A Cutting-Edge Ensemble of Vision Transformer and ResNet101v2 Based Transfer Learning for the Precise Classification of Leukemia Sub-types from Peripheral Blood Smear Images

Acute Lymphoblastic Leukemia (ALL) stands as the most prevalent form of cancer among children and its diagnosis predominantly involves microscopic blood evaluations of bone marrow. A quick and accurate diagnosis is crucial for effective treatment and better survival rates. The intricate challenge emerges from categorizing leukemia into distinct sub-types aligned with WHO standards. This task deviates from binary classification due to the striking similarity of inter-class features, leading to misclassification. In response, a ViT-CNN ensemble model was introduced in this study to aid in the automated diagnosis of ALL. The proposed ensemble architecture seamlessly integrated Vision Transformer (ViT) with Convolutional Neural Network (CNN) to achieve accurate classification of leukemia sub-types. The ViT-CNN ensemble model orchestrated the extraction of cell image features through two distinct pathways, culminating in enhanced classification outcomes. Leveraging a publicly accessible dataset comprising blood cell images adhering to WHO standards, this study empirically showcased the efficacy of the approach. The proposed ViT-ResNet101v2 ensemble model achieved an exceptional overall accuracy of 99.39%, outperforming numerous prior methods on this dataset. This achievement represents a significant advancement compared to the existing research on leukemia. The proposed methodology adeptly discriminated between leukemia sub-types, thus serving as an efficacious computer-aided diagnostic tool for ALL.