The possibility to control actions from thoughts is becoming reality thanks to Brain Machine Interfaces (BMIs). However modern BMIs are mostly unidirectional and aimed at restoring lost motor functions. In a wider perspective, neural interfaces must be bi-directional devices that substitute motor, sensory or cognitive circuits within the brain, that might have been damaged as a result of an injury or a disease. The project goes towards this direction: it will provide the knowledge to realize a new class of neuroprostheses aimed at treating those diseases where a portion of brain tissue is damaged (e.g. lesion). The ultimate goal of the project is to connect in vitro neuronal assemblies to an artificial system (a neuromorphic chip) which aims at restoring the lost neuronal functionality, with the long-term perspective to be implanted in humans affected by invalidating brain diseases. This will allow us to reach three main scientific objectives: (i-DECODING) improve the performance of BMIs by investigating optimal decoding schemes to extract useful information by multisite acquisitions from neuronal assemblies; (ii-PROCESSING) provide an artificial link between two previously connected assemblies by designing computational models (sw and hw) which mimic the behavior of the injured part of the network; (iii-CODING) design new strategies by developing stimulation protocols aimed at effectively sending appropriate information to cell assemblies. These goals will be reached by using an engineering approach to neuroscience: a fully controlled and measurable experimental system where the 'biological' neural element will be investigated at different levels of anatomical complexity: from random, unstructured circuits up to the intact brain. Thanks to this approach we will be able to exploit the peculiarity of each biological model to study different coding/decoding schemes in order to optimize the performance of advanced neuroprostheses.