SpikeZip: a retinomorphic encoding method for low-latency and loss-less data compression and telemetry of large-scale in-vivo neural recordings

Informations projet

SpikeZip

N° de convention de subvention: 101138283

DOI

10.3030/101138283

Projet clôturé

Date de signature de la CE 19 Octobre 2023

Date de début 1 Novembre 2023

Date de fin 30 Avril 2025

Financé au titre de

• European Research Council (ERC)

Coût total

Aucune donnée

Contribution de l’UE

€ 150 000,00

Coordonné par

STICHTING IMEC NEDERLAND
Netherlands

Periodic Reporting for period 1 - SpikeZip (SpikeZip: a retinomorphic encoding method for low-latency and loss-less data compression and telemetry of large-scale in-vivo neural recordings)

Période du rapport: 2023-11-01 au 2025-04-30

Résumé du contexte et des objectifs généraux du projet

SpikeZIP aims to develop and validate a proof of concept prototype that perform low-loss compression on the neural data. There are three key objectives performed in three work packages
1. PoC hardware and software building:
- To design and fabricate a proof-of-concept hardware (HW) prototype capable of acquiring, compressing, transmission and reception high-channel-count neural signals using the SpikeZip encoding algorithm.
- To develop and implement a real-time software (SW) algorithm for reconstructing and analysing the signal quality of the retinomorphic encoded extracellular neural data.
- To optimize the hardware and software for small form factor, low-power consumption, low-latency data transfer, and compatibility with existing neural recording systems and PC.
2. validation with in-vivo neural recording
- To validate the SpikeZip encoding algorithm using in vivo neural recordings in mice models.
- To analyze the fidelity and compression performance of the SpikeZip-encoded neural data compared to traditional methods.
- To investigate the effect of SpikeZip encoding on waveform features of neural signals for spike sorting.
3. business development
- To develop a comprehensive business plan for the commercialization of the SpikeZip technology. This includes defining the target market, identifying potential customers, and developing a strategy for intellectual property protection. The ultimate goal of this work package is to create a roadmap for bringing the SpikeZip technology to market and maximizing its impact.

Travail effectué depuis le début du projet jusqu’à la fin de la période considérée dans le rapport et principaux résultats atteints jusqu’à présent

PoC hardware and software building (WP1)
The following activities have been performed
- Completion of spikeZip and spikeUnzip hardware design
- Demonstrated targeted compression ratio and reconstruction quality with the HW and SW using both synthetic data and in-vivo recording data.
Key achievements: the PoC prototype HW and SW demonstrates a compression ratio >7x, and a low compression error below 7 micro-volts RMS (well below the typical noise floor of the sensor readout front-end)

Validation with in-vivo neural recording (WP2)
- Record in-vivo data from mice with strong artifact as stress test data, and demonstrate targeted compression and reconstruction error
- Demonstrate the analysis result with widely used spike sorting tool, Kilosort, and identify the influence from the compression.
Key achievements: an in-vivo recording from mice with strong licking artifact is recorded, and is used as input data of the compression. The spike sorting analysis shows that >97% of the spikes from the original synthetic data with ground truth can be recovered with a correlation coefficient >0.95. This shows the compression method proposed in SpikeZip has very low information loss.

Progrès au-delà de l’état des connaissances et impact potentiel prévu (y compris l’impact socio-économique et les conséquences sociétales plus larges du projet jusqu’à présent)

Overview of the results:
- SpikeZIP compression and encoding hardware module
- SpikeZIP decoding hardware (clock and data recovery) and software reconstruction module
- Preclinical in-vivo neural recording dataset from two area of mice brain, based on 2x Neuropixels v2 probe (up to 768 channels)
- Preclinical in-vivo neural recording dataset with licking artifact
- IPR: 3 patents being related to retinomorphic compression being filed

Potential impacts:
- economical impacts:
- societal impacts: improve the precision of the BCI and neuroprosthesis to enhance patients' quality of life (from grab cup to tying shoelace)

Needs for further uptake:
- to achieve the impacts, following funding will be required to further increase TRL (preclinical in-vivo validation) and the development of MVP that fits to the market need