DL-ALL: Deep Learning-Powered Acute Lymphoblastic Leukemia Classification

Acute leukemia, particularly acute lymphoblastic leukemia (ALL), is a lethal and fatal cancer that progresses rapidly if untreated. So, a prompt and precise diagnosis is critical to save numerous lives. This study aims to develop Computer Aided Diagnosis (CAD) system to help in the diagnosis of ALL using peripheral blood smear images. The study presents an accuracy of different hybrid models developed. The four models developed are fusion-based hybrid models combining ResNet50, VGG16, ShuffleNet, ViT and MobileNet; handcrafted features using wavelet transform; prompt embedding and machine learning classifiers like SVM, XGBoost and Random Forest. The publicly available Kaggle dataset is used for training and testing of the hybrid fusion model. Experimental results demonstrate that the fused architecture involving Resnet50, ViT, Wavelet and prompt embeddings framework achieved superior classification performance of 100% compared to all the other models developed. The findings highlight the boosting techniques in medical image analysis and the robustness of hybrid deep learning models, facilitating a way for more efficient and reliable leukemia diagnosis.