Evaluating human and automated interpretation of visual field data in perimetry.

The complexity of visual field result evaluation has increased with the advent of automated perimetry. This in turn makes software assistance such as provided by the new Octosmart program interesting and desirable. Its accuracy and reliability in visual field evaluation must be tested, though, and the testing methodology itself must be developed. For this purpose, 27 visual fields were evaluated by three human interpreters and the Octosmart program. Based on a newly developed compound index, the performance of the three interpreters and Octosmart is described and a high degree of consistency is demonstrated.