Uncertainty quantification of TMS simulations considering MRI segmentation errors.

Area of Science:

• Neuroscience
• Biophysics
• Medical Imaging

Background:

• Transcranial magnetic stimulation (TMS) is crucial for brain research and therapy.
• Computational electric field (E-field) dosimetry for TMS relies on head models derived from medical images.
• Segmentation inaccuracies in medical imaging introduce geometric errors, limiting the precision of TMS E-field predictions.

Purpose of the Study:

• To investigate the impact of segmentation accuracy on the fidelity of TMS-induced E-field predictions.
• To quantify how geometric uncertainties in tissue boundaries affect E-field dosimetry.
• To identify critical tissue interfaces influencing E-field prediction accuracy.

Main Methods:

• Modeled segmentation errors as geometric uncertainty in tissue boundary shapes.
• Employed a boundary element method for forward propagation analysis.
• Quantified the effect of tissue boundary uncertainties on induced cortical E-fields.

Main Results:

• E-field predictions showed minimal sensitivity to segmentation errors in scalp, skull, and white matter.
• E-field predictions were highly sensitive to cerebrospinal fluid (CSF) segmentation errors.
• Segmentation errors at CSF-gray matter interfaces increased uncertainty in gyral crowns; CSF-white matter interfaces increased uncertainty in sulci.
• Averaging cortical E-fields over regions reduced uncertainty compared to point-wise estimates.

Conclusions:

• Cerebrospinal fluid (CSF) segmentation accuracy is a primary limitation for precise cortical E-field simulations in TMS.
• Regional averaging of E-fields provides a more robust dosimetry measure, resilient to segmentation inaccuracies.
• Future TMS dosimetry should consider regional averages to enhance reliability.