Related Experiment Videos
The procrustean bed and the self-anointed raters
1Neuroimaging Section, National Institutes of Health, Bethesda, MD 20892.
Radiology
|February 1, 1989
Summary
Formalized statistical assessments of diagnostic radiology are unreliable due to inconsistent data handling. The inherent nature of image evaluation in radiodiagnostics makes it unsuitable for rigid statistical appraisal, especially for rapidly evolving techniques.
Area of Science:
- Radiology
- Medical Imaging
- Bibliometrics
Background:
- A recent journal rating surprisingly placed Radiology at number 3.
- This rating prompts a critical examination of statistical assessment methods in diagnostic radiology.
Purpose of the Study:
- To critically analyze formalized statistics-based assessments of diagnostic radiologic activities.
- To evaluate the suitability of statistical appraisal for image-based diagnostic processes.
Main Methods:
- Analysis of a published journal rating methodology.
- Critical review of statistical approaches applied to diagnostic radiology.
Main Results:
- The journal ranking was found to result from inconsistent treatment of primary data.
- Diagnostic radiology, based on image evaluation, is argued to be unsuitable for rigid statistical appraisal.
Conclusions:
- Formalized statistical assessments in diagnostic radiology are flawed and unreliable.
- The dynamic nature of radiodiagnostic techniques challenges meaningful statistical evaluation.
- Meaningful and timely information on rapidly changing radiodiagnostic techniques is unlikely to be obtained through rigid statistical methods.