Factors affecting recognition of scintigraphic abnormalities
| Autor: | Lawrence C. Kohlenstein, Lloyd G. Knowles, Alvin G. Schulz |
|---|---|
| Rok vydání: | 1973 |
| Předmět: |
Computers
business.industry Pattern recognition Observer (special relativity) Models Theoretical Statistical power Lesion Visual recognition Evaluation Studies as Topic Nuclear medicine imaging Count density medicine Radiology Nuclear Medicine and imaging Lack of knowledge Artificial intelligence False positive rate Diagnostic Errors medicine.symptom Radionuclide Imaging business Nuclear medicine |
| Zdroj: | Seminars in Nuclear Medicine. 3:327-341 |
| ISSN: | 0001-2998 |
| DOI: | 10.1016/s0001-2998(73)80026-1 |
| Popis: | Predicting the recognition of marginally detectable abnormalities in scintigrams by the physician is a key weakness in our present capability for evaluating the effectiveness of nuclear medicine imaging systems. Development of a method for evaluating scintigraphic systems and for predicting lesion-detection performance has been impeded by lack of knowledge of the visual response of the observer to the poor resolution, low contrast, and random intensity fluctuations that are characteristic qualities of scintigrams. Three parameters of the image appear to govern visual recognition of abnormalities: count density in the general area of the lesion, contrast between the suspected area of the image and adjacent normal areas, and the lateral dimension of the perturbation in count density caused by the lesion. The effects of various characteristics of the clinical problem, the radiopharmaceutical, and the imaging system on these image parameters are discussed. Available data on the role of each image parameter in detection of localized lesions are presented. Tradeoffs involving features of the clinical problem and image system that affect more than one of these image parameters simultaneously are discussed. Experimental measurements of the performance of observers in detecting simulated localized lesions are reviewed, and a model is developed from these data that predicts the values of lesion contrast, count density, and image lesion size required for a 50% probability of detection. These data are compared to the detection achieved by a statistical mathematical observer constrained to the same average false positive rate to determine how close an observer comes to taking full advantage of the statistical information in the image. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|