This repository contains the implementation code for my dissertation research on evaluating deep learning approaches for pneumonia detection from chest X-ray images. The study compares multiple convolutional neural network (CNN) architectures to assess their effectiveness in classifying pneumonia from chest radiographs.
Pneumonia remains a leading cause of morbidity and mortality worldwide, particularly in developing countries where access to expert radiologists is limited. This research explores the potential of deep learning models to assist in automated pneumonia screening, potentially reducing diagnostic delays and improving patient outcomes.
The repository includes implementations of four deep learning architectures:
| Model | File | Status | Description |
|---|---|---|---|
| Prototype CNN | prototype.ipynb |
✅ Complete | Simple baseline CNN used for feasibility study in Semester 1 |
| Baseline CNN | baseline_cnn.ipynb |
✅ Complete | Custom CNN architecture serving as primary baseline |
| ResNet | resnet_model.ipynb |
✅ Complete | Residual Network implementation with skip connections |
| DenseNet | densenet_model.ipynb |
✅ Complete | Densely Connected CNN with feature reuse |
| EfficientNet | efficientnet_model.ipynb |
✅ Complete | State-of-the-art model with compound scaling |
- Data Preprocessing: Image normalization, augmentation, and train/validation/test splitting
- Transfer Learning: Pre-trained weights from ImageNet for ResNet, DenseNet, and EfficientNet
- Performance Metrics: Accuracy, precision, recall, F1-score, and confusion matrices
- Comparative Analysis: Side-by-side evaluation of all architectures
- Python 3.8 or higher
- TensorFlow 2.x
- Jupyter Notebook / JupyterLab