EfficientNet-B3

Acute Lymphoblastic Leukemia (ALL), the most common pediatric cancer, requires early detection, and this study evaluates deep transfer learning as a faster, more efficient alternative to traditional diagnostic methods. Using over 10,000 blood cell images, EfficientNet-B3 significantly outperformed VGG-19, achieving 96% accuracy and superior performance on minority classes, while VGG-19 showed limited recall and F1 scores. The results support EfficientNet-B3 as a promising tool for early ALL detection, though broader datasets and multimodal integration are needed to ensure clinical generalizability.