Treatment Potential for Macular Cone Vision in Leber Congenital Amaurosis Due toCEP290orNPHP5Mutations: Predictions From Artificial Intelligence

Abstract

PURPOSE. To use supervised machine learning to predict visual function from retinal structure in retinitis pigmentosa (RP) and apply these estimates to CEP290-and NPHP5-associated Leber congenital amaurosis (LCA) to determine the potential for functional improvement. METHODS. Patients with RP (n ¼ 20) and LCA due to CEP290 (n ¼ 12) or NPHP5 (n ¼ 6) mutations were studied. A patient with CEP290 mutations but mild retinal degeneration was included. RP patients had cone-mediated macular function. A machine learning technique was used to associate perimetric sensitivities to local structure in RP patients. Models trained on RP data were applied to predict visual function in LCA. RESULTS. The RP and LCA patients had comparable retinal structure. RP patients had peak sensitivity at the fovea surrounded by decreasing sensitivity. Machine learning could successfully predict perimetry results from segmented or unsegmented optical coherence tomography (OCT) input. Application of machine learning predictions to LCA within the residual macular island of photoreceptor structure showed differences between predicted and measured sensitivities defining treatment potential. In patients with retained vision, the treatment potential was 4.6 6 2.9 dB at the fovea but 16.4 6 4.4 dB at the parafovea. In patients with limited or no vision, the treatment potential was 17.6 6 9.4 dB. CONCLUSIONS. Cone vision improvement potential in LCA due to CEP290 or NPHP5 mutations is predictable from retinal structure using a machine learning approach. This should allow individual prediction of the maximal efficacy in clinical trials and guide decisions about dosing. Similar strategies can be used in other retinal degenerations to estimate the extent and location of treatment potential.

Foveal Therapy in Blue Cone Monochromacy: Predictions of Visual Potential From Artificial Intelligence

“…The supervised machine learning approach used data from patients with retinitis pigmentosa with only cone-mediated macular function remaining. The results in these two forms of severe genetic retinal blindness allowed for prediction of maximal efficacy ( Sumaroka et al, 2019 ).…”

“…A pattern has been that if VA improves (to some degree) compared to baseline, success is announced, and the trial may actually move forward to later phases, given regulatory approval of course. For BCM [and CEP290 -LCA and IQCB ( NPHP5 )-LCA; Sumaroka et al, 2019 ], we can now advance to quantitative prediction of efficacy outcome. Once a patient has a clinical and molecular diagnosis of BCM, relevant parameters of function and structure would be quantified.…”

“…All IRD patients had retained foveal outer nuclear layer thickness (range, 24–150 μm). We assumed that the cones in this cohort of IRD patients were functioning proportional to their remaining quantum catch and there was no additional de-sensitization beyond the partial loss of photoreceptors and shortening of outer segments (OSs) among the surviving cones ( Sumaroka et al, 2019 ). Patients with cystoid macular edema or foveal atrophy were not included.…”

“…The first technique, a supervised machine learning approach [random forest (RF)], was used to model the relationship between foveal function [foveal sensitivity (FS); VA] and retinal structure. Following our previous data analyses ( Sumaroka et al, 2019 ), two groups of models based on the form of input variables were constructed. The model in each group was applied to predict FS and VA. Group 1 (Model I-FS and I-VA) used the input features derived from segmentation parameters: thicknesses of ONL, IS, COS, and RPE, and the number of distinct layers between ELM and RPE2/BrM.…”

“…Machine learning techniques have recently been used to predict treatment potential in two forms of Leber congenital amaurosis (LCA) caused by mutations in CEP290 or IQCB1 ( NPHP5 ), that had little or no measurable vision but some measurable central retinal structure ( Sumaroka et al, 2019 ). The present work uses these methods to try to predict from cross-sectional retinal structure images with OCT in BCM patients of different genotypes the best possible foveal visual outcomes in a clinical trial of intravitreally delivered vector-gene, acknowledging that it would only target foveal cones and not extracentral dysfunctional cones in these retinas.…”