Validity of the Gaussian assumption in the analysis of ROC data obtained from scintigraphic-like images.

The Gaussian assumption (GA) has been found experimentally to provide an acceptable fit to ROC data elicited from visual experiments. When the GA is satisfied the two-parameter index D(delta m, s) determines the ROC completely. It is however not known whether the GA is an intrinsic property of the eye-brain system, or whether it is a consequence of the type of visual stimuli selected for the experiments. Using scintigraphic-like images to test the validity of the GA, it is found that (i) series of visual stimuli with as few variables as possible produce ROC data compatible with the GA, but (ii) it is possible to find stimuli for which the GA provides a very poor account of experimental data. These results lead us to doubt that the GA is an intrinsic property of the human visual system.

[1]  D. Turner,et al.  An intuitive approach to receiver operating characteristic curve analysis. , 1978, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[2]  A. Houston,et al.  An intercomparison of computer assisted image processing and display methods in liver scintigraphy. , 1979, Physics in medicine and biology.

[3]  A S Houston,et al.  An intercomparison of computer assisted data processing and display methods in radioisotope scintigraphy using mathematical tumours. , 1977, Physics in medicine and biology.

[4]  J Mallard,et al.  The measurement of the performance of the display system of radioisotope imaging device: the multi-element band display. , 1976, The British journal of radiology.

[5]  C. Metz,et al.  Statistical significance tests for binormal ROC curves , 1980 .

[6]  A S Houston.  A comparison of four standard scintigraphic TV displays: concise communication. , 1980, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[7]  D. Dorfman,et al.  Maximum-likelihood estimation of parameters of signal-detection theory and determination of confidence intervals—Rating-method data , 1969 .

[8]  P. Guignard,et al.  A comparative method based on ROC analysis for the quantitation of observer performance in scintigraphy. , 1982, Physics in medicine and biology.

[9]  A. Cuarón,et al.  Interobserver variability in the interpretation of myocardial images with Tc-99m-labeled diphosphonate and pyrophosphate. , 1980, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[10]  W. I. Kayes.  An analysis of multi-level scan display systems , 1973 .

[11]  L B Lusted,et al.  Radiographic applications of receiver operating characteristic (ROC) curves. , 1974, Radiology.

[12]  J. H. Gallagher,et al.  ROC analysis of diagnostic performance in liver scintigraphy. , 1981, Journal of nuclear medicine : official publication, Society of Nuclear Medicine.

[13]  J. Swets.  ROC analysis applied to the evaluation of medical imaging techniques. , 1979, Investigative radiology.

[14]  Byron J. T. Morgan,et al.  Some aspects of ROC curve-fitting: Normal and logistic models , 1972 .

[15]  C. Metz.  Basic principles of ROC analysis. , 1978, Seminars in nuclear medicine.