Usefulness of the Entropy of the Bayesian Posterior Distribution of the Threshold to Assess the Reliability of Measured Visual Field.

PURPOSE: To investigate the association between reliability of visual field (VF) and the entropy of the Bayesian posterior distribution of the threshold.

MATERIALS AND

METHODS

Subjects were comprised of 122 eyes of 73 patients with primary open-angle glaucoma. The VF was measured using the newly developed variational Bayes linear regression visual field (VBLR-VF) twice within 3 months. The likelihood of this final probability distribution of the thresholds in the measured VF was calculated (Entropy index). The mean absolute error (MAE) of the total deviation (TD) values of the test-retest VFs was measured, and its relationship to fixation losses (FLs), false positives (FPs), false negatives (FNs), and the Entropy index (average of the 2 VFs) was investigated. In addition, the association between the difference in the mean deviation (MD; between 2 VFs) and the difference in the Entropy index (between 2 VFs) was investigated.

RESULTS

The Entropy index was significantly associated with FL, FP, and FN, as well as with the MAE (p < .001). Furthermore, the difference in the Entropy index was significantly negatively associated with the difference in MD (p < .001).

CONCLUSION

The Entropy index was significantly associated with FL, FP, FN, and test-retest reproducibility. In addition, a higher Entropy index was associated with advanced VF loss, potentially indicating lower test reliability.