Task-Driven IQA for Lung CT
Quantitative evaluation of image quality metrics for AI-based lung cancer detection (Duke Virtual Imaging Trials)
In collaboration with the Virtual Imaging Trials team at Duke University (Prof. Joseph Y. Lo), I analyze clinical and simulated CT images to study the false positives that arise in AI-based lung cancer detection. The central question is how image quality, acquisition, and harmonization affect downstream clinical tasks such as nodule detection and classification — connecting image-quality behavior to task performance rather than relying on generic visual metrics alone.