Optimizing input parameter of the randomized singular value decomposition algorithm for compressing technetium-99m...

Area of Science:

• Medical Imaging
• Computational Science

Background:

• Technetium-99m L,L, ethylenedicysteine (Tc-99m LLEC) renal dynamic (RD) studies are crucial for assessing kidney function.
• Image compression techniques can reduce data size but risk losing critical clinical information.
• Singular Value Decomposition Sketch (SVDsketch) offers adaptive matrix approximation using a tolerance (tol) parameter.

Purpose of the Study:

• To optimize the tolerance (tol) value for SVDsketch compression of Tc-99m LLEC RD studies.
• To achieve efficient compression while preserving essential clinical information.
• To minimize computation time for image compression.

Main Methods:

• Fifty Tc-99m LLEC RD studies were compressed using SVDsketch with varying tol values (0.00012, 0.1, 0.01, 0.05).
• Nuclear medicine physicians qualitatively assessed image fidelity at tol = 0.1.
• Quantitative image quality metrics (error, SSIM, brightness, GCF, CPP, blur) and split renal function accuracy were evaluated.

Main Results:

• Compressed images at tol = 0.1 were deemed identical to original images by expert physicians.
• Image quality metrics showed compressed images were less noisy, brighter, and had better contrast.
• There was no statistically significant difference in split renal function estimation.
• Compression achieved up to 74.53% with a computation time of 0.04725 seconds.

Conclusions:

• The optimal tol value for compressing Tc-99m LLEC RD studies is 0.1.
• This optimized compression effectively preserves clinical information and enhances image quality.
• SVDsketch provides a rapid and efficient method for Tc-99m LLEC RD study compression.