Feature-Driven Acute Lymphoblastic Leukemia Detection From Blood Smears Using Machine Learning Ensemble Classifiers

This study delineates a machine learning-based method for using peripheral blood smear images to diagnose acute lymphoblastic leukemia early. The proposed method captures a wide range of image features by using various feature extraction techniques, including Gabor filters, Sobel edge detection, local binary pattern, and histogram of oriented gradients. The features were classified using Random Forest, Gradient Boosting, and AdaBoost, among which Random Forest reported the best accuracy of 91.00%. The study's contribution is its multi-feature extraction methodology that improves the diagnostic accuracy and model robustness. The proposed method presents a valuable and automated facility to support hematologists in identifying acute lymphoblastic leukemia correctly and in good time.