A robust noise reduction technique for time resolved CT | JoVE Visualize

Area of Science:

Medical Imaging
Image Processing
Computed Tomography

Background:

Time-resolved CT generates dynamic image series crucial for diagnosing conditions like cardiac arrhythmias and brain perfusion deficits.
Acquiring these dynamic datasets often requires compromises between temporal resolution, radiation dose, and image noise.
Significant patient motion during scanning further exacerbates noise and artifacts, complicating image interpretation.

Purpose of the Study:

To develop and evaluate a novel noise reduction technique, PArtial TEmporal Nonlocal (PATEN) means, for time-resolved CT data.
To assess PATEN's effectiveness in reducing image noise and improving image quality, particularly in the presence of patient motion.
To validate PATEN's performance in both cardiac CT and CT brain perfusion applications.

Main Methods:

The PArtial TEmporal Nonlocal (PATEN) means algorithm was developed, utilizing spatial and temporal information to identify similar pixel profiles for noise reduction.
PATEN was applied to 4D cardiac CT datasets with reduced radiation doses and to simulated low-dose CT brain perfusion datasets.
Image noise, CT number accuracy, contrast-to-noise ratio, and perfusion parameter accuracy were quantitatively and qualitatively assessed.
PATEN's performance was compared against filtered backprojection and time-intensity profile similarity (TIPS) bilateral filtering.

Main Results:

PATEN filtering significantly reduced image noise in reduced-dose cardiac CT phases by 53.8%, approaching the noise levels of full-dose images.
In CT brain perfusion, PATEN achieved a 55.8%-66.3% decrease in image noise for low-dose scans, improving contrast-to-noise ratio.
Neuroradiologists preferred PATEN-filtered images for anatomical differentiation and quantitative perfusion analysis compared to low-dose and TIPS methods.
Perfusion parameters derived from PATEN-filtered images showed good agreement (within 12-20%) with full-dose data.

Conclusions:

PATEN effectively reduces noise in time-resolved CT, significantly improving image quality in reduced-dose cardiac CT and enabling substantial dose reduction in CT brain perfusion.
The method demonstrates robustness to patient motion, a common challenge in dynamic CT imaging.
PATEN filtering enhances quantitative perfusion analysis and anatomical visualization, offering a valuable tool for low-dose, motion-affected dynamic CT studies.