An Efficient Deep CNN-Based AML Detection:
Pradeep Kumar Das, Adyasha Sahu, Sukadev Meher · 2024
Acute myeloid leukemia (AML) is a potentially fatal hematological malignancy that necessitates accurate and prompt detection in order to provide effective treatment and enhance the well-being of patients. In this chapter, we have suggested an automatic AML detection system that makes use of the resilient EfficientNetB0 model, a cutting-edge deep learning architecture known for its outstanding performance in image recognition tasks. One of the major obstacles in creating accurate AML detection algorithms is the limited number of images available in existing AML databases, which can limit the system&s;s ability. To get rid of this issue, an augmentation approach has been carried out, named RandAugment. In addition, a reasonable trade-off between classification efficiency and computing cost is maintained by adaptive and uniform scaling of resolution, width, and depth, which results in an effective detection model. The experiments have been carried out in depth on the ASH database, a recognized standard for AML detection, to assess the effectiveness of the suggested approach. The experimental results reveal that the EfficientNetB0 model, when supplemented with RandAugment, shows the best AML detection performances with 96.25% accuracy, 97.44% precision, 97.50% specificity, 95% sensitivity, and 0.9620 F1 score.