Computer-aided assessment of diagnostic images for epidemiological research

Diagnostic images are often assessed for clinical outcomes using subjective methods, which arelimited by the skill of the reviewer. Computer-aided diagnosis (CAD) algorithms that assist reviewers in theirdecisions concerning outcomes have been developed to increase sensitivity and specificity in the clinical setting.However, these systems have not been well utilized in research settings to improve the measurement of clinicalendpoints.

Reductions in bias through their use could have important implications for etiologic research.

Methods: Using the example of cortical cataract detection, we developed an algorithm for assisting a reviewer inevaluating digital images for the presence and severity of lesions. Available image processing and statisticalmethods that were easily implementable were used as the basis for the CAD algorithm.

The performance of thesystem was compared to the subjective assessment of five reviewers using 60 simulated images. Cortical cataractseverity scores from 0 to 16 were assigned to the images by the reviewers and the CAD system, with each imageassessed twice to obtain a measure of variability.

Image characteristics that affected reviewer bias were alsoassessed by systematically varying the appearance of the simulated images.

Results: The algorithm yielded severity scores with smaller bias on images where cataract severity was mild tomoderate (approximately less than or equal to 6/16ths). On high severity images, the bias of the CAD system exceeded that of thereviewers.

The variability of the CAD system was zero on repeated images but ranged from 0.48 to 1.22 for thereviewers. The direction and magnitude of the bias exhibited by the reviewers was a function of the number ofcataract opacities, the shape and the contrast of the lesions in the simulated images.

Conclusions: CAD systems are feasible to implement with available software and can be valuable when medicalimages contain exposure or outcome information for etiologic research.

Our results indicate that such systemshave the potential to decrease bias and discriminate very small changes in disease severity. Simulated images area tool that can be used to assess performance of a CAD system when a gold standard is not available.

Author: Alison AbrahamDonald DuncanStephen GangeSheila West
Credits/Source: BMC Medical Research Methodology 2009, 9:74

Published on: 2009-11-11

Copyright by the authors listed above - made available via BioMedCentral (Open Access). Please make sure to read our disclaimer prior to contacting 7thSpace Interactive. To contact our editors, visit our online helpdesk. If you wish submit your own press release, click here.

Comments Page 0 of 0

There are currently 0 comments to display.

+ Add New Comment

Custom Search