A Novel Algorithm For Performance Analysis Of Bayesian Convolutional Neural Networks In Detecting Acute Lymphoblastic Leukemia

Acute Lymphoblastic Leukemia is a bone hub and blood-related white platelet sickness that is dangerous. Blood testing done by hand is frequently inaccurate, sluggish, and time-consuming. Although methods for classification employing Convolutional Neural Networks have shown promise in the analysis of leukemia pictures, measuring their degree of uncertainty is still not achievable. Since a stochastic deep convolution neural network model frequently yields the best classification accuracy along an uncertainty estimate, we use a pre-trained Bayesian CNN model for automatic ALL detection that performs superior against over fitting on short datasets. The Bayesian model uses the Gaussian process' variational estimation method to approximate the probability distributions across the kernels. Using the classification of malignant cells as harmless, early, pre-, and pro-cancerous; an accuracy of 98.28% was obtained for ALL identification. Accuracy, Precision, Recall and F1 score information were utilized to direct the adequacy examination.