LeukemiaVisionNet-21 Deep Neural Network for Automated Acute Lymphoblastic Leukemia (ALL) Classification
S Harithanush, Kappeta Poojitha, R Shalini, B.S. Vandana, R. Nithya · 2025
Acute lymphoblastic leukemia (ALL) is a blood cancer that affects the people of all ages. ALL disease is caused by immature development of too many lymphocytes which are generated in bone marrow. Early detection and treatment is essential to save the life of the many individuals. This disease affects more children than adults. Blood smear microscopy is a general screening procedure to examine the blood sample for the measurement of total blood cells count, size and shape. These blood cell measures are helpful to identify the cancer and other blood disorders. The manual blood cell examination is performed by the pathologist but it may leads to missing the abnormality findings. It is essential to develop a computer aided diagnosis (CAD) system to automate the blood cell screening process. In this study, novel LeukemiaVisionNet-21 convolution neural network model is proposed which integrates 21 layered customized CNN and extreme gradient boost (XGBoost) classifier. The efficiency of the proposed CNN architecture is compared with the deep network transfer learning models such as AlexNet, VGG16, ResNet-50 and DarkNet-53. 500 benign and 500 malignant peripheral blood smears (PBS) images from the kaggle dataset are randomly selected to train and test the proposed methodology for automated ALL classification. The proposed LeukemiaVisionNet-21 model obtained the classification accuracy of 99.5% and outperforms the transfer learning models in terms of classification accuracy.