题目：Neuroimage-based Diagnosis of Brain Disorders
主讲嘉宾： Dinggang Shen，美国北卡罗莱纳大学-教会山分校教授，放射学系IDEA实验室主任
时间： 2015年7月13日上午 10:30 – 11:30
地点： 深圳大学南校区医学院 附楼学术报告厅
主持人： 汪天富 教授
This talk will summarize our recent work on using novel machine learning techniques to predict or synthesize one modality image from other modality image(s). The typical applications, i.e., in PET/MRI scanner, include 1) predicting high-quality PET from both low-quality PET and MRI to reduce the tracer dose without degrading the PET image quality, and 2) predicting CT from MRI for attenuation correction of the acquired PET image. We have developed various sparse learning techniques to establish the internal relationships among different modality images and then use them for prediction or synthesis of target image. We have also used the structured random forest and auto-context model to learn the deep architecture for mapping source image(s) (i.e., MRI) to the target image (i.e., CT). All these techniques and their respective applications will be discussed in this talk.