The effect of noise removal filters on the quality of digital radiography used low exposure technique: A phantom study

Abstract

Objective: To study the effect of noise removal filters on the quality of digital radiography used low exposure techniques.

Methods: The method began with finding an appropriate exposure technique for anthropomorphic chest phantom and subsequently decreasing the exposure technique, taking radiography, calculating SNR on digital radiographic images in bone regions and background regions. Next, filtering noise in the low exposure images by using 4 filters included with low pass filter, Gaussian filter, median filter and modal filter. All of images were assessed on image quality by 3 radiologists.

Results: The appropriate exposure technique for anthropomorphic chest phantom was 81 kVp 8 mAs. The low exposure techniques were 81 kVp and mAs values were 6.3, 4, 2 mAs. When we calculated the SNR on the digital radiographic images in bone region before using noise removal filters, we found the SNR were 70.29, 59.77, 55.85, and 45.71, respectively. In background region, the SNR were 14.29, 12.84, 12.28 and 10.69, respectively. The results of using four types of noise removal filters showed the SNR increasing. One way ANOVA showed no significant difference (p>0.05). However, filter which provided the highest SNR was modal filter. In bone region, the SNR were 99.56, 84.41 and 85.32, respectively. In background region, the SNR were 22.33, 43.37 and 21.63, respectively. The results of image quality assessment by 3 radiologists had similar values between low exposure images with filters and appropriate exposure images (ICC=0.929).

Conclusions: The result of the study demonstrated that the effect of noise removal filters of low exposure images affect an increase of SNR while maintaining images of adequate quality for diagnosis and also help patients receive less radiation dose by setting the lowest exposure technique 81 kVp 2 mAs.