Study of Mammography Medical Imaging Sample Selection Based on CSAL
Jiyun Li, Wenlong Mi · 2019
Medical imaging is widely used in the diagnosis and treatment of clinical practice. At present, medical imaging technology has entered a new era of digitalization. In order to solve the scarcity of medical image standard data sets and the high cost of manual labeling, it is necessary to use active learning methods to select the most valuable data from the vast image data pool for labeling which the purpose is to make the labeling costs lower, reduce the pressure on professional physicians and make clinical decision-making easier. This article focuses on mammography medical imaging, analyzes the structure and characteristics of the images, deals with mammography medical imaging data and trains the model using machine learning model, called CSAL (CNN-Seq2Seq-Attention + Active Learning), in order to reduce the loss of image data sequence after graying, and filter out valuable data from the data pool by active learning. This paper uses two different active learning methods and two different data sets for comparative experiments. The finally experimental results achieved satisfactory effort.