Deep learning-based source imaging framework (DeepSIF) that integrates computational neural models that are motivated by biophysics into the source imaging pipeline to deliver an accurate, usable, and objective Electrophysiological Source Imaging solver.
A deep neural network (DNN) was trained to recognize macroscale scalp electromagnetic signals, using artificial training data produced by a large-scale brain dynamics model made up of connected NMMs that resembled realistic mesoscale neuronal activity. When provided noninvasive scalp electrophysiological recordings, the DeepSIF algorithm can once be trained, to produce estimates of spatiotemporally distributed brain activities. Through a series of numerical tests, we proved that the trained DNN can reliably and precisely reconstruct the sources’ location, size, and temporal dynamics. Additionally, by imaging the epileptogenic brain tissue in a cohort of 20 patients with drug-resistant focal epilepsy, we empirically showed the generalizability of DeepSIF.
The effectiveness of this new strategy was evaluated in a study by Bin He, a professor of biomedical engineering at Carnegie University, which was just published in Proceedings of the National Academy of Sciences (PNAS), by mapping sensory and cognitive brain responses in 20 healthy adults. The identification of epileptogenic tissue in a cohort of 20 patients with drug-resistant epilepsy has also undergone rigorous validation by contrasting the outcomes of non-invasive AI-based imaging with those of invasive measures and surgical excision.
“With this new approach, you only need a centralized location for brain modeling and deep neural network training,” Bin He. “After collecting data in a clinical or research setting, clinicians and researchers could remotely send the data to centralized, well-equipped deep neural networks and quickly obtain accurate analysis results. This technology can speed up diagnosis and help neurologists and neurosurgeons plan surgery better and faster.”
“As part of a decades-long effort to develop innovative, non-invasive functional neuroimaging solutions, I have been working on a dynamic brain imaging technology that can provide precision, be effective and easy to use, to better serve clinicians and researchers,” said Bin He.