Deep Learning Denoising Improves CT Perfusion Image Quality in the Setting of Lower Contrast Dosing: A Feasibility...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Recent iodinated contrast media shortages necessitate reduced contrast protocols.
• Maintaining computed tomography perfusion (CTP) image quality with lower contrast doses is crucial.
• Artificial intelligence (AI) offers a potential solution for optimizing CTP protocols.

Purpose of the Study:

• To compare reduced vs. standard iodinated contrast media doses in CTP.
• To evaluate the impact of deep learning denoising on CTP image quality.
• To assess the effect of reduced X-ray milliampere-second (mAs) dose on CTP image quality.

Main Methods:

• Preclinical swine and clinical patient CTP studies were conducted.
• Reduced and standard iodinated contrast media doses were used.
• A residual encoder-decoder convolutional neural network (RED-CNN) was developed for denoising.

Main Results:

• Reduced contrast and mAs doses decreased CTP image SNR, with contrast dose being more impactful.
• Both k-space-weighted image average and RED-CNN denoising improved SNR.
• RED-CNN denoising yielded the highest SNR and was qualitatively rated as the best image quality.

Conclusions:

• Deep learning denoising significantly improves CTP image quality with reduced iodinated contrast media.
• AI-based denoising can help achieve diagnostic quality comparable to standard contrast doses.
• This approach may also enhance image quality in low mAs CTP acquisitions.