RNDC-Lab: a real-time calcium imaging-based adaptive neural decoding framework

Real-time neural decoding is critical for precise neuromodulation, which is of increasing interest in the treatment of brain disorders. This paper presents RNDC-Lab, a computational framework for designing, optimizing, and experimenting with real-time, calcium imaging-based, end-to-end neural decoding systems.

RNDC-Lab integrates calcium imaging preprocessing with adaptive neural decoding using compact Deep Neural Networks (DNNs). RNDC-Lab supports dynamic retraining of a neural decoding model concurrently with inference, thereby providing robustness to changes in the statistical properties of input calcium imaging signals. We integrate NDSEP and NeuroGRS - two recently- developed tools for neural decoding and signal extraction, and optimized generation of DNN models, respectively-into RNDC-Lab. This results in the first real-time neural decoding system based on calcium imaging with a dynamic, DNN-based decoding model. This paper also introduces a number of simulated datasets to assist in system evaluation and experimentation using RNDC-Lab, and more broadly, as resources for experimentation with arbitrary neural decoding models and systems. The experimental capabilities of RNDC-Lab are further enhanced by four new datasets-the Brain Phantom, N80, N90 and Reward Learning (for multiclass classification) datasets-that we introduce in this work as companion resources for working with RNDC-Lab.

To demonstrate the capabilities of RNDC-Lab, we conduct extensive experiments to evaluate the functionality and performance of the proposed framework. The experiments incorporate the aforementioned simulated datasets as well as real calcium imaging data collected from mouse models. Our results demonstrate that the RNDC-Lab system, combined with NDSEP and NeuroGRS, achieves efficient and accurate real-time neural decoding from both real and synthetic calcium imaging data. Moreover, the results demonstrate RNDC-Lab's capability for dynamic model retraining and characterize the associated trade-offs involving decoding performance and execution time.

While the components within NDSEP, NGSynth, and NeuroGRS are used to concretely demonstrate the capabilities of RNDC-Lab in this paper, the rigorous dataflow- based architecture of RNDC-Lab enables these components to be selectively replaced with alternative modules that provide the same general functionality using different algorithms or implementations (e.g., a new module for motion correction or model retraining). This makes RNDC-Lab a powerful tool for experimenting with novel aspects of neural decoding in the context of an overall real-time framework.