CNS*2026 Halifax

Introduction
Functional Source Separation (FSS) extends Blind Source Separation (BSS) by incorporating prior knowledge about the functional characteristics of neural responses during source extraction. FSS has previously been used to estimate primary motor (FS_M1) and somatosensory (FS_S1) cortical representations from stimulus-evoked EEG responses [1], and to identify FS_M1 in passive subjects [2]. Here we propose a population-based FSS approach that estimates spatial filters from pooled cross-subject data. We test whether these population-derived filters can identify motor and sensory cortical sources in new individuals using resting-state EEG


Methods
EEG recordings were obtained from 22 healthy subjects during median nerve stimulation (Sensory condition), weak isometric handgrip (Motor condition), and resting state with eyes open and closed. FSS was applied in two ways: (i) FSS_individual, computed separately for each subject, and (ii) FSS_population, estimated from pooled cross-subject data. Both approaches were used to identify FS_S1 and FS_M1 sources in both hemispheres. The similarity between sources extracted with the two approaches was assessed using mutual information (MI), time correlation coefficient (TCC), and Kullback–Leibler (KL) divergence. Cortico-muscular coherence (CMC) was computed in the Motor condition to evaluate the functional relevance of the motor sources.


Results
In the Motor condition, CMC between the population-derived motor source and the prime mover muscle during isometric handgrip did not differ from the coherence obtained using the individual motor source in either hemisphere (p = 0.3). Similarly, the population-derived sensory source showed responses to median nerve stimulation comparable to those obtained with the individual sensory source (p = 0.1). Across all four sources, the broadband activity of each population-derived source showed higher MI and TCC, and lower KL divergence, with its corresponding individual source than with other sources (p < 0.03).

Discussion
These results show that motor and sensory cortical sources can be identified from resting-state EEG using population-derived FSS filters that generalize to new subjects. This approach may be particularly useful in clinical settings where task-related activity is difficult to obtain, such as in patients with stroke or limb amputation [3,4]. Because FSS is deterministic, it may also improve the reproducibility of EEG-based studies of regional neuronal activity. Future work should extend this framework to larger populations, pathological conditions, and additional cortical regions such as auditory and visual cortices.


References
1. Barbati, G., Sigismondi, R., Zappasodi, F., Porcaro, C. et al. (2006). Functional source separation from magnetoencephalographic signals. Hum. Brain Mapp. doi:10.1002/hbm.20232
2. Porcaro, C., Cottone, C., Cancelli, A., Salustri, C., Tecchio, F. (2018). Functional semi-blind source separation identifies primary motor area without active motor execution. Int. J. Neural Syst. doi:10.1142/S0129065717500472
3. Delcamp, C., Srinivasan, R., Cramer, S. C. (2024). EEG provides insights into motor control and neuroplasticity during stroke recovery. Stroke. doi:10.1161/STROKEAHA.124.048458
4. Liu, S., Fu, W., Wei, C., Ma, F. et al. (2022). Interference of unilateral lower limb amputation on motor imagery rhythm and remodeling of sensorimotor areas. Front. Hum. Neurosci.doi:10.3389/fnhum.2022.1011463





Acknowledgement
The authors thank Annalisa Pascarella, Gian Luca Loli, Rosario Ribecco, and Filippo Zappasodi for their scientific contributions over the years, and the participants who took part in the recordings.