Pulmonary cancer is one of the most malignancies worldwide. We introduce a novel three-dimensional Convolutional Neural Network (3D-PulCNN) to classify three pulmonary subtypes based on the microscopic hyperspectral image because it can supply spatial information as well as spectral information. Combining the difference of spectral curves and texture feature between the three subtypes, we can classify LUAD, LUSC and SCLC more accurately in cytodiagnosis.

Further details can be found in the article by Qing Zhang | Yan Wang | Song Qiu | Jiangang Chen | Li Sun and Qingli Li (e202100142).