CNN-Based Classification of Acute Myeloid Leukemia Blood Samples

Acute Myeloid Leukaemia (AML) diagnosis is often hampered by a lack of technology and time-consuming procedures. Manual interpretation of blood sample images by radiologists is error-prone and time-consuming. The objective of this research is: i) to develop a CNN model for AML classification, addressing challenges in obtaining high-quality training datasets. To predict and classify normal WBC and AML cells from microscopic peripheral blood cell images, a CNN classifier with AlexNet, SqueezeNet, MobileNetV2, and ResNet18 is used. Each classifier’s efficacy is assessed using evaluation metrics, confusion matrices, and ROC curves. ResNet18 outperformed the other models, identifying normal WBC and AML cells with an accuracy of 97.191% and F1-score of 0.97. This study demonstrates CNNs’ potential in AML diagnosis, including accurate classification and improved patient care.