The central objective of this multidisciplinary collaborative project is to overcome individual resolution limits of Calcium (poor temporal resolution) and ECoG signals (poor spatial resolution) by designing novel signal processing algorithms that exploit their complementary strengths. 

  On the fabrication front, we developed a new fabrication process for 16-electrode transparent graphene arrays on completely-clear substrates. It exploits an impedance reduction technique to provide uniform impedance distribution across the array. Using a transgenic mouse model, we demonstrated simultaneous electrical recording of cortical activity with high fidelity while imaging calcium signals at various cortical depths right beneath the transparent microelectrodes. We also developed a new fabrication process to build transparent graphene arrays with ultra-small electrodes scaled into single cell dimensions and used it in multimodal imaging experiments with electrophysiological recordings and 2-photon calcium imaging. We demonstrated that our model successfully decodes stimulus-induced multi-unit activity recorded with transparent graphene electrodes from cellular spikes recorded from layer II-III neurons in mouse cortex using calcium imaging.

On the signal processing front, our group developed novel modeling techniques and algorithms for super-resolution spike reconstruction (inference of spike timings at very high rates) from calcium measurements acquired at a lower sampling. The separation between two consecutive neural spikes can be a few milliseconds, whereas the sampling rate of typical 2-photon calcium imaging system is <100 Hz. Existing algorithms can mostly infer the spiking activity at the same resolution as the sampling rate of the signal. In contrast to these methods, we developed a new “multirate” measurement model for calcium data, which allows us to infer spiking activity at rates higher than calcium sampling rate. Using this multirate generative model, it becomes possible to decouple the spike recovery problem into recovering (super-resolving) a block of spikes between any two frames of the calcium signal. By designing simple inequality and equality comparison tests that exploit the inherent binary nature of the spiking activity, we show that it is possible to decode the underlying high-resolution spiking activity. Since not all spiking patterns are equally likely, we can flexibly design the search strategy based on prior information on the spiking patterns. This enables us to significantly reduce the expected number of comparisons needed to decode the spiking pattern compared to performing a naïve search. We conducted extensive benchmarking to evaluate the performance of the proposed technique compared to the existing spike deconvolution techniques used in the neuroscience community for calcium imaging. We also evaluated our method on real neuroscience experimental tasks (conducted by our collaborators lab) where the goal is to predict the decoding time from the deconvolved spikes. Our algorithm is the first of its kind that can decode spikes at higher rates from calcium signals acquired at lower rate, and shows significantly improved performance over existing algorithms. This can also be very useful for wide-field imaging applications where the scanning time nevitably results in a smaller frame rate. Our algorithm can enable the acquisition of wide field data without losing valuable temporal information due to the lower sampling rate.

Following is a partial list of publications produced from our project:

1.  Thunemann, M., Lu, Y., Liu, X., Kılıç, K., Desjardins, M., Vandenberghe, M., ... & Kuzum, D. (2018) “Deep 2-photon imaging and artifact-free optogenetics through transparent graphene microelectrode arrays." Nature communications, 9(1), 2035.

2. Li, W.L., Chu, M.W., Wu, A., Suzuki, Y., Imayoshi, I.#, and Komiyama, T.# (2018) “Adult-born neurons facilitate olfactory bulb pattern separation during task engagement.” eLife, 2018 Mar 13;7. pii: e33006. doi: 10.7554/eLife.33006 

3. P. Sarangi and Piya Pal, "“Superresolution via bilinear fusion of multimodal imaging data.” In Proceedings of SPIE Big Data: Learning, Analytics, and Applications, (Vol. 10989, p. 109890I)", May 2019.

4. Pulak Sarangi, Mehmet Can Hucumenoglu and Piya Pal, "Understanding Sample Complexities for Structured Signal Recovery from Non-linear Measurements", IEEE CAMSAP, 2019 [Best Student Paper Award, first position].

5. Pulak Sarangi, Mehmet Can Hucumenoglu, and Piya Pal, "Effect Of Undersampling on non-negative blind deconvolution with Autoregressive Filters", Proceedings of IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP) 2020.

 6. Pulak Sarangi and Piya Pal, “No Relaxation: Guaranteed Recovery of finite Valued Signals from Undersampled Measurements”, submitted to IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP) 2021.

7. Pulak Sarangi, Ryoma Hattori, Takaki Komiaya, Piya Pal, “Super-resolution Spike Reconstruction from undersampled Calcium Signals using Multirate Autoregressive Generative Models”, (tentative title), Manuscript in preparation. (Our main results are being reported in this manuscript).

8. Ding, D., Lu, Y., Zhao, R., Liu, X., De-Eknamkul, C., Ren, C., Mehrsa, A., Komiyama, T. and Kuzum, D., 2020. Evaluation of Durability of Transparent Graphene Electrodes Fabricated on Different Flexible Substrates for Chronic In Vivo Experiments. IEEE Transactions on Biomedical Engineering, 67(11), pp.3203-3210. 

 Our project also provided interdisciplinary training for a team of graduate students specializing in electronic devices, signal processing and data analytics and experimental neuroscience. 

Last Modified: 12/30/2020
Modified by: Piya Pal