Real-time posterior probabilistic TMS motor mapping with robust individualized stopping criteria
Background. Electric-field-informed regression provides high-resolution transcranial magnetic stimulation (TMS) motor maps, but conventional nonlinear fitting is too slow to update the map during acquisition, so sessions rely on predefined stimulation budgets, despite substantial inter-individual variability in the data required for stable localization. Objective. To develop a real-time analysis…
Electrical-field guided regression techniques enable high-resolution Transcranial Magnetic Stimulation (TMS) motor maps, yet standard nonlinear fitting methods are too time-consuming to provide real-time updates during a session. This delays stimulation budgets, even though people exhibit significant differences in the data required for accurate localization.
Researchers aimed to create a system that updates a posterior probability map between pulses, then compares its hotspots with functional MRI data from finger-tapping tasks.
The study examined two distinct groups of participants (12 and 13 individuals, with one group collected recently) for each hand muscle individually. Instead of using the classic per-element sigmoid fit, the research team replaced it with Spearman's rank correlation coefficient. The resulting score map was then converted into a posterior probability map using marginal-likelihood evidence. The researchers stopped the acquisition when the posterior mass within 5mm of the hotspot reached a predetermined confidence level.
The Spearman and linear scores were able to update the region of interest within milliseconds, whereas sigmoid fitting took tens of seconds. Spearman's method more accurately matched the complete-data sigmoid hotspot in the actual sessions and converged to the true source faster and more precisely in simulated subjects. Across both fast scores, the stopping criterion was achieved after a median of 74-159 pulses, ranging from 26-454 pulses per individual.
The Functional Data Index (FDI) hotspots were found to be 8.3 ± 4.3mm away from the center of gravity of the finger-tapping fMRI data, located in the precentral hand region.
In conclusion, employing rank correlation reduces computation time without sacrificing accuracy compared to sigmoid fitting. The resulting posterior probability map offers a session-specific stopping criterion, eliminating the need for a fixed pulse budget.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
