Quantitative study of total variation (TV) noise reduction algorithm with chest X-ray imaging

Abstract

The noise in chest X-ray image causes the decrease in the pulmonary nodule diagnosis accuracy and the increase in the misdiagnosis rate. For this reason, the noise decrease is very important in the diagnosis image, and total variation (TV) noise reduction algorithm was suggested and developed in a very effective method. In this study, we quantitatively evaluated and analyzed image performances as function of acquisition parameter in chest X-ray image using TV noise reduction algorithm. We performed simulation study with MATLAB and experimental study in chest X-ray image for evaluation of image performance. For that purpose, median filter, Anscombe's transform and proposed TV noise reduction algorithm were modeled to apply to each image. We acquired image with respect to the mAs (1.2, 3.6 and 5.9) at fixed 120 kVp. Also, we acquired image with respect to the kVp (70, 90 and 110) at fixed 3.8 mAs. Normalized noise power spectrum (NNPS), coefficient of variation (COV) and contrast to noise ratio (CNR) were used in this study. Both the simulation and experimental results, the noise was greatly improved compared to conventional noise reduction methods when using TV noise reduction algorithm so it was confirmed the NNPS, COV and CNR performances were all improved. In particular to analyze CNR, if TV noise reduction algorithm is used at 1.2 mAs and 70 kVp, it will be greatly improved compared to when using no technique at 5.9 mAs and 110 kVp, which will make a great contribution to the decrease in the exposure dose. In conclusion, as TV noise reduction algorithm can obtain the image performance excellently in the chest X-ray image, it is expected to be used substantially for the field of the diagnostic imaging.