Task-Oriented Low-Dose CT Image Denoising

The extensive use of medical CT has raised a public concern over the radiation dose to patient. Reducing leads increased image noise and artifacts, which can adversely affect not only radiologists judgement but also performance downstream analysis tasks. Various low-dose denoising methods, especially recent deep learning based approaches, have produced impressive results. However, existing methods are all downstream-task-agnostic neglect diverse needs applications. In this paper, we introduce novel Task-Oriented Denoising Network (TOD-Net) with task-oriented loss leveraging knowledge from Comprehensive empirical shows that complements other task-agnostic losses by steering denoiser enhance quality in task related regions interest. Such enhancement turn brings general boosts on various for task. presented work may shed light future development context-aware methods. Code is available at https://github.com/DIAL-RPI/Task-Oriented-CT-Denoising_TOD-Net.