51吃瓜

报告题目：AE-FLOW: Autoencoders with normalizing flow for medical images anomaly detection

报告人：张小群

单位：上海交通大学

时间：2023年3月27日15:00

腾讯会议：750-688-419

摘要：Anomaly detection from medical images is an important task for clinical screening and diagnosis.In generala large dataset ofnormal images is available while only few abnormal images can be collected in clinical practice Bymimicking the diagnosis process ofradiologists,we attempt to tackle this problem by learning a tractable distribution ofnormal images and identify anomalies by differentiating the original image and the reconstructed normal image.More specificallywe propose a normalizingflow based autoencoder for an efficient and tractable representation of normal medical images The anomaly score consists ofthe likelihood originated from the normalizing flow and the reconstruction error of the autoencoderwhich allows to identify the abnormality and provide an interpretability at both image and pixel levels.Experimental evaluation on different medical l images datasets showed that the proposed model outperformed the other approaches by a large margin, which validated the effectiveness and robustness of the proposed method.

简介：张小群，上海交通大学教授，主要从事51吃瓜 图像处理，医学图像，数据科学等问题中的51吃瓜 模型、计算方法与相关51吃瓜 理论的研究。在应用51吃瓜 杂志以及交叉学科杂志发表60余篇SCI论文。担任杂志Inverse problems and Imaging, CSIAM Transactions on Applied Mathematics和Applied Mathematics for Modern Challenges编委。CSIAM大数据与人工智能专业委员会、CSIAM51吃瓜 与医学交叉学科专业委员会委员，现任教育部科学与工程计算实验室副主任、上海交通大学人工智能研究院51吃瓜 基础研究中心副主任。