In a groundbreaking announcement, four advanced AI medical models developed by the Hong Kong University of Science and Technology (HKUST) have emerged as innovative tools that can assist healthcare professionals in diagnosing up to 30 types of cancers and various other diseases. This revolutionary technology, including the newly introduced “medical ChatGPT,” promises to significantly streamline diagnostic processes, potentially reducing the time it takes to reach a diagnosis by as much as 40 percent.
During a recent media briefing, Chen Hao, an assistant professor in the HKUST Department of Computer Science and the lead researcher on this pioneering initiative, introduced MedDr, which serves as a versatile AI medical generalist that embodies the characteristics of a “medical GPT.” This sophisticated AI tool is designed to generate comprehensive reports and offer preliminary diagnoses through the analysis of medical images, enabling healthcare providers to submit close-up images for expert assessment and receive tailored treatment recommendations directly from the AI chatbot.
Chen expressed his aspirations for MedDr, stating that he hopes it can save medical professionals between 30 and 40 percent of their time. This substantial time reduction is expected to enhance their overall efficiency, allowing them to allocate more focus toward higher-value tasks within patient care.
To ensure the tool’s accuracy and reliability in real-world applications, Chen confirmed that HKUST will engage in collaborative efforts with local hospitals and consult with an extensive network of physicians.
According to HKUST, MedDr stands out as one of the largest open-source software initiatives available in the field of general medicine. A recent evaluation conducted by the Shanghai AI Laboratory has recognized MedDr as one of the foremost generalist models currently in existence.
Another noteworthy model, MOME, is specifically designed to tackle one of the most prevalent cancers affecting women in Hong Kong — breast cancer. Chen highlighted that MOME aids doctors in the analysis of breast MRIs, enabling them to quickly differentiate between benign and malignant tumors, which could potentially prevent unnecessary and invasive pathological biopsies.
This innovative model also possesses the capability to predict how patients will respond to chemotherapy, thereby assisting healthcare professionals in devising appropriate and effective treatment plans, Chen elaborated.
Remarkably, Chen noted that the diagnostic accuracy of this model approaches that of experienced radiologists, achieving a precision rate that exceeds an impressive 90 percent. This success was made possible through strategic partnerships with five hospitals across mainland China, coupled with the thorough analysis of over 10,000 patient cases.
Additionally, the suite of four models comprises specialized pathology assistant tools known as mSTAR and XAIM, which meticulously investigate the decision-making processes utilized by AI medical systems.
HKUST confirmed that the research underpinning these AI models was conducted utilizing the university’s advanced AI computing resources. The deployment of substantial computational power enables these sophisticated medical systems to effectively process extensive datasets, resulting in exceptional performance outcomes.
Guo Yike, the Provost of HKUST, elaborated during the media briefing that this initiative represents a significant step forward in the modernization of healthcare practices. He further emphasized that HKUST is actively pursuing the establishment of Hong Kong’s third medical school, marking the development of AI-integrated modern healthcare as a major strategic direction for the university’s future.
Published in Dawn, November 4th, 2024
**Interview with Chen Hao – Lead Researcher of HKUST’s AI Medical Models**
**Interviewer:** Thank you for joining us today, Dr. Chen. Your team’s recent work at HKUST has been making waves in the medical field. Can you explain to us what these AI models, particularly MedDr, are designed to do?
**Chen Hao:** Thank you for having me. MedDr is designed as a versatile AI medical generalist—essentially a “medical GPT.” Its primary function is to analyze medical images and generate comprehensive reports, offering preliminary diagnoses. By allowing healthcare providers to submit imagers for expert analysis, we aim to streamline the diagnostic process considerably.
**Interviewer:** That sounds incredibly promising. You mentioned that this technology could reduce the diagnostic time by up to 40 percent. How do you foresee this impacting patient care?
**Chen Hao:** We’re hopeful that this reduction in time will significantly enhance the efficiency of healthcare professionals. By offloading some preliminary evaluation tasks to the AI, doctors can focus more on higher-value activities such as patient interaction and comprehensive care planning. Ultimately, this could lead to faster treatments and improved patient outcomes.
**Interviewer:** Collaboration with local hospitals is key to ensuring the model’s reliability. Can you elaborate on how that partnership will work?
**Chen Hao:** Certainly. We plan to work closely with local hospitals and a network of physicians to conduct real-world evaluations of MedDr and other models. Their feedback will be critical in further refining the technology, ensuring that the AI meets the practical needs of medical professionals and patients alike.
**Interviewer:** It’s interesting to hear about the collaborative efforts. You also mentioned another model, MOME, focused on breast cancer diagnosis. What sets this model apart?
**Chen Hao:** MOME is specifically tailored to analyze breast MRI images, which is crucial for diagnosing breast cancer—one of the leading cancers affecting women in Hong Kong. This model provides targeted insights that can assist doctors in making more accurate diagnoses and tailored treatment plans for patients.
**Interviewer:** With MedDr recognized as one of the largest open-source initiatives in general medicine, what are your future aspirations for AI in healthcare?
**Chen Hao:** Our goal is to continue pushing the boundaries of what AI can achieve in medicine. We envision a future where AI assists healthcare professionals not just in diagnostics but in ongoing patient management and personalized treatment plans. By making our models open-source, we hope to encourage further innovation and collaboration within the medical community.
**Interviewer:** Thank you, Dr. Chen, for sharing these revolutionary advancements. It sounds like the future of AI in healthcare is very bright!
**Chen Hao:** Thank you for having me! I’m excited about the potential to improve patient care through these technologies.