This project will study and model observers' detection-and-localization of target objects sought in backgrounds of image noise. The model assumes that: 1) the observer's detection response and first-choice of target location both depend on the """"""""maximally suspicious"""""""" finding on the image, 2) the evaluation of the normal finding regarded as the """"""""most suspicious"""""""" does not depend on whether or not a target is present and 3) a correct localization of the actual target occurs if and only if its localization was regarded as the most suspicious. Formalizations of these assumptions lead to very strong predications about the relation between detection and localization accuracies, and between the ROC curve and the """"""""Localization-Response"""""""" (LROC) curve that plots (against the false-positive rate) the joint probability that a true-positive response will also result in a correct localization of the actual target. The parametric version of the model can be used to develop maximum-likelihood procedures for fitting either the ROC curve alone or both the ROC and LROC curves. The same model can be extended to multiple-report interpretations of images that may contain multiple target objects, provided that both the observer's detection capability and strategy for reporting possible targets remain invariant across all images and positive reports. This extension of the model leads to formulations for so-called """"""""Free-Response"""""""" (FROC) curves and a recently proposed """"""""Alternative FROC"""""""" (AFROC) curves. The experiments will apply this model, and procedures developed to fit observer's ROC, LROC, FROC and AFROC data, in tasks that require both single-report and multiple-report interpretations. The target-objects to be found at unknown image locations will be either: a)low visibility objects randomly located in uniform-noise backgrounds or b) easily-visible objects that need to be discriminated (by contrast, size or shape) from other visible """"""""nontarget"""""""" objects in the image. Physical calculations on the images (from a realized cross-correlator) will be used to study how closely the systematic changes in observer performance can be predicted from the image information.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Research Project (R01)
Project #
Application #
Study Section
Biomedical Library and Informatics Review Committee (BLR)
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Brigham and Women's Hospital
United States
Zip Code
Swensson, R G (2000) Using localization data from image interpretations to improve estimates of performance accuracy. Med Decis Making 20:170-85
Wester, C; Judy, P F; Polger, M et al. (1997) Influence of visual distractors on detectability of liver nodules on contrast-enhanced spiral computed tomography scans. Acad Radiol 4:335-42
Swensson, R G (1996) Unified measurement of observer performance in detecting and localizing target objects on images. Med Phys 23:1709-25
Swensson, R G; Judy, P F (1996) Measuring performance efficiency and consistency in visual discriminations with noisy images. J Exp Psychol Hum Percept Perform 22:1393-415
Seltzer, S E; Judy, P F; Adams, D F et al. (1995) Spiral CT of the chest: comparison of cine and film-based viewing. Radiology 197:73-8
Seltzer, S E; Judy, P F; Swensson, R G et al. (1994) Flattening of the contrast-detail curve for large lesions on liver CT images. Med Phys 21:1547-55
Judy, P F; Swensson, R G; Nawfel, R D et al. (1992) Contrast-detail curves for liver CT. Med Phys 19:1167-74
Nawfel, R D; Chan, K H; Wagenaar, D J et al. (1992) Evaluation of video gray-scale display. Med Phys 19:561-7
Seltzer, S E; Swensson, R G; Nawfel, R D et al. (1991) Visualization and detection-localization on computed tomographic images. Invest Radiol 26:285-94
Swensson, R G; Theodore, G H (1990) Search and nonsearch protocols for radiographic consultation. Radiology 177:851-6

Showing the most recent 10 out of 12 publications