The Hong Kong University of Science and Technology (HKUST) has unveiled four advanced AI models designed to revolutionize medical diagnostics and prognostics for over 30 types of cancers and diseases.

As a leader in AI research, HKUST has previously developed models for brain tumor prognosis, early detection of Alzheimer’s disease, and support in lung cancer surgeries.

One of the standout models, MOME, focuses on breast cancer diagnosis, utilizing 3D mpMRI fusion technology to accurately differentiate between malignant and benign lumps, achieving results comparable to seasoned radiologists.

MOME also predicts patient responses to neoadjuvant chemotherapy, offering a less invasive alternative to traditional biopsies.

The second tool, mSTAR, acts as a pathology assistant, enhancing the speed and accuracy of generating pathology reports by modeling whole slide images, thus enabling pathologists to efficiently perform up to 40 diagnostic tasks.

The third model, MedDr, is a multimodal language model that generates medical reports and initial diagnoses based on medical images, facilitating quicker clinical decision-making.

The fourth AI model, XAIM, serves as an explainable AI framework, helping healthcare professionals understand AI decision-making processes through visual and textual explanations, thereby promoting confidence in these technologies.

Project lead Prof. Hao Chen emphasized that the primary goal of these AI tools is to assist doctors in improving diagnostics and personalizing treatment plans while alleviating their workloads.

This announcement comes in light of findings that medtech companies may face a 4% annual revenue loss due to inefficiencies, which these AI innovations aim to address, particularly in reducing physician burnout.

Experts predict that AI will emerge as a significant economic driver in the medtech industry in the coming years, as highlighted in GlobalData's AI in Healthcare report.

These AI models are backed by HKUST's new AI supercomputing facility, which leverages extensive data sets and innovative machine training strategies to enhance their effectiveness.