Optimizing Medical Image Reconstruction with Incomplete Projection Data  

Author:

Inioluwa Nelson Aina

Co-Author:

Crismar Amaro

Mentor:

Younes Benkarroum

Abstract:

Radiation exposure from medical imaging poses long-term health risks, including DNA damage and increased cancer susceptibility. To mitigate these risks, clinicians often reduce radiation doses or limit scanning angles. However, this leads to incomplete projection data and degraded image quality in modalities such as CT, X-ray, and MRI. Reconstructing diagnostically useful images under these constraints remains a critical challenge in medical imaging. This study investigates the application of Filtered Back Projection (FBP) as a reconstruction technique for recovering high-quality images from incomplete projection data. We evaluate FBP’s performance limitations under sparse and missing data conditions and compare it against iterative reconstruction approaches designed to compensate for data gaps. Our results show that FBP offers clear advantages in computational speed and produces fine-grained noise characteristics compared to other deblurring methods. This allows for clear visualization of soft-tissue boundaries and fine anatomical structures even when projection data is limited. These findings suggest that FBP, while limited in certain incomplete-data scenarios, remains a strong baseline and can inform hybrid reconstruction strategies that balance speed, image quality, and clinical usefulness.