Mask-Guided Attention Deep Learning Model for COVID-19 Diagnosis
Published:
Project overview
This research project focuses on developing a mask-guided attention deep learning model for automated COVID-19 diagnosis using CT scan images. The model leverages lung masks to guide the attention mechanism toward clinically relevant regions, allowing the classifier to focus on areas most indicative of COVID-19 infection.
The project integrates CT scan datasets from multiple public sources, forming a unified and diverse medical imaging dataset suitable for robust model training. This work was conducted as part of my Ph.D. research at the SMART Laboratory under the guidance of Dr. James Kong.
My role and tools
- Role: Model design, dataset integration, training pipeline development, evaluation, and manuscript preparation
- Tools: Python, PyTorch, NumPy, OpenCV, Scikit-learn, Google Colab, CUDA
What the project shows
- How to build a deep learning classifier using CT scan images
- How mask-guided attention improves diagnostic accuracy by highlighting clinically relevant regions
- How to construct and clean a multi-source medical imaging dataset
- How to evaluate model performance using accuracy, sensitivity, specificity, ROC curves, and Grad-CAM
- How to structure a reproducible research pipeline in Python for medical imaging problems
- How to prepare a research workflow that can be replicated and extended by other researchers
Links
📄 Research Manuscript (Submitted):
A Mask-guided Attention Deep Learning Model for COVID-19 Diagnosis💻 GitHub Repository:
Corona CT Classification – Source Code