Two-view topogram-based anatomy-guided CT reconstruction for prospective risk minimization.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Computational Anatomy

Background:

• Accurate prospective dose estimation for CT scans requires precise anatomical information.
• Current CT reconstruction methods struggle to accurately represent anatomical structures from limited projections.
• Minimizing patient risk in CT imaging necessitates advanced dose estimation techniques.

Purpose of the Study:

• To develop an optimized CT reconstruction model for generating accurate 3D volumes from minimal projections.
• To enhance the visualization and identification of anatomical structures within reconstructed CT volumes.
• To enable sophisticated patient risk-minimizing methods through improved spatial dose estimation.

Main Methods:

• A generative adversarial network (GAN) was trained to reconstruct 3D CT volumes from anterior-posterior and lateral projections.
• Integration of a pre-trained organ segmentation network and 3D perceptual loss to enhance anatomical details.
• Evaluation using standard metrics (PSNR, RMSE, SSIM) and organ segmentation performance (Dice score).

Main Results:

• The proposed GAN model achieved superior reconstruction quality with PSNR of 26.49, RMSE of 196.17, and SSIM of 0.64.
• Significant enhancement of organ shapes and boundaries, facilitating straightforward anatomical identification.
• Improved average organ Dice score of 0.71 compared to 0.63 for the baseline method, confirming enhanced anatomical accuracy.

Conclusions:

• The optimized GAN-based CT reconstruction method effectively reconstructs detailed 3D volumes from limited projections.
• The approach successfully enhances anatomical structures, outperforming conventional metrics and baseline methods.
• This technique holds promise for improving prospective dose estimation and patient risk management in CT imaging.