A Shallow Convolution Neural Network for Predicting Lung Cancer Through Analysis of Histopathological Tissue Images

Anjani Singh, Neha Sengar, Malay Kishore Dutta, Radim Burget

Abstract: Cancer remains an ailment that affects thousands of people resulting in the loss of life. Early diagnosis is essential for effective cancer treatment as cancer is not treatable once it has progressed to later stages. An optimized low-computational-cost shallow convolutional neural network-based framework for cancer diagnosis through analysis of histopathological tissue samples of the lung has been proposed. The proposed framework comprises a novel lightweight, efficient and robust diagnostic clas-sification network. The framework was trained on the LC25000 dataset and found to perform with an accuracy of 99.8% on the training set, 98.2% on the validation set, and 98.1 % on the test set. The performance of the proposed network was compared with other state-of-the-art networks and it was found that the proposed model was definitive for diagnosing unseen lung cancer images. The framework is capable of performing better than resource-intensive machine learning algorithms and can be deployed with limited computation and memory resource requirements.