RESUMO
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.
RESUMO
BACKGROUND: Optic nerve head measurements extracted from optical coherence tomography (OCT) show promise for monitoring clinical conditions with elevated optic nerve heads. The aim of this study is to compare reliability within and between raters and between image acquisition devices of optic nerve measurements derived from OCT scans in eyes with varying degrees of optic nerve elevation. METHODS: Wide angle line scans and narrow angle radial scans through optic nerve heads were obtained using three spectral domain(SD) OCT devices on 5 subjects (6 swollen optic nerves, 4 normal optic nerves). Three raters independently semi-manually segmented the internal limiting membrane(ILM) and Bruch's membrane(BM) on each scan using customized software. One rater segmented each scan twice. Segmentations were qualitatively and quantitatively compared. Inter-rater, intra-rater and inter-device reliability was assessed for the optic nerve cross sectional area calculated from the ILM and BM segmentations using intraclass correlation coefficients and graphical comparison. RESULTS: Line scans from all devices were qualitatively similar. Radial scans for which frame rate could not be adjusted were of lower quality. Intra-rater reliability for segmentation and optic nerve cross sectional area was better than inter-rater reliability, which was better than inter-device reliability, though all ICC exceeded 0.95. Reliability was not impacted by the degree of optic nerve elevation. CONCLUSIONS: SD-OCT devices acquired similar quality scans of the optic nerve head, with choice of scan protocol affecting the quality. For image derived markers, variability between devices was greater than that attributable to inter and intra-rater differences.
RESUMO
Background/Aims: High intracranial pressure (ICP) is associated with changes in peripapillary Bruch's membrane (pBM) shape on optical coherence tomography (OCT) images of the optic nerve head. It is not known if image acquisition pattern and analysis method impact this association. Materials and Methods: Cross sectional OCT scans of the optic nerve head were obtained at six angles using a radial scan pattern in 21 subjects immediately prior to ICP measurement via lumbar puncture. On each image, Bruch's membrane was manually segmented and defined by either 14 or 16 semi-landmarks and either rater identified, or distance identified boundaries. For each of these four image analysis strategies, geometric morphometric analysis identified the first principal component of Bruch's membrane shape for all images and for the set of images taken at each angle. Repeated measures ANOVA of the first principal component magnitude (PC1) for all images assessed for shape difference between image angles. Linear generalized estimating equation models assessed association between angle specific first principal component magnitudes (PC1) and ICP for each angle. Receiver operating characteristic analysis assessed angle specific PC1s' ability to differentiate elevated from normal ICP. Results: The first principal component represented deflection into the vitreous for all scan angles, but quantitatively differed across scan angles (p < 0.005, repeated measures ANOVA). Angle specific first principal components were positively correlated with ICP (p < 0.005 for all angles, generalized estimating equation models). All angle specific first principal components showed excellent ability to classify ICP (area under curve ≥ 0.8 for all). These results were independent from image analysis strategy. Discussion: Though qualitative changes in Bruch's membrane shape are similar regardless of cross-sectional angle of the 2-D OCT scan, they differ quantitatively between OCT scan angles, meaning that pBM is not axially symmetric and therefore PC1 extracted from different 2-D scan angles can't be compared between individuals. However, we do not identify an optimal scan angle for classification of ICP since there is a similarly strong linear relationship between the first principal component of shape and ICP and angle specific first principal components of Bruch's membrane shape showed similarly excellent ability to differentiate elevated from normal ICP. The results support development of Bruch's membrane shape extracted from 2-D cross sectional optic nerve head OCT scans as a biomarker of ICP and emphasize the importance of consistency of scan angle. This is relevant for developing diagnostic protocols that use OCT to detect high ICP states.