Design of X-ray Images Based on Quantitative Analysis: Predictive Model of Image Quality and Physical Quantity Factors by Multiple Regression Equation of Visual Evaluation (Scheffé’s) and Physical Quantities (Dose and Noise)

Abstract

Purpose: The Scheffé’s method obtains the difference between pair comparisons with that of the interval scale and can judge the superiority or inferiority of the sample to be compared with no restriction in the observation image by the statistical significant difference. However, the Scheffé’s method cannot be judged as a single image quality indicator. Therefore, I examined a method that can evaluate the association of average degree of preference of Scheffé’s method and the physical quantities that make up the image. Methods: This study focuses on the fact that the average degree of preference of the Scheffé’s method is quantitative data on the interval scale and that multiple regression analysis is possible. In the multiple regression analysis, the average degree of preference by imaging simulated pulmonary adenocarcinoma with different exposure doses was used as the objective variable and the exposure doses and noise (standard deviation [SD]) were used as the explanatory variables. The Scheffé’s method used the Nakaya’s modified method. Results: In the multiple regression analysis, SD was P=0.027. By substituting the threshold value of the intersection of the exposure doses and SD into the multiple regression equation (predictive model), the average degree of preference (Ŷ) was calculated. Ŷ_{(Scheffé;Gy,SD)} was −0.147, which was about 1/2 of exposure doses (−0.150). Conclusion: The multiple regression analysis of Scheffé’s (average degree of preference) and physical quantity factors (exposure doses and noise) has made it possible to design images that can reduce exposure doses while maintaining adequate image quality.