Deep Learning Approaches for Intraoperative Pixel-based Diagnosis of Colon Cancer Metastasis in a Liver from Phase-contrast Images of Unstained Specimens

摘要

There is a need for computer-aided diagnosis (CAD) systems to relieve the workload onpathologists. This seems to be especially important for intraoperative diagnosis during surgery, forwhich diagnostic time is very limited. This paper presents preliminary results of intraoperativepixel-based CAD of colon cancer metastasis in a liver from phase-contrast images of unstainedfrozen sections. In particular, two deep learning networks: the U-net and the structured autoencoderfor deep subspace clustering, were trained on eighteen phase-contrast images belonging to fivepatients and tested on eight images belonging to three patients. Spectrum angle mapper was alsoused in comparative performance analysis. The best result achieved by the U-net yielded balancedaccuracy of 83.70%±8%, sensitivity of 94.50%±8%, specificity of 72.9%±8% and Dice coefficientof 45.20%±25.4%. However, factors such as absence of tissue fixation and ethanol-induceddehydration, melting of the specimen under the microscope and/or frozen crystals in the specimencause variations in quality of phase-contrast images of unstained frozen sections. This, in return,affects reproducibility of diagnostic performance.