijing. 

Other Asian health systems have been working on similar generative AI projects. In October, the Singaporean Ministry of Health announced a new investment to support a national project that will roll out genAI in the public health system by end-2025. Docquity from Singapore is also assisting community health centres in West Java, Indonesia to adopt the genAI virtual assistant TehAI, supporting health workers in diagnosing tuberculosis, stunting, and hypertension.

share Paylaş facebook pinterest whatsapp x print

Similar Content

Drug Survival Rates in Pediatric Patients With HS Compared
Drug Survival Rates in Pediatric Patients With HS Compared
FDA finalizes AI-enabled medical device life cycle plan guidance
FDA finalizes AI-enabled medical device life cycle plan guidance
OpenAI's general purpose speech recognition model is flawed, researchers say
OpenAI’s general purpose speech recognition model is flawed, researchers say
Allergy medications: Know your options
Chili
Ambient AI doesn't improve efficiency across the board, study finds
Ambient AI doesn’t improve efficiency across the board, study finds
Trump taps former Rep. Doug Collins to head VA
Trump taps former Rep. Doug Collins to head VA
Frontline Reports | © 2024 | News