Comparison of Machine Learning Approaches to Improve Diagnosis of Optic Neuropathy Using Photopic Negative Response Measured Using a Handheld Device

The photopic negative response of the full-field electroretinogram (ERG) is reduced in optic neuropathies. However, technical requirements for measurement and poor classification performance have limited widespread clinical application. Recent advances in hardware facilitate efficient clinic-based recording of the full-field ERG. Time series classification, a machine learning approach, may improve classification by using the entire ERG waveform as the input. In this study, full-field ERGs were recorded in 217 eyes (109 optic neuropathy and 108 controls) of 155 subjects. User-defined ERG features including photopic negative response were reduced in optic neuropathy eyes (p < 0.0005, generalized estimating equation models accounting for age). However, classification of optic neuropathy based on user-defined features was only fair with receiver operating characteristic area under the curve ranging between 0.62 and 0.68 and F1 score at the optimal cutoff ranging between 0.30 and 0.33. In comparison, machine learning classifiers using a variety of time series analysis approaches had F1 scores of 0.58–0.76 on a test data set. Time series classifications are promising for improving optic neuropathy diagnosis using ERG waveforms. Larger sample sizes will be important to refine the models.

Variability of Retinal Vessel Tortuosity Measurements Using a Semiautomated Method Applied to Fundus Images in Subjects With Papilledema

Purpose

To develop methods to quantitatively measure retinal vessel tortuosity from fundus images acquired in subjects with papilledema and assess sources of variability in these measurements.

Methods

Digital fundus images from 30 eyes of subjects with untreated idiopathic intracranial hypertension and papilledema were analyzed. Retinal vein and artery tortuosities for three to four vessels of each type were measured in a region of interest 1.8 to 2.7 mm from the center of the optic nerve head. Measurements were averaged to generate a venous tortuosity index (VTI) and arterial tortuosity index (ATI) for each eye. One image of each eye was analyzed two times by the same rater. Two images of each eye, differing by focal depth, were analyzed by the same rater. Correlations between VTI and ATI for the same image and different images were calculated.

Results

Intrarater Pearson correlations (r) were 0.8 (95% confidence interval [CI], 0.59–0.9) and 0.90 (95% CI, 0.73–0.96) for VTI and ATI, respectively, with one outlier removed. Interimage r values were 0.72 (95% CI, 0.48–0.87) and 0.96 (95% CI, 0.89–0.99) for VTI and ATI, respectively, with one outlier removed. The intraclass correlation coefficients for agreement and consistency were similar, suggesting that the discrepancy between measurements was due to residual random error.

Conclusions

The finding of similar intrarater and interimage variability suggests that intrarater variability may be a more dominant source than physiology and image acquisition.

Translational Relevance

Standardizing rater procedures and averaging multiple measuring sessions are strategies to reduce variability and improve reliability of detecting retinal vessel tortuosity changes in images of eyes with papilledema.