Predictive analysis on histopathological images using metaheuristics and machine learning method

The computer-based diagnosis system using histopathology images has always been the centre of research and improvement. Automated histopathological image classification systems have advanced to greater heights as a result of development in digital histopathology for computer-aided diagnosis. Due to the complexity of these images, a high dimension feature map is generated that makes the process difficult. The usage of metaheuristics in the classification of these images has grown due to their remarkable results. To conduct predictive analysis on histopathology images using nature-inspired algorithms and machine learning techniques. The feature extraction methods are applied to determine the statistical and texture-based feature. Furthermore, whale optimisation algorithm (WOA), cat swarm optimisation (CSO), lion optimisation algorithm (LOA), adaptive particle swarm optimisation (APSO), golden eagle optimisation (GEO), hybrid LOA_CSO (H-LOA_CSO) are used as feature selection algorithm for acquiring optimal subset of features. The classification are performed using support vector machine (SVM) and artificial neural network (ANN) to determine the predictive analysis. The observations show that the H-LOA_CSO algorithm performed best with ANN, giving an accuracy of 98%, while APSO showed the worst performance with both ANN and SVM with an accuracy of 72%.