BLOOD ONCOLOGY -REVOLUTIONIZING BLOOD CANCER DETECTION AND CLASSIFICATION WITH DEEP LEARNING

Abstract

Blood cancer affects the production and functioning of blood cells, making diagnosis difficult due to the presence of multiple subtypes and overlapping symptoms. This research proposes a novel blood cancer detection system based on a fine-tuned VGG16 Convolutional Neural Network (CNN) optimized using the Adam optimizer. The framework begins with an image preprocessing stage that includes standardization to a resolution of 224 × 224 pixels and data augmentation techniques such as rotation, flipping, and zooming to improve model robustness against data variability. The VGG16 model effectively extracts both low-level and high-level features such as boundaries, textures, shapes, and complex patterns from microscopic blood cell images, enabling accurate classification. The use of the Adam optimizer allows continuous learning rate optimization, ensuring faster convergence, improved generalization, and better handling of sparse or noisy datasets. The optimized VGG16 model is deployed through a Flask-based web application for real-time medical image analysis, achieving an accuracy of 96%. This system offers a precise, efficient, and user-friendly solution to assist healthcare professionals in the early detection and classification of blood cancer, thereby supporting improved diagnosis and treatment outcomes.