Purpose:
Therapeutic intervention for inherited retinal degenerations is gene-specific and time sensitive, i.e. depends on preserved function of rod and cone system. The purpose was to estimate topographic distribution of functioning cone and rod system in inherited retinal degenerations with confirmed gene mutations.
Methods:
The dataset was created based on 190 eyes from genotyped subjects with the following phenotype pattern: 60 Usher syndrome, 40 retinitis pigmentosa, 40 morbus Stargardt, 20 occult macula dystrophy and 30 healthy control eyes. ISCEV standard protocols were used for cone and rod electrical response recordings and normalized energy responses were evaluated to assess cone and rod systems functional output. Photoreceptors morphology was evaluated via tracing photoreceptors layers in spectral domain OCT b-scans within 30 degrees diameter around fovea centralis. Approximation of quantitative distribution of functioning cones and rods was performed using learning neural network developed by author. The accuracy of automatic topographic mapping of functioning cone and rod systems was evaluated.
Results:
The normalized energy of the functioning rod system was decreased by 90(±5)%, 50(±5)%, 20(±5)%, 0.1(±0.01)% for Usher syndrome, retinitis pigmentosa, Morbus Stargardt and occult macula dystrophy, respectively, comparing them to healthy controls (p<0.01). The normalized energy of the functioning cone system was decreased by 50(±7)%, 30(±7)%, 85(±9)%, 20(±1)% for Usher syndrome, retinitis pigmentosa, morbus Stargardt and occult macula dystrophy, respectively, comparing them to healthy ones (p<0.01). Cone and rod systems functional output loss was significantly positively associated with photoreceptors morphological changes as r=0.89, r=0.35, r=0.89 and r=0.56 in Usher syndrome, retinitis pigmentosa, Morbus Stargardt and occult macula dystrophy, respectively (p<0.001). Averaged sensitivity and specificity was 0.91 and 0.89 for calculated automatic maps, assessing functioning cones and rods system topographic distribution for specific phenotype-genotype patterns (p<0.01).
Conclusions:
We have proven that the newly developed learning artificial neural network is accurate for estimating the topographic distribution of functioning photoreceptors in inherited retinal degenerations for specific phenotype-genotype patterns. Topographic mapping of cone and rod system's function will assist in the identification of candidates suitable for therapeutic intervention.
Keywords: retinal degenerations: hereditary • imaging/image analysis: clinical • cell survival