多视角判别度量学习的乳腺影像检索方法

Multi-view metric learning with Fisher discriminant analysis and its applications for breast image retrieval

传统的医学影像检索使用单幅影像,但单幅影像中的影像信息有限,且不能有效利用不同角度拍摄的医学影像。为解决这一问题,提出了一种多视角判别度量学习的医学影像检索方法。基于Fisher判别模型在多个视角之间学习鲁棒的度量空间,使得相似的医学影像在度量空间紧密地映射,不相似的医学影像尽可能地彼此分离。同时,设置视角权重因子充分利用每个视角特征的不同表征信息。在“乳腺癌数字存储库”中与4种多视角方法比较,本文提出的方法检索准确率提高7%,识别率更高。

The traditional medical image retrieval uses a single image,but the image information in a single image is limited,and the medical images taken from different angles can not be effectively utilized.To solve this problem,a medical image retrieval method based on multi-view metric learning with Fisher discriminant analysis(MVML-FDA)is proposed.The MVML-FDA learns robust metric space from multiple views based on Fisher discriminant model.A robust metric space allows the medical images of the same class to be closely mapped in the metric space,and different classes of medical images are separated from each other as much as possible.At the same time,the view weighting factor is set to make full use of the different representation information of each view feature.According to the distance metric learning,the MVML-FDA is applied to the diagnosis of breast diseases.The experimental results on the Breast Cancer Digital Repository dataset show that the proposed method has better performance,and improves retrieval accuracy by 7%compared with four multi-view methods.