Efficient Kidney Cancer Classification from CT Images Using a Lightweight Convolutional Neural Network Optimized with an Enhanced Crow Swarm Optimization Algorithm

Abstract

Kidney cancer is among the fifty most common cancers in the global statistics, therefore, early and accurate classification could enhance the prognosis. However, present classification models are more of a challenge to handle with data obtained from CT imaging. Our study proposes a lightweight and automated classification for kidney cancer detection using a hybrid feature extraction approach and a novel lightweight convolutional neural network improved by a hybrid Crow Swam Optimization (CSO) algorithm. Two datasets were used to develop and validate the model: the CT Normal – Kidney dataset containing 6,101 CT images and the CT Cyst, Tumor & Stone Kidney – Normal dataset comprising 6,345 CT images together and the Kidney Cancer dataset with 8,400 images. The technique used for feature extraction involved the use of multiple descriptors where useful image features were obtained. This was followed by optimising the Hybrid CSO algorithm with better results observed on augmented feature selection for better classification. The experiments’ outcomes were an accuracy of 100%, an F1-score of 97.49%, a Precision of 97.97%, a recall of 98.28% fast processing and the model’s successful differentiation of kidney pathologies. This more efficient and accurate framework, based on the application of both deep learning and conventional methods depending on levels of accuracy, opens up a valuable window on real-time kidney cancer classification that should directly assist radiologists in clinical diagnosis and raise detection reliability.