Receiver operating characteristic analysis and its potential role in evaluating the effect of head movement in PET of the brain

Full text: We outline an example of receiver operating characteristic (ROC) analysis in the assessment of image quality. ROC analysis is a measure of image quality that accounts for the consequences of the decision and the role of the observer. Kim and Haynie (Nuclear Diagnostic Imaging: Practical Clinical Applications. Melbourne: Macmillan, 1987) describe ROC analysis as an 'objective approach to the evaluation of diagnostic decision making'. ROC analysis is an ideal technique for evaluating images of a Hoffman brain phantom obtained using positron emission tomography. Images have been acquired with the phantom in different positions. The position of the phantom and the time the phantom remained in each position was based on the measurements of head movement during simulated brain imaging (Patterson et al., Technologists Symposium, ANZSNM, 1998). This study was undertaken to explore the potential of ROC analysis in determining the effect of movement on the ability to detect lesions of various sizes