Comparison of clinicians and an artificial neural network regarding accuracy and certainty in performance of visual field assessment for the diagnosis of glaucoma.

摘要: Purpose: To compare clinicians and a trained artificial neural network (ANN) regarding accuracy certainty of assessment visual fields for the diagnosis glaucoma. Methods: Thirty physicians with different levels knowledge experience in glaucoma management assessed 30-2 SITA Standard field printouts that included full Statpac information from 99 patients glaucomatous optic neuropathy 66 healthy subjects. Glaucomatous eyes perimetric mean deviation values worsethan -10 dB were not eligible. The graded on scale 1-10, where 1 indicated absolute certaintyand 10 signified glaucoma; 5.5 was cut-off between same classified by previously ANN. ANN output transformed into linear matched used subjective assessments. Classification using classification error score. Results: Among physicians, sensitivity ranged 61% to 96% (mean 83%) specificity 59% 100% 90%). Our achieved 93% 91% specificity, it significantly more sensitive than (p < 0.001) at similar level specificity. score equivalent top third scores all never high degree any its misclassified tests. Conclusion: Our results indicate performs least as well assessments