The purpose of this work was to develop a quantitative analysis that allows, in healthful and patient with visual loss, to separate registers with signal correlated with the visual stimulation, of the registers that contain noise. System VERIS Science v6.0.5d was used to extract kernel first-class of the registers, and posterior exportation of the data for analysis in the MATLAB. In this environment of programming was carried the analysis of the main components (PCA), that is used to reduce the dimensionalidad of the data conserving its variation and minimizing the influence of the noise; the signal-to-noise analysis (SNR) of the multifocal responses (mfERG) using cumulative distribution SNR carried through in two intervals of time, one understanding only the signal and another one only the noise, for classification of the valid, diminished or absent responses. We have found that in a group healthy subjects, the noise and signal SNR cumulative distributions have not overlapped and occupy distinct ranges of SNR values in all subjects. By using a 1% criterion for false positives and false negatives, it was possible to define the borders of noise, low response, and signal SNR ranges. We have also applied this protocol to a three subject with visual loss due to ocular toxoplasmosis scares. In this case, there was a degree of overlapping between the noise and signal SNR distributions. It was still possible to separate the noise, low response, and signal SNR range by using the confidence limits for the distance between noise and signal boundaries obtained from the group of healthy subjects. The results of SNR evaluation were then used to display the mfERG waveforms in topographical plots that discriminate between reliable responses (signal), low responses, and no responses (noise). Using this approach, features of retinal topography, such as the optic nerve head, are more easily discriminate, whilst artifacts such as an erroneous central peak, are more easily rejected. The combinations of PCA to reconstruct mfERG waveforms with SNR evaluation are useful tools to analyze retinal topography both in healthy and impaired conditions.
|
