Quantitative coronary angiography using image recovery techniques for background estimation in unsubtracted images

Area of Science:

• Medical imaging
• Quantitative angiography

Background:

• Digital subtraction angiography (DSA) is susceptible to misregistration artifacts from patient motion.
• Quantitative densitometry using subtracted images is clinically limited by motion artifacts.
• Unsubtracted images offer a potential alternative for accurate densitometry.

Purpose of the Study:

• To evaluate image recovery techniques for densitometry on unsubtracted coronary angiograms.
• To compare the accuracy and precision of Local Averaging (LA), Morphological Filtering (MF), Linear Interpolation (LI), and Curvature-Driven Diffusion (CDD) image inpainting.
• To assess the clinical feasibility of quantitative densitometry without background mask images.

Main Methods:

• Humanoid phantom and swine models were used for in vitro and in vivo studies.
• Iodinated vessel phantoms and coronary arteries were imaged.
• LA, MF, LI, and CDD were applied to estimate and remove background signals from unsubtracted images.
• Iodine mass quantification was performed and compared to known values and DSA measurements.

Main Results:

• LI and CDD achieved ~3% root mean square error in iodine mass quantification in both phantom and swine studies.
• LA and MF showed higher errors (~6-12%) and were deemed inadequate.
• LI and CDD demonstrated significantly lower standard deviations in iodine mass measurements compared to DSA, indicating higher precision.

Conclusions:

• Linear interpolation (LI) and curvature-driven diffusion image inpainting (CDD) are effective for quantitative densitometry using unsubtracted coronary angiograms.
• These techniques accurately estimate background signals, overcoming motion artifacts inherent in DSA.
• LI and CDD show promise for applications requiring precise quantification where background masks are challenging to obtain, such as lumen volume and blood flow analysis.