Automated extraction of electrode coordinates from structural MRI to assess tDCS placement accuracy

Abstract

Accurate electrode placement is essential for studies using transcranial direct current stimulation (tDCS) since it determines whether - and to what extent - the intended brain region was modulated. Using in-scanner tDCS, placement accuracy can be determined by extracting electrode coordinates from structural MRI and by comparing actual and intended positions. Previous studies extracted coordinates manually, which is time-consuming and prone to error. Here, we present an algorithm to automate this process. The proposed algorithm first segments electrode and electrode-gel blobs visible on MRI. It then separates this segmentation into individual electrode regions and extracts their coordinates as the centroids of these regions. Finally, coordinates are assigned to the intended target positions using linear assignment. We evaluated the algorithm with T1-weighted MRI data from 65 individuals acquired at diferent scanners with diferent magnetic field strengths, image resolutions, and two high-definition montages by comparing the extracted coordinates with manual annotations from two independent readers. The median localization error of the algorithm was 2.4 mm (IQR 1.5 mm). The median variability between human readers was 2.7 mm (IQR 1.7 mm). Distances between algorithm-derived coordinates and the reference were smaller than distances between readers. Our proposed algorithm obviates the need for manual electrode segmentation and allows an objective evaluation of electrode placement accuracy for in-scanner tDCS. It may be used to relate electrode position errors to stimulation efects.