Premium
The determination of the contrast necessary for the detection of small positive or negative contrasting areas in flood field images by employing ROCs
Author(s) -
Geldenhuys E. M.
Publication year - 1991
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.596682
Subject(s) - contrast (vision) , optics , receiver operating characteristic , mathematics , image quality , statistics , field (mathematics) , scintillation , quality (philosophy) , physics , artificial intelligence , computer science , image (mathematics) , detector , pure mathematics , quantum mechanics
Flood field images are acquired and inspected visually during weekly quality control procedures for scintillation cameras. Nonuniformities are quantified by parameters such as integral and differential uniformity (IU and DU). This study is a first step toward understanding the relation between the standard quantitative parameters measured in regular quality control procedures and the performance of clinicians in the interpretation of studies conducted with the same scintillation camera. This study quantified the performance of observers to detect nonuniformities in an ideal case. Flood field images (64×64) were simulated on a computer. One contrasting area (2×2) was superimposed at random positions on each image. Positive and negative contrast values of 10%, 8%, 6%, 4%, 2%, and 0% were employed. The linearly scaled computer images were transferred to film. These films were evaluated by 11 observers to obtain receiver operating characteristic (ROC) curves. The areas ( A ) below the average ROCs for every contrast value were utilized as an indication of the detectability of the contrasting areas superimposed on the flood images. The results indicate areas with positive contrasts >6% and negative contrasts <−8% were detected with 95% probability ( A ≥0.95). These contrasts correspond to optical density differences of >0.051 and <−0.072, respectively (assuming a gamma, G , or gradient of 2.0 for the x‐ray film characteristic curve). The contrast values and optical density differences obtained may serve as guide values, beyond which action must be taken to correct the scintillation camera nonuniformity to ensure optimum imaging. These results can be utilized by other institutions to predict the threshold contrasts of their imaging systems if G can be measured or estimated for the systems.