SVM-Based Optical Detection of Retinal Ganglion Cell Apoptosis

Abstract

Background: Retinal ganglion cell (RGC) loss is crucial in eye diseases like glaucoma. Axon damage and dendritic degeneration precede cell death, detectable within optical coherence tomography (OCT) resolution, indicating their correlation with neuronal degeneration. The purpose of this study is to evaluate the optical changes of early retinal degeneration. Methods: The detection of optical changes in the axotomised retinal explants was completed in six C57BL/6J mice. OCT images were acquired up to 120 min from enucleation. A grey-level co-occurrence-based texture analysis was performed on the inner plexiform layer (IPL) to monitor changes in the optical speckles using a principal component analysis (PCA) and a support vector machine (SVM). In parallel tests, retinal transparency was confirmed by a comparison of the modulation transfer functions (MTFs) at 0 and 120 min. Results: Quantitative confirmation by analysis of the MTFs confirmed the non-degradation of optical transparency during the imaging period: MTF (fx) = 0.267 ± 0.02. Textural features in the IPL could discriminate between the optical signals of RGC degeneration. The mean accuracy of the SVM classification was 86.3%; discrimination was not enhanced by the combination of the SVM and PCA (81.9%). Conclusions: Optical changes in the IPL can be detected using OCT following RGC axotomy. High-resolution OCT can provide an index of retinal neuronal integrity and its degeneration.

Keywords

• modulation transfer function
• optical coherence tomography
• retinal ganglion cell
• support vector machine
• principal component analysis