Tools
Tools
Govind, Pavan Kumar (2024) Pneumonia detection using Transfer learning. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (2MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (2MB) | Preview |
Abstract
Pneumonia is still a huge burden to the world, especially targeting children, the elderly and persons with a weakened immune system. Early diagnosis is critical in order to have a lower number of fatalities and better prognosis. This work examines the use of transfer learning in improving the identification of pneumonia from Chest X-ray images using Deep Learning models namely DenseNet121, EfficientNetB0, and ResNet50. They can utilize knowledge from other large datasets like ImageNet, through transfer learning, to overcome the two big problems of a limited annotated training set and the high computation that comes with it. The models were trained and tested using a chest X-ray dataset that is available to the general public, which incorporated data enhancement, early stopping mechanism, and the Grad-CAM technique for model interpretation. DenseNet121 appeared to be the most efficient on average, with a test accuracy 0.9822 while the other models also showed promising results with EfficientNetB0 and ResNet50 having the most significant difference from DenseNet121. To increase trust among clinicians, Grad-CAM visualization was incorporated, and the output was presented showing important areas in the X-rays that affected the models’ decisions. Of course, there were triumphs, but problems of different nature, including overfitting, variability of a dataset, and generalization of the model to non training data were revealed.
It is only until now that this study shows how deep learning and transfer learning can dramatically enhance diagnosis of pneumonia, especially in restricted healthcare facilities. However, for the model to become helpful in clinical practice, research should focus to enhancing the interpretability of the model, the model’s generalization ability across different populations, and clinical validation of obtained results.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Raj, Kislay UNSPECIFIED |
| Uncontrolled Keywords: | Pneumonia detection; deep learning; transfer learning; DenseNet121; EfficientNetB0; ResNet50; chest X-rays; Grad-CAM; model interpretability; dataset generalization |
| Subjects: | Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science R Medicine > Healthcare Industry Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HM Sociology > Information Science > Communication > Medical Informatics |
| Divisions: | School of Computing > Master of Science in Artificial Intelligence |
| Depositing User: | Ciara O'Brien |
| Date Deposited: | 19 Jun 2025 15:51 |
| Last Modified: | 19 Jun 2025 15:51 |
| URI: | https://norma.ncirl.ie/id/eprint/7949 |
Actions (login required)
 |
View Item |