An Intelligent Multisampling Tensor Model for Oral Cancer Classification
Chenxi Huang, Guokai Zhang, Sirui Chen, Victor Hugo C. de Albuquerque
Abstract
Oral cancer has been one of the most mortal diseases in the world, and accurate and timely treatment will efficiently improve the survival and cure rate of the patients. However, the traditional diagnosis ways by the clinicians could be laborious and easily misdiagnosed, and oral cancer usually with different morphological features, which makes it challenging to achieve the high accuracy classification automatically. To address this challenge, in this article, we propose an intelligent multisampling tensor model to achieve oral cancer and cyst classification from magnetic resonance imaging (MRI). Specially, our approach first encodes the input image by four simple sampling operations, which enables the model to learn more regional, global, influential, and correlative features, and then a representation fusion strategy is adopted to fuse those extracted representations. Afterward, those are contracted by sequences of matrix product states, which map the input representation into high dimensional space to conduct a classifier operation. Finally, we evaluate our proposed method on oral MRI cancer data, and the experimental result demonstrates that our approach could achieve competitive classification results.