Harnessing CNN and Smartphone Microscopy: A Mobile Application for Automated Leukemia Detection and Classification

Abstract

This paper presents a novel approach to leukemia detection and classification by integrating Convolutional Neural Networks (CNNs) with a custom-built smartphone microscope and a mobile application. Leveraging the portability of smartphones and the power of deep learning, we developed a cost-effective diagnostic tool designed to capture high-resolution blood smear images using a dual-magnification smartphone microscope and process them in real-time through a CNN, built with Keras, for the automated detection and classification of leukemia. The system is capable of classifying four primary types of leukemia: Acute Lymphoblastic Leukemia (ALL), Acute Myeloid Leukemia (AML), Chronic Lymphocytic Leukemia (CLL), and Chronic Myeloid Leukemia (CML) directly on the smartphone without the need for cloud-based services. This enables on-site diagnostics, particularly in resource-limited or remote settings. The proposed solution addresses the limitations of conventional laboratory microscopy and cloud-dependent systems, offering a portable, scalable, and accurate tool for medical professionals. Extensive testing with a novel blood smear dataset has demonstrated the efficiency of our approach, achieving high accuracy in both detection and classification tasks. By eliminating reliance on external infrastructure and focusing on mobile-based computation, this system brings affordable and accessible healthcare to the forefront, with potential applications in both clinical settings and field diagnostics. This work highlights the potential of combining smartphone microscopy and deep learning for early and accurate disease detection, representing a significant advancement in portable medical diagnostics.