The Chinese University of Hong Kong

A team of engineers at The Chinese University of Hong Kong has developed an AI-based blood pressure management system featuring a large language model from DeepSeek. HOW IT WORKS The system called Dr PAI (personalised AI doctor) integrates wearable technology with DeepSeek to enable 24/7 monitoring and tracking of dynamic blood pressure changes and heart rate variability. It also collects multimodal data, including photoplethysmography signals, which it analyses to provide accurate, continuous readings and deliver health assessments, dietary advice, and exercise recommendations. Dr PAI can "'communicate' with patients like a real doctor, analyse long-term communication records and incorporate daily monitoring data to provide more reliable diagnostic references and treatment recommendations for patients and healthcare professionals," further explained CUHK professor Zhang Yuanting, who heads the research team. The beta version of the system is currently being tested in clinical trials. It is targeted for launch by yearend, based on a media release. WHY IT MATTERS Researchers emphasise the need for a cost-effective blood pressure management system, particularly for individuals living in poor and rural areas, amid rising cases of hypertension, which remains a major risk factor for cardiovascular diseases and stroke. "This open-source model can adopt a lightweight AI architecture that runs on low-computation devices, eliminating the need for expensive equipment," Prof Yuanting said about their genAI-driven solution. "This ensures that people in rural and remote areas can access intelligent blood pressure management system for free, promoting equitable access to health services and, in the long run, reducing the burden on healthcare systems." THE LARGER TREND DeepSeek LLMs, which have recently gained global traction for rivalling OpenAI's ChatGPT, are being increasingly applied in health IT systems across hospitals and IT providers in China. Healthcare SaaS provider ClouDr and Shenzhen University's South China Hospital are now running on DeepSeek AI. E-commerce platform Akso Health Group has recently announced DeepSeek integration to upgrade its diagnostic and medical assistant systems. Waterdrop's smart insurance service solutions are also runnin on DeepSeek.

Engineers from the University of Hong Kong have developed a generative AI-based imaging method that supports cancer diagnosis. HOW IT WORKS CytoMAD (Cyto-Morphology Adversarial Distillation) combines AI and microfluidic technology to do label-free imaging of tumours at the individual cell level. It can capture cell images "at great speed, enhance these images, and automatically correct inconsistencies in cell imaging. "We use [genAI] technology to render much clearer label-free images with useful information such as whether a treatment has had a positive effect," explained research lead and HKU professor Kevin Tsia. It also allows simultaneous brightfield to quantitative phase image translation to reveal additional cellular information. Dr Michelle Lo, a postdoctoral researcher at HKU and the main developer of CytoMAD, further explained that the technology can be trained to extract mechanical properties and molecular information of cells from brightfield images that are undetectable to the human eye. "In other words, we could uncover important properties of cells that underpin cell functions, bypassing the use of standard fluorescence markers and their limitations in costs and time." The research team demonstrated the use of the genAI imaging tool on lung cancer patients in a study with Li Ka Shing Faculty of Medicine and Queen Mary Hospital. The team is now working to secure funding for clinical trials of CytoMAD with lung cancer patients. They also intend to train the genAI tool to enable clinicians to predict cancer and other diseases. WHY IT MATTERS Single-cell imaging as part of cancer diagnosis and research can be lengthy and expensive. The process often applies stains and labels to samples to resolve low visibility. Results can also take time. "Until now, there is no cost-effective technique to do single-cell analysis through imaging mainly because of the limitation in scale. Under the traditional methods, the imaging throughput is not fast enough and the cell images are not clear and informative enough," noted Prof Tsia, who also heads HKU's Biomedical Engineering Programme. His team sought to address these limitations with their imaging innovation. According to HKU, besides its cost-effectiveness and fast imaging throughput, CytoMAD also allows unbiased cell image analysis since it does not require any data assumption, unlike current ML-driven methods requiring a priori knowledge of data. THE LARGER TREND Another university in Hong Kong has also developed healthcare large language models (LLM), including one for detecting breast cancer pathologies in MRI images. The research team at Hong Kong University of Science and Technology is now preparing for clinical pilots of these AI models with hospitals. Other Hong Kong-based institutes have also hopped on the genAI development trend. Late last year, the Chinese University of Hong Kong unveiled its genAI model for automated eye disease diagnosis called VisionFM. The Centre for Artificial Intelligence and Robotics created a medical chatbot named CARES Copilot 1.0, powered by Meta's Llama 2 LLM. Meanwhile, a radiology-specific LLM, developed by Australian company Harrison.ai, was introduced in September last year. Harrison.rad.1 was found to outperform popular models, including OpenAI's GPT-4o and Microsoft LLaVA-Med, in a competitive examination for radiologists.

Researchers from The Chinese University of Hong Kong and Beijing Tongren Hospital have developed a foundational AI model for automated eye disease diagnosis. WHAT IT'S ABOUT The generative AI model, VisionFM, was pre-trained on a large ophthalmic data cohort containing 3.4 million images from over 500,000 individuals worldwide. The pre-training dataset covers eight eye imaging modalities, various diseases, and clinical scenarios. VisionFM has been tested on an ophthalmic database comprising 53 public and 12 private datasets for multiple applications. These include eye disease diagnosis, progression prediction, systemic biomarker prediction through ocular imaging, intracranial tumour prediction, and lesion, vessel, and layer segmentation. Findings from this test have been published in NEJM AI of the New England Journal of Medicine. FINDINGS The study found that the genAI model demonstrated accuracy in diagnosing eye diseases comparable to an opthalmologist with between four and eight years of clinical experience. "In a comparative study of diagnostic accuracy of 12 ocular diseases from fundus photographs, VisionFM shows diagnostic accuracy close to that of intermediate-level ophthalmologists." It even accurately graded diabetic retinopathy (over 90%) using an imaging modality it was never exposed to during pre-training. Another interesting finding was VisionFM's ability to predict the presence of intracranial tumours from fundus images, which the researchers noted is a breakthrough. "While it is relatively hard for ophthalmologists and radiologists to infer the presence of an intracranial tumour from a fundus image, VisionFM can make accurate predictions directly from fundus images, and it is the first AI model that has been validated for this task." Additionally, the model can accurately predict glaucoma progression from fundus photographs. WHY IT MATTERS CUHK and Beijing Tongren Hospital claim their foundational model is a first of its kind that goes beyond the limits of similar existing AI models for automated ophthalmic diagnosis. These models, according to the researchers, rely on huge amounts of labelled data, which could be costly and time-consuming to collect. Many existing models also target a single or a few number of eye diseases and utilise only one imaging modality – particularly fundus photographs. Their significant finding of the model's ability to predict tumours directly from low-cost retinal images – a first among similar models – "holds great potential for the early detection in community and primary care," they emphasised. According to CUHK, VisionFM has been deployed in China's Henan Province to support the screening of common eye conditions. THE LARGER TREND Indeed, most AI tools for screening ophthalmic diseases utilise fundus images, or images of the back of the eye, captured through either professional or smartphone cameras. Such innovation has been reported in Taiwan, South Korea, Singapore and India. About four years ago in China, a multi-institutional research team also built a fundus-based AI system. Meanwhile last year, Google's AI model specifically intended for screening diabetic retinopathy was licensed to its partners in Thailand and India.