Premium
Extended random‐dot models for radiographic image noise analysis [In Japanese]
Author(s) -
Yamada Isao
Publication year - 1997
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 - 10.1118/1.598047
Subject(s) - noise (video) , autocorrelation , shot noise , radiography , granularity , transmittance , physics , computer science , optics , artificial intelligence , image (mathematics) , statistical physics , algorithm , mathematics , statistics , detector , nuclear physics , operating system
The object of this thesis was to study the basic characteristics of radiographic image noise using a microscopic theoretical model. Radiographic images obtained by screen‐film systems contain various noise sources: quantum mottle, structure mottle and film granularity. These noise sources affect the signal detectability in the radiographic image. A two‐dimensional microstructure model for radiographic image noise is proposed, i.e., an extended random‐dot model which obeys triple Poisson process. Rigorous expressions of mean transmittance and autocorrelation function were obtained. By using numerical calculations, it is shown that this model can explain basic characteristics of radiographic image noise. Next, a random‐dot model containing a signal is used to evaluate the theoretical detectability of computer‐generated random‐dot images, i.e., idealized models of nuclear‐medicine images. Theoretical detectability of a random‐dot model containing a signal are calculated and compared with the results of receiver operating characteristics analysis (ROC). The results of ROC analysis were in agreement with the calculated theoretical detectability.