A Deep Learning Framework for Semantic Segmentation of Nucleus for Acute Lymphoblastic Leukemia Detection

Acute lymphoblastic leukemia is a form of blood cancer in which the bone marrow overproduces immature white blood cells. A novel semantic segmentation of nucleus for detection of Acute lymphoblastic leukemia is proposed here. The input images are obtained from public database ‘‘ALLIDB2’’. Resizing, SMOTE and Augmentation are carried out as preprocessing. After pre-processing, segmentation of nucleus is performed by SegNet and ResUNet. The performance of SegNet and ResUNet are compared. The segmented images are given as input to the classification models. Using Xception, Inception-v3 and ResNet50 models, the segmented images are classified as healthy and blast cells. It is found that Inception-v3 performs better than Xception and ResNet50 with an accuracy of 93.74%. This will be helpful to detect Acute lymphoblastic leukemia at the earliest.