Evaluation of Low-dose Computed Tomography Images Reconstructed Using Artificial Intelligence-based Adaptive...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Radiation Dose Optimization

Background:

• Radiation dose reduction is crucial in medical imaging to minimize risks.
• Iterative Reconstruction (IR) and deep learning (DL) algorithms are key techniques for improving low-dose image quality.
• PixelShine (PS) is a vendor-neutral DL tool for denoising low-dose scans.

Purpose of the Study:

• To evaluate the diagnostic performance of PS-reconstructed images at various low radiation doses.
• To compare PS image quality against standard dose (SD) and other low-dose reconstruction methods.

Main Methods:

• Retrospective analysis of 85 patient cases.
• Qualitative and quantitative evaluation of low-dose images reconstructed with PS, IR, and filtered back projection (FBP).
• Doses were reduced to 70%-50% (FBP) and 50%-33% (IR) of SD.

Main Results:

• At 50% dose reduction, PS and IR yielded similar image quality and contrast-to-noise ratio (CNR).
• At 33% dose reduction, PS outperformed IR in image quality and CNR.
• Low-dose PS images (50% reduction) were equivalent to standard-dose FBP images.

Conclusions:

• AI-based denoising algorithms, such as PS, provide effective low-dose image reconstruction.
• PS demonstrates comparable performance to IR at 50% dose reduction and superior performance at 33% dose reduction.