The Showdown against PD0325901 And Ways To Suceed in It

Provided a representative and adequately large training data set in relation to the input parameters, this approach produces robust classifiers that are able to generalize well, even in the absence of a model of the underlying process, and are insensitive to noise and outliers in the data. Using combinations of RNFL parameters provided by the Stratus OCT (Grewal et?al. 2008), Heidelberg retinal tomography (Naithani et?al. 2007) or GDx (Grewal et?al. 2008; Zhu et?al. 2010), some authors have performed neural network analysis to evaluate detection of glaucoma or to detect visual field deficit progression. These methods produced an AROC larger than 0.95 and were more effective for discriminating between normal and pathologic eyes than the commercially available RNFL parameters (Burgansky-Eliash et?al. 2005; Naithani et?al. 2007). Fourier-domain OCT, however, has not been evaluated using ANN. While ANN methods have been established for distinguishing between control subjects and patients with multiple sclerosis using the P300 cognitive evoked potential (Wu et?al. 1994), a neural network model has not been used to assess the diagnostic ability of OCT in MS. In this study, we analysed whether a selective combination of Spectralis RNFL parameters could further optimize RNFL decrease detection in patients with multiple sclerosis. The design of the study followed the tenets of the Declaration of Helsinki. The study protocol was approved by the Clinical Research Ethics Committee of Aragon (Zaragoza, Spain), and informed written consent was obtained from all participants. Required inclusion criteria were as follows: best-corrected visual acuity (BCVA) of 20/40 or better, refractive error within ��5.00 dioptres equivalent sphere and ��2.00 dioptres astigmatism, transparent ocular media (nuclear colour/opalescence, cortical or posterior subcapsular lens opacity