The exact distinction of malignant leukocytes at minimal expenses during earlier phase of illness, which is a significant difficulty in disease diagnosis, is a real concern in the field of disease detection. Flow cytometry equipment is few, and the procedures available at laboratory diagnosis institutes are time- consuming, despite the high frequency of leukaemia. The current systematic review was undertaken to examine the works aimed at discovering and classifying leukaemia using machine learning, which was motivated by the possibilities of machine learning (ML) in disease detection. This research proposenovel technique in detection of Leukemia from smear blood images based on deep learning techniques. Here the input image has been processed for noise removal and image resize with smoothening the image. Then this processed image has been extracted using probabilistic neural network with lasso regression (PNN_LR) from which the images with leukemia has been extracted. The extracted image classification has been utilizing Support vector machine (SVM) in which the accuracy has been Enhanced from the classification output. For the test datasets, the method was capable of classifying leukaemia type containing accuracy, specificity, and sensitivity of 97.69 percent, 97.86 percent, and 100 percent, respectively, and for the validation datasets, 97.5 percent, 98.55 percent, and 100 percent, respectively. In addition, the system had a 94.75 percent accuracy rate for WBC counts, which included both lymphocytes and monocytes. In comparison to unsupported manual procedures, the computer-assisted diagnostic machine takes below 1 min to process and assign the leukaemia kinds.

leukemia, smear blood images, deep learning, classification, PNN_LR, SVM