CT-based synthetic iodine map generation using conditional denoising diffusion probabilistic model.

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence

Background:

• Iodine maps from dual-energy CT (DECT) are valuable in radiology and radiation oncology.
• DECT scanners and iodine contrast agents have accessibility limitations.
• Alternative methods for generating iodine maps are needed.

Purpose of the Study:

• To develop a method for synthesizing iodine maps from non-contrast single-energy CT (SECT) images.
• Utilize a conditional denoising diffusion probabilistic model (DDPM) for synthetic iodine map generation.

Main Methods:

• Retrospective analysis of 126 head-and-neck patient scans (SECT and DECT).
• Generation of ground truth iodine maps from DECT using commercial software.
• Implementation and evaluation of a conditional DDPM, compared against GAN and CycleGAN.

Main Results:

• The DDPM achieved high accuracy with MAE of 1.039 ± 0.345 mg/mL, SSIM of 0.89 ± 0.10, and PSNR of 25.4 ± 3.5 dB.
• The proposed DDPM method outperformed reference GAN-based techniques.
• Statistical validation confirmed the superior performance of the DDPM.

Conclusions:

• A conditional DDPM framework can successfully generate synthetic iodine maps from non-contrast SECT images.
• This approach offers a viable solution for obtaining iodine maps when only SECT is available.
• Potential for broader clinical application in scenarios with limited DECT access or contrast agent contraindications.