Non-invasive neural decoding at millisecond resolution via ultrasound-gated molecular transducers

High-bandwidth brain-machine interfaces (BMIs) are currently limited by a fundamental 8 trade-off between invasiveness and signal fidelity. While invasive electrodes provide single9 neuron resolution, they carry significant surgical risks; conversely, non-invasive modalities 10 lack the spatiotemporal precision required for complex motor control. Here we report MAN11 TIS (Molecular-Acoustic Neural Transduction and Interfacing System), a platform that 12 transforms the brain parenchyma into an active nonlinear acoustic medium. By utilizing 13 biofunctionalized nanotransducers that modulate tissue nonlinearity (βNL) in response to 14 membrane voltage, we demonstrate the ability to read neural activity using focused ul15 trasound. We present a deep learning decoder, the Temporal Transformer with Harmonic 16 Attention (TTHA), which reconstructs firing rates from acoustic backscatter with < 20 ms 17 latency. Our computational validation suggests that MANTIS achieves sub-millimeter reso18 lution while adhering to FDA Class III safety limits, offering a scalable path to surgically-free, 19 high-performance neural interfacing.