A comprehensive analysis of the impact of head model extent on electric field predictions in transcranial current stimulation

Abstract

Objective. MRI-based head models are used to predict the electric field (E-field) in the brain in transcranial current stimulation. The standard field of view of clinical MRI often only covers the head down to the skull base, which has usually lead to models truncated at the level of the nose. Although recent pipelines can artificially extend the head model to the neck, the need for implementing full head models preserving skull holes such as the foramen magnum remains controversial. The objective of this work is to analyse the impact of head model extent on E-field accuracy, with emphasis on specific electrode montages. Approach. A full head model containing an open foramen magnum and a cut head model with closed skull were compared in terms of predicted E-field. Several electrode montages, including fronto-occipital montages recently used in validation studies, were simulated. Local and global metrics were used to evaluate the error for both E-field magnitude and distribution, along with tangential and normal components over different cortical areas. The percentage of current flowing through the lower head was also computed. Results. Regarding E-field magnitude, small relative differences (RDs) below 7% were found in grey matter for classical montages. Although considerably higher RDs near 50% were found for fronto-occipital montages, absolute errors of 0.1 V m−1 were only found in non-targeted regions such as the cerebellum. Differences in tangential and normal E-fields were similar and followed the same trend observed for E-field magnitude. Our results also showed a high correlation between the percentage of current shunted through the lower head and the absolute E-field differences. Significance. The influence of head model extent on E-field accuracy depends on electrode montage. Standard cut head models provide sufficiently accurate predictions for both E-field magnitude and distribution in targeted brain areas. However, fronto-occipital montages exhibited larger errors, which might be considered in further validation studies.