Autoregressive Modeling of Visual Evoked Potentials and Its Applications to Optic Nerve Diseases-Ischemic Optic Neuropathy and Optic Neuritis

It is important to differentiate the diagnosis of ischemic optic neuropathy (ION) and optic neuritis (ON) for prognostic and therapeutic reasons. In most cases, differentiation is accomplished by assessing the disc appearance, the presence or absence of retrobulbar pain, the age of the patient, the mode of onset and other features of clinical and laboratory evaluation. However, in certain groups of patients, diagnosis may be difficult because of overlapping clinical profiles in these two disorders. In this paper, an attempt is made to overcome clinically overlapping profiles and to evolve indices to classify and delineate clearly ION and ON groups by differential diagnosis of the visual evoked potentials (VEP) using autoregressive (AR) modeling. In the present work of AR modeling, the data sequence x(n) as the output of a linear system has been carried out using digitized VEP waveform. An appropriate optimal order p for the AR model is chosen based on the Akaike information criterion (AIC). Accordingly, AR model has eight coefficients for each data sequence. These AR model coefficients are computed using Burg’s algorithm. These AR coefficients with different combinations were plotted in the feature plane representations, for distinction between the ION and ON group of patients. It was found that, the feature plane plot of a2 verses a7 has a potential to distinguish clearly the ION and ON patients with respect to normal subjects. This novel technique using the AR feature plane representation is more efficient and thus, enables the neurologist in early therapy planning. KeywordsVisual evoked potential; Ischemic optic neuropathy; Optic neuritis; AR modeling; Feature plane representation; Akaike information criterion __________________________________________________*****_________________________________________________