Advancements in CT Image Reconstruction: An Exploration of Conventional and Deep Learning-Driven Approaches

This paper offers a comprehensive review of CT image reconstruction methods (FBP, CNN, ART, SART, ATV), tracing their evolution from traditional analytical techniques to recent deep learning-based approaches. It covers the principles of CT image reconstruction, including data collection and conventional methods like filtered back-projection and iterative reconstruction. The emergence of deep learning, particularly CNNs, is discussed, highlighting their potential for improving image quality, noise reduction, and clinical efficiency. The paper also addresses challenges, including the need for diverse training data and model interpretability issues, while emphasizing the importance of optimizing radiation dosage levels for patient safety.