Organoid Array

October 10, 2026 Textile EMG grids and the front-end lesson for MEAs

A graphics paper records 64 channels of facial surface EMG on two textile grids and decodes 383 expression blendshapes at 100 Hz from muscle activity alone. The neural network is unremarkable; the instrumentation lessons for anyone building dense microelectrode arrays are not.

October 8, 2026 Spiking neuron models and the real-time loop twin

A survey from Innopolis University classifies single-compartment spiking neuron models from integrate-and-fire to Hodgkin-Huxley, with a comparison table of variables, parameters, and FLOPs. Its real value for array work is an explicit compute budget for the model that must run faster than the tissue in any closed-loop setup.

October 8, 2026 Frontier decomposition and the MEA channel budget

A hybrid quantum-classical paper solves dense subset-selection by optimizing only a small rotating frontier of variables at a time. The mechanism, freezing the rest into reduced coefficients and crawling across stages, is a directly readable template for channel scheduling on high-density microelectrode arrays.

October 7, 2026 A movable crystal as a trainable analog kernel

Researchers built an all-optical classifier where a single nonlinear crystal performs the feature expansion of a support vector machine kernel, retuned by translating the crystal along the beam axis. The lesson for microelectrode array instrumentation is not optical; it is about what the acquisition chain already computes in physics, and how cheaply it can be calibrated.

October 6, 2026 Inverse design is coming for the analog front end

A perspectives article from the University of Vienna surveys inverse design in magnonics, where algorithms now find device structures that manual design cannot, including optimization run directly on physical hardware with no simulation. The same methodology reframes how MEA front ends and the electrode-tissue interface could be designed, calibrated, and specified.

October 6, 2026 A quantum compiler wall shows where array scaling breaks

ARGON, a compilation framework for neutral-atom quantum processors, breaks a scheduling bottleneck by precomputing hardware-certified spatial layouts offline and guiding routing with a graph neural network. The same spatiotemporal decoupling maps directly onto the configuration-management problem that large-channel-count MEAs have been quietly accumulating.