Yui Hayashi, Presenter: Nothing to Disclose

Yukiyoshi Kimura BS, Abstract Co-Author: Nothing to Disclose

Naotoshi Fujita, Abstract Co-Author: Nothing to Disclose

Yoshie Kodera PhD, Abstract Co-Author: Nothing to Disclose

Maki Yamada BS, Abstract Co-Author: Nothing to Disclose

Akiko Horii, Abstract Co-Author: Nothing to Disclose

We found that the correlation coefficient of SNRi was the highest among the three SNR types. SNRi, which incorporated the visual characteristics, explained the visual image quality well. Therefore, we conclude that SNRi is useful for evaluating.Because SNRi which includes human visual characteristics is close to the visual image quality, we can evaluate image quality more precisely and save people the trouble of doing perceptual evaluation.

The effects of various imaging parameters on detectability have not yet been clarified, and image quality indices do not equate with visual image quality. We attempted to investigate the relationship between three types of SNR and the signal detection rate.

We calculated three types of SNR corresponding to the amplitude model (SNRa), matched filter model (SNRm), and internal noise model (SNRi). SNRa is calculated from the amplitude of the signal and the standard deviation of the noise. SNRm depends only upon the physical characteristics of the image. SNRi takes into consideration the visual spatial frequency response of the human visual system and the internal noise in the eye-brain system. We employed a 16-AFC procedure for the observer performance experiments. The observer is asked to choose one of 16 boxes as the signal location in a test image. The test image involved 15 different imaging cases. We displayed the test images on a 5-megapixel liquid crystal display (5-MP LCD; Totoku) and examined the detection performance. The signal detection rate was defined as the number of detected signals divided by the total number of images. We studied the relationship between SNR and the signal detection rate using Spearman’s rank correlation coefficient.

The correlation coefficient of SNRi was 0.925, making it the highest among the three SNR types. That of SNRm was 0.904; it correlated at the same level as SNRi although it is not considered human visual characteristics. That of SNRa was 0.446; there is poor correlation between SNRa and detectability because SNRa only depends upon the contrast of the signal and the standard deviation of the noise, and the signal size is not taken into account.

Hayashi, Y, Kimura, Y, Fujita, N, Kodera, Y, Yamada, M, Horii, A, Study of Signal-to-Noise Ratios Considered as Human Visual Characteristics.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11014478.html