Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions
Abstract White blood cells (WBCs) play a crucial role in the immune system, protecting the body from infections and foreign invaders. Abnormalities in WBCs can be indicative of various conditions, including leukemia. WBCs classification is pivotal for diagnosing hematological disorders. This study a...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Big Data |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40537-025-01235-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849342190400045056 |
|---|---|
| author | M. Hussein Faten Abd El-Sattar Zahran El-Mougi |
| author_facet | M. Hussein Faten Abd El-Sattar Zahran El-Mougi |
| author_sort | M. Hussein |
| collection | DOAJ |
| description | Abstract White blood cells (WBCs) play a crucial role in the immune system, protecting the body from infections and foreign invaders. Abnormalities in WBCs can be indicative of various conditions, including leukemia. WBCs classification is pivotal for diagnosing hematological disorders. This study advances automated WBCs analysis through an 8-class classification framework encompassing rare but clinically critical subtypes: neutrophils, eosinophils, basophils, lymphocytes, monocytes, immature granulocytes (IGs), erythroblasts, and platelets. Leveraging a dataset of 17,092 CellaVision DM96-generated images standardized for clinical relevance, we implement rigorous preprocessing (normalization, resizing) and dynamic augmentation (rotations, flips) to enhance robustness. Six architectures are evaluated: ResNet50, InceptionV3, EfficientNetB3, MobileNetV3, Swin Transformer, and a custom convolutional neural network (CNN). ResNet50 emerged as the top performer 98.83% accuracy, followed by InceptionV3 98.77% and Swin Transformer 98.71%, demonstrating the efficacy of transfer learning and transformer-based attention mechanisms. Class-weighted loss mitigated dataset imbalance, achieving > 0.98 F1-scores for 6/8 classes. Computational efficiency analysis revealed MobileNetV3 as optimal for deployment (3.43 ms/inference). The study addresses key challenges—class imbalance, model interpretability via Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations—and validates improved diagnostic precision over prior work. By integrating clinically critical subtypes and state-of-the-art architectures, it provides a robust tool for medical education and practice, enabling early detection of leukemia, sepsis, and myelodysplastic syndromes. This study can enhance the training of medical students and doctors, equipping them with better tools for diagnosis and decision-making. Furthermore, the ability to classify a broader range of WBCs types could lead to more accurate and early diagnoses of diseases, ultimately improving patient care. |
| format | Article |
| id | doaj-art-0787e90e0b7e45b599cdf3b420128b41 |
| institution | Kabale University |
| issn | 2196-1115 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | Journal of Big Data |
| spelling | doaj-art-0787e90e0b7e45b599cdf3b420128b412025-08-20T03:43:27ZengSpringerOpenJournal of Big Data2196-11152025-07-0112113110.1186/s40537-025-01235-1Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutionsM. Hussein0Faten Abd El-Sattar Zahran El-Mougi1Department of Computer Science, Faculty of Specific Education, Mansoura UniversityDepartment of Computer Science, Faculty of Specific Education, Mansoura UniversityAbstract White blood cells (WBCs) play a crucial role in the immune system, protecting the body from infections and foreign invaders. Abnormalities in WBCs can be indicative of various conditions, including leukemia. WBCs classification is pivotal for diagnosing hematological disorders. This study advances automated WBCs analysis through an 8-class classification framework encompassing rare but clinically critical subtypes: neutrophils, eosinophils, basophils, lymphocytes, monocytes, immature granulocytes (IGs), erythroblasts, and platelets. Leveraging a dataset of 17,092 CellaVision DM96-generated images standardized for clinical relevance, we implement rigorous preprocessing (normalization, resizing) and dynamic augmentation (rotations, flips) to enhance robustness. Six architectures are evaluated: ResNet50, InceptionV3, EfficientNetB3, MobileNetV3, Swin Transformer, and a custom convolutional neural network (CNN). ResNet50 emerged as the top performer 98.83% accuracy, followed by InceptionV3 98.77% and Swin Transformer 98.71%, demonstrating the efficacy of transfer learning and transformer-based attention mechanisms. Class-weighted loss mitigated dataset imbalance, achieving > 0.98 F1-scores for 6/8 classes. Computational efficiency analysis revealed MobileNetV3 as optimal for deployment (3.43 ms/inference). The study addresses key challenges—class imbalance, model interpretability via Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations—and validates improved diagnostic precision over prior work. By integrating clinically critical subtypes and state-of-the-art architectures, it provides a robust tool for medical education and practice, enabling early detection of leukemia, sepsis, and myelodysplastic syndromes. This study can enhance the training of medical students and doctors, equipping them with better tools for diagnosis and decision-making. Furthermore, the ability to classify a broader range of WBCs types could lead to more accurate and early diagnoses of diseases, ultimately improving patient care.https://doi.org/10.1186/s40537-025-01235-1Deep learningTransfer learningWhite blood cellsConvolutional neural networksMedical institutionsSwin transformer |
| spellingShingle | M. Hussein Faten Abd El-Sattar Zahran El-Mougi Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions Journal of Big Data Deep learning Transfer learning White blood cells Convolutional neural networks Medical institutions Swin transformer |
| title | Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions |
| title_full | Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions |
| title_fullStr | Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions |
| title_full_unstemmed | Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions |
| title_short | Integrating deep learning and transfer learning: optimizing white blood cells classification in medical educational institutions |
| title_sort | integrating deep learning and transfer learning optimizing white blood cells classification in medical educational institutions |
| topic | Deep learning Transfer learning White blood cells Convolutional neural networks Medical institutions Swin transformer |
| url | https://doi.org/10.1186/s40537-025-01235-1 |
| work_keys_str_mv | AT mhussein integratingdeeplearningandtransferlearningoptimizingwhitebloodcellsclassificationinmedicaleducationalinstitutions AT fatenabdelsattarzahranelmougi integratingdeeplearningandtransferlearningoptimizingwhitebloodcellsclassificationinmedicaleducationalinstitutions |