Unsupervised dimensionality reduction of medical hyperspectral imagery in tensor space

Hongmin Gao, Meiling Wang, Xinyu Sun, Xueying Cao, Chenming Li, Qin Liu, Peipei Xu

Abstract: Highlights•Our study proposed a novel unsupervised dimensionality reduction (DR) method of cholangiocarcinoma (CCA) microscopy hyperspectral pathological images. ​Particularly, entropy rate superpixel (ERS) segmentation algorithm is introduced to generate homogeneous regions. Low-rank collaborative graph weight matrix is constructed on homogeneous regions instead of all training samples, which greatly improve the efficiency and robustness of the proposed method. In addition, the whole process of DR is carried out in tensor space, which can greatly preserve the structure of hyperspectral data and avoid the loss of structure information via vectorization operation.