Residual Network with Squeeze-And-Excitation Block for White Blood Cell Classification in Acute Myeloid Leukemia

Leukemia is a cancer that occurs when white blood cells are produced excessively in the bone marrow. Acute Myeloid Leukemia (AML) is a type of leukemia that targets myeloid stem cells, which will differentiate into white blood cells. The diagnosis of leukemia is performed by directly observing the morphology of white blood cells, but this method is time-consuming, energy-intensive, costly, and susceptible to errors. Recent research has explored deep learning methods for the automatic classification of AML white blood cells. This study proposes the use of ResNet-50 method with the integration of the Squeeze-and-Excitation (SE Block) to enhance the accuracy of AML white blood cell classification. The dataset used is The Munich AML Morphology Dataset, consisting of 18,365 AML white blood cell images. Data augmentation techniques are employed to address the imbalanced dataset. Experimental results show that SE-ResNet-50 model achieves an accuracy of 0.9743, precision of 0.98, recall of 0.97, and F1-score of 0.98. This result surpasses previous research that using ResNet-50 with an accuracy of 0.9657, precision of 0.97, recall of 0.97, and F1-score of 0.97.