Comparison of Deep Learning and Machine Learning Techniques for Automated Diagnosis of Acute Lymphoblastic Leukemia
Zeina Mahmoud, Mohammed Abdulmuttaleb Abdullah · Maǧallaẗ al-handasaẗ al-rāfidayn · 2025
Acute lymphoblastic leukemia (ALL) presents major diagnostic and categorization challenges due to its wide range of clinical manifestations and the invasive nature of existing diagnostic procedures. In this work, we examine how Deep Learning (DL) and Machine Learning (ML) approaches can be used to enhance ALL diagnosis and classification using bone marrow images. We do an extensive investigation of the performance of various DL and ML models, such as Convolutional Neural Network (CNN), Feed-forward neural networks (FNN), Naive Bayes (NB), and Decision Trees (DT), using the ALL-IDB1 dataset. A developed feature extraction method based on VGG16 is proposed with feature selection based on recursive feature elimination. Our study includes fine-tuning pre-trained models, feature extraction with VGG16, and model optimization. F1 score, accuracy, and recall measures are used to assess the performance of the model. The investigation produced encouraging results, with both DL and ML models recording 100\% accuracy and excellent classification scores. Additionally, it is confirmed that automated systems based on DL and ML models have the potential to improve patient outcomes by speeding up and improving diagnosis accuracy.