Enhancing central visual field loss representation with a hybrid unsupervised approach.

Purpose: To effectively represent central visual field (VF) loss for individual patients using a hybrid unsupervised approach.

Methods: We obtained 7927 10-2 VF test data from 3328 patients in 5 hospitals. We propose a hybrid approach that combines archetypal analysis (AA) and fuzzy c-means (FCM) to identify characteristic patterns and decompose 10-2 VF without loss. To compare the performance between hybrid approach using FCM and AA single approach, mean deviation (MD) change prediction was performed through supervised learning using decomposition coefficient changes and a linear mixed-effects model was built to investigate the relationship between the MD slope and baseline decomposition coefficients.

Results: We identified 10 representative archetypes for 10-2 VF test. The hybrid approach using FCM outperformed the AA single approach in predicting MD change, achieving lower mean squared error and higher pearson correlation coefficient (all P ≤ 0.039). According to the linear mixed-effects model, the hybrid approach using FCM provides a better fit for predicting MD slope compared to the AA single approach, as reflected by lower akaike information criterion (AIC) and bayesian information criterion (BIC) scores (AIC decrease: 20.31, BIC decrease: 13.33). Eyes with baseline VFs with more inferior and both hemifield loss and less intact field and nearly total loss were associated with faster central VF progression (all P ≤ 0.026).

Conclusion: A hybrid approach combining AA and FCM to analyze 10-2 VF can visualize central VF tests in characteristic patterns and enhance prediction of central VF progression with minimized projection loss decomposition compared with AA single approach.