BCIwiki (bciwiki.com) — According to Istanbul Medipol University, researchers at the university introduced NeuroGate, a framework for predicting an unrecorded brain region's waveform from a recorded region. The preprint was uploaded to bioRxiv on 2026-09-15 and has not been peer reviewed. Its authors are Zareh, A.; Ozdemir, M. K.; and Yagci, R.
The framework was tested across 56 pathways spanning rodent local field potentials (LFP), human stereo-EEG (sEEG) and electrocorticography (ECoG), and scalp EEG. NeuroGate predicted waveforms more accurately than 13 deep-learning and linear baselines. The authors also report low across-session variance and high median accuracy. These outputs are model-predicted waveforms, not additional direct recordings from the target region.
To test whether the predictions reflected circuit directionality, the researchers silenced piriform-cortex output using tetanus toxin light chain. Piriform-to-bulb predictions collapsed in all 8 folds, while predictions in the reverse direction were preserved. The authors interpret this directional difference as consistent with monosynaptic anatomy.
The study demonstrates two applications: quantifying how recording redundancy limits the information contributed by each additional source, and developing a reusable circuit-perturbation protocol for testing directional signal translation. The latter is proposed as a protocol for use in other circuits, rather than a result already validated across all circuits.