Leukaemia BLOOD CANCER PREDICTION USING DEEP LEARNING

Abstract

Blood cancer, particularly Leukemia, is a serious disease that affects the blood and bone marrow. Early detection of leukemia is very important for improving patient survival rates and enabling timely medical treatment. Traditionally, leukemia diagnosis is performed manually by medical experts through microscopic examination of blood smear images, which can be time-consuming and prone to human error. This project proposes an intelligent deep learning–based system for automatic detection of leukemia from microscopic blood smear images. The system uses a Convolutional Neural Network (CNN) based on the EfficientNetB0 architecture to extract high-level visual features from blood cell images. These extracted features are further processed using the XGBoost machine learning classifier to improve prediction accuracy and classification reliability. The model is trained on a labeled dataset containing leukemia-infected and healthy blood cell images. Data preprocessing techniques such as image normalization, augmentation, and balanced sampling are applied to improve model generalization. The trained model is integrated into a REST API-based web system that allows Lab Technicians to upload blood smear images, Doctors to review predictions and approve reports, and Administrators to monitor system performance and analytics. The proposed hybrid CNN-XGBoost approach improves diagnostic accuracy and provides fast automated analysis of blood smear images. This system can assist medical professionals in early leukemia detection and help reduce diagnostic workload in hospitals and laboratories.