Enhancing Blood Cancer Recognition through Deep Learning Techniques: A Comparative Study of Deep Learning Models for Blood Cancer

Abstract

For efficient treatment and positive patient outcomes, early and precise diagnosis of blood malignancies is crucial in the field of medical diagnostics. PBSs, or peripheral blood smears, are important diagnostic tools for several blood-related conditions. However, manually reviewing these photographs can be time-consuming and prone to mistakes. So, this study proposes three deep learning models such as Convolutional Neural Network (CNN), the Modified U-Net, and VGG16 for recognition of blood cancer. The dataset consists of PBS pictures from patients with acute lymphoblastic leukemia (ALL) is used to train the models. Furthermore, metrics like accuracy, recall, and F1-score are used to assess how well the models perform in classifying various blood cancer subtypes. The CNN model achieved an accuracy of 42%, while the Modified U-Net model demonstrated an accuracy of 52%. Remarkably, the VGG16 model outshines its counterparts with an impressive accuracy rate of 99%, underscoring the potential of deep learning for intricate medical image analysis.