Nonlinear control of neurodegenerative diseases. A case study on optical illusion networks disrupted by diabetic retinopathy

We present an efficient computational framework for the design of optimal drug delivery control strategies that can successfully treat a family of neurodegenerative diseases that originate from channelopathies and synaptopathies. To this end, we extend our previously introduced scalable and adaptable modelling framework that models heterogeneous Hodgkin-Huxley (HH) neuronal networks to account for the modular organisation of the neurons in the brain, e.g. interconnecting sub-networks of heterogeneous neurons. Based on this framework, we introduce a novel design of lateral inhibition networks to successfully reproduce 2D optical illusions that are known to occur in the human retina. We model the dynamic behaviour of Diabetic Retinopathy (DR), a neurodegenerative disease that progressively hinders the inherent ability of patients to distinguish optical illusions. We implement nonlinear control on accurate models of diabetic lateral inhibition neuronal networks to recover their functionality and investigate the effects of virtual drug administration. We utilise the healthy and diabetic optical illusions generated by these networks as a 'computational' phenotype to design therapies based on an adaptive terminal error iterative learning controller (TE-ILC). Therefore, we provide a comprehensive computational framework that is able to imitate the dynamics of healthy and diseased neuronal networks and we propose an adaptive nonlinear control strategy based on the error between output images that correspond to healthy and diseased conditions.