TY - GEN
T1 - Deep Learning with Transfer Learning for Detecting Abnormalities in X-Ray Images
AU - Abd, Dina Jasim
AU - Jebur, Sabah Abdulazeez
AU - Alkhazraji, Lafta
AU - Alsammarraie, Riyam M.
AU - Hussain, Abir Jaafar
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Deep learning (DL) applications in medical imaging systems have revolutionized the detection of abnormalities, particularly in chest X-ray examinations. Current challenges in medical image analysis arise from insufficient data availability and class imbalance. Transfer learning (TL), an effective strategy that leverages pre-trained models to enhance performance and operational efficiency, addresses these limitations. This paper employs a modified InceptionV3 model trained on the COVID-19 Radiography Database-which includes four categories: COVID-19, Lung Opacity, Normal, and Viral Pneumonia-to classify chest X-ray images. The proposed method fine-tunes InceptionV3 by optimizing its final layers to improve feature extraction and classifier performance. The model achieved superior results, with an accuracy of 90.06%, recall of 88.02%, precision of 9 2. 1 9 %, and an F1-score of 89.75% in chest X-ray classification. This research demonstrates that transfer learning with InceptionV3 enhances abnormality detection in medical imaging, paving the way for reliable automated diagnostic systems.
AB - Deep learning (DL) applications in medical imaging systems have revolutionized the detection of abnormalities, particularly in chest X-ray examinations. Current challenges in medical image analysis arise from insufficient data availability and class imbalance. Transfer learning (TL), an effective strategy that leverages pre-trained models to enhance performance and operational efficiency, addresses these limitations. This paper employs a modified InceptionV3 model trained on the COVID-19 Radiography Database-which includes four categories: COVID-19, Lung Opacity, Normal, and Viral Pneumonia-to classify chest X-ray images. The proposed method fine-tunes InceptionV3 by optimizing its final layers to improve feature extraction and classifier performance. The model achieved superior results, with an accuracy of 90.06%, recall of 88.02%, precision of 9 2. 1 9 %, and an F1-score of 89.75% in chest X-ray classification. This research demonstrates that transfer learning with InceptionV3 enhances abnormality detection in medical imaging, paving the way for reliable automated diagnostic systems.
KW - COVID-19
KW - InceptionV3
KW - RFMiD dataset
KW - Transfer Learning
UR - https://www.scopus.com/pages/publications/105034165596
U2 - 10.1109/DeSE68208.2025.11367956
DO - 10.1109/DeSE68208.2025.11367956
M3 - Conference contribution
AN - SCOPUS:105034165596
T3 - Proceedings - 18th International Conference on Developments in eSystems Engineering, DeSE 2025
SP - 373
EP - 377
BT - Proceedings - 18th International Conference on Developments in eSystems Engineering, DeSE 2025
A2 - Obe, Dhiya Al-Jumeily
A2 - Assi, Sulaf
A2 - Mustafina, Jamila
A2 - Hussain, Abir
A2 - Jayabalan, Manoj
A2 - Radvan, Roxana
A2 - Bita, Bogdan
A2 - Tawfik, Hissam
A2 - Rowe, Neil
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 18th International Conference on Developments in eSystems Engineering, DeSE 2025
Y2 - 10 November 2025 through 12 November 2025
ER -