Pneumonia detection with transfer learning

I built a pneumonia classification system using transfer learning on chest X-ray images, with a preprocessing pipeline tailored to noisy medical imaging: histogram equalization, erosion, dilation, and augmentation via OpenCV, TensorFlow, and Keras.

The model achieved 97.4% recall on test data across seven pneumonia types, which mattered more than headline accuracy in a screening context. The work was published at IEEE ICAIS 2021 (DOI: 10.1109/ICAIS50930.2021.9395895).

Focus

• Transfer learning for medical image classification
• Custom preprocessing for chest X-ray quality variation
• Peer-reviewed publication at IEEE ICAIS