A Hybrid Approach Using Grey Wolf Optimization with Discrete Wavelet Transform and LeNet Features for Tuberculosis Classification in Chest X-Ray

The aim is to provide a hybrid methodology that fuses Discrete Wavelet Transform with LeNet-based Convolutional Neural Network features for the classification of chest X-ray images. DWT decomposes the images into frequency sub-bands, while feature extraction is covered by LeNet. These combined features are subjected to the evaluation of SVM, Random Forest, and AdaBoost classifiers. The results indicated enhanced classification performance regarding the detection of tuberculosis in chest radiographs. Feature selection can further be optimized by Grey Wolf Optimizer, leading to even better results. The optimized Random Forest classifier yielded an accuracy of 97.1 %.