Comparative Analysis of CNN Models for ALL Diagnosis

Abstract—Acute Lymphoblastic Leukemia (ALL) is a serious pediatric cancer where early and accurate diagnosis is vital for improving treatment outcomes. This project explores the use of Convolutional Neural Networks (CNNs) to develop an effective diagnostic system for identifying ALL from microscopic blood smear images. Leveraging the ALL IDB1 dataset, six pre-trained CNN models—VGG16, VGG19, ResNet101V2, DenseNet201, Mo- bileNetV2, and InceptionV3—were fine-tuned and assessed based on key performance metrics including accuracy, precision, recall, F1-score, and AUC-ROC. To improve model robustness and reduce overfitting, data augmentation techniques were applied. A comparative evaluation highlighted the strengths and weaknesses of each model, as well as their potential for real-world clinical use. The findings demonstrate the power of deep learning in medical image analysis, offering faster, more affordable, and accurate alternatives to traditional diagnostic methods. This research not only validates the role of CNNs in ALL detection but also moves closer to integrating AI-driven diagnostics into everyday clinical practice, enhancing the accessibility and reliability of healthcare delivery.