An in–vivo validation of ESI methods with focal sources

• DOI: 10.1016/j.neuroimage.2023.120219
• Publication Date: 2023-06-10T17:01:17Z
• Authors: Annalisa Pascarella; Ezequiel Mikulan; Federica Sciacchi...; Simone Sarasso; Annalisa Rubino; Ivana Sartori; Francesco Cardinale; Flavia Zauli; Pietro Avanzini; Lino Nobili; Andrea Pigorini; Alberto Sorrentino
• Keywords: Python; Regularization; Ground truth

Electrophysiological source imaging (ESI) aims at reconstructing the precise origin of brain activity from measurements electric field on scalp. Across laboratories/research centers/hospitals, ESI is performed with different methods, partly due to ill-posedness underlying mathematical problem. However, it difficult find systematic comparisons involving a wide variety methods. Further, existing rarely take into account variability results respect input parameters. Finally, are typically using either synthetic data, or in-vivo data where ground-truth only roughly known. We use an high-density EEG dataset recorded during intracranial single pulse electrical stimulation, in which true sources substantially dipolar and their locations precisely compare ten implementation MNE-Python package: MNE, dSPM, LORETA, sLORETA, eLORETA, LCMV beamformers, irMxNE, Gamma Map, SESAME dipole fitting. perform under multiple choices parameters, assess accuracy best reconstruction, as well impact such parameters localization performance. Best reconstructions often fall within 1 cm source, most accurate methods hitting average error 1.2 outperforming least ones erring by 2.5 cm. As expected, sparsity-promoting tend outperform distributed For several regularization parameter turned out be one principle associated low SNR, despite high SNR available dataset. Depth weighting played no role for two six implementing it. Sensitivity varied widely between While would expect being solution, this not always case, some producing highly variable error, other stable error. In particular, recent provide significantly better than older we repeated tests "conventional" (32 channels) dense (64, 128, 256 recordings, observed little number channels accuracy; however, denser montages smaller spatial dispersion. Overall findings confirm that reliable technique point therefore reinforce importance may have clinical context, especially when applied identify surgical target potential candidates epilepsy surgery.