HPC

medicine

AI-based earlier medicine development leveraging TWCC HPC to aid cancer prediction research

AI-based earlier medicine development leveraging TWCC HPC to aid cancer prediction research

Artificial Intelligence (AI) is shaping the future of global medical industries. The practice of medicine is changing with the development of AI methods of machine learning. As the increasing accuracy of predictive medicine, AI technology, based on analyzing patient’s medical records, is entailing predicting the probability of disease in order to either further diagnosis of disease allowing for the estimation of disease risks or significantly decrease the cost to deal with its impact upon the patient. The AI based prediction medicine is a new type of earlier medicine

Hsuan-Chia Yang, assistant professor of the Graduate Institute of Biomedical Informatics, Taipei Medical University, explains Prediction of Principle Health Threat (PROPHET) project. Led by Dr. Li Yu-Chuan, a pioneer of AI in Medicine and Medical Informatics Research, earlier medicine for fatal diseases is leveraging AI technology and data mining systems to provide a personal, real-time, accurate and manageable healthcare program. The PROPHET project provides the prediction of cancer risks and boosts the new business opportunity of start-ups. Taiwan Ministry of Science and Technology provides the funding support for this kind of projects.

Taking breast cancer detection as an example, there are 5 persons confirmed as positive out of every 1000 people screening. Applying the AI earlier medicine perdition method, the effective rate will be reduced to 5 confirmed out of 233 people check. There are 77% saving of breast cancer earlier diagnosis. The saved cost is obvious.

The basic of PROPHET project is making AI Bio-maker model using AI technology to screen cancer and provide the prediction. Transforming the patient medical records to time matrix data diagrams, the skill is setting to predict 10 kinds of cancer risks after one year time frame based on sequential medical records to develop a prediction model. Each prediction of various cancers could reach 85% AUROC (Area under the receiver operating characteristic) curves. Taiwan Healthcare insurance program preserves every citizen’s healthcare digital records of treatments and medicine usage. PROPHET takes this strength to analyze three-year personal data records to predict the cancer risks of next 12-month. These lower cost AI-based cancer predictions allow healthcare professions to participate in the decision about whether or not it is appropriate testing or detection priority for patients.

From the technical point of view, the dynamic prediction value of personal diseases is a time-dependent scenario. The time matrix combined with personal medicine usage records and various diseases could make a two dimensional health diagram. The vertical axis is thousands of variables including medicine usage, set of medical signs and symptoms. The horizontal axis is time listings based on week or month. There are about 250 thousand health diagrams to use in the AI training process to get effective prediction AI models. After requiring repeat fine-tuning in training new AI models of each cancer, it can be derived effective prediction models based on above AI Bio-marker.

However, the huge compute power to perform these AI training tasks requires huge support

Read More