時 間：113年5月14日（星期二下午14:10~15:00）
地 點：管理學院新大樓M240
主講人：楊欣洲 研究員兼所長(中央研究院 統計科學研究所)
講 題：Transforming Precision Healthcare: Harnessing Statistical Data Sciences and Artificial Intelligence to Unite Genetics and Medical Imaging for Advanced Disease Risk Prediction
Abstract
In the era of precision medicine, our study holds immense promise, tapping into the power of big data analytics from extensive biobank sources to predict disease risk. This study integrates statistical data sciences and artificial intelligence methodologies to comprehensively analyze genetic and medical imaging data obtained from the Taiwan Biobank and international collaborative projects. Innovative analysis techniques were developed, resulting in models with remarkable classification and prediction accuracy (with an area under the receiver operating curve exceeding 0.94). Our focus is on Type 2 diabetes, a multifactorial and polygenic inheritance condition. By employing convolutional-network deep-learning models for imaging analysis and statistical-learning models for genetic analysis, we fuse these modalities using machine-learning methods. Our findings from imaging analysis underscore the superiority of pixel-based analysis over feature-centric methods in terms of accuracy. Furthermore, the incorporation of multi-modality analysis enhances accuracy compared to single-modality approaches. By leveraging statistical data sciences and artificial intelligence for the holistic assessment of genetic, medical imaging, and demographic factors, our study offers potential contributions to the realms of early detection and precision healthcare for Type 2 diabetes, paving the way for a future where disease risk can be accurately predicted and managed.
Keywords: Type 2 diabetes (T2D), risk assessment, Taiwan Biobank, single nucleotide polymorphism (SNP), medical imaging, polygenic risk score (PRS), multi-image risk scores (MRS), eXtreme Gradient Boosting (XGBoost).
Joint work with Miss Yi-Jia Huang and Dr. Chun-houh Chen