Electric field calculations based on numerical methods and increasingly realistic head models are more and more used in research on Transcranial Magnetic Stimulation (TMS). However, they are still far from being established as standard tools for the planning and analysis in practical applications of TMS. Here, we start by delineating three main challenges that need to be addressed to unravel their full potential. This comprises (i) identifying and dealing with the model uncertainties, (ii) establishing a clear link between the induced fields and the physiological stimulation effects, and (iii) improving the usability of the tools for field calculation to the level that they can be easily used by non-experts. We then introduce a new version of our pipeline for field calculations (www.simnibs.org) that substantially simplifies setting up and running TMS and tDCS simulations based on Finite-Element Methods (FEM). We conclude with a brief outlook on how the new version of SimNIBS can help to target the above identified challenges.

Despite advances in data augmentation and transfer learning, convolutional neural networks (CNNs) difficultly generalise to unseen domains. When segmenting brain scans, CNNs are highly sensitive to changes in resolution and contrast: even within the same MRI modality, performance can decrease across datasets. Here we introduce SynthSeg, the first segmentation CNN robust against changes in contrast and resolution. SynthSeg is trained with synthetic data sampled from a generative model conditioned on segmentations. Crucially, we adopt a domain randomisation strategy where we fully randomise the contrast and resolution of the synthetic training data. Consequently, SynthSeg can segment real scans from a wide range of target domains without retraining or fine-tuning, which enables straightforward analysis of huge amounts of heterogeneous clinical data. Because SynthSeg only requires segmentations to be trained (no images), it can learn from labels obtained by automated methods on diverse populations (e.g., ageing and diseased), thus achieving robustness to a wide range of morphological variability. We demonstrate SynthSeg on 5,000 scans of six modalities (including CT) and ten resolutions, where it exhibits unparallelled generalisation compared with supervised CNNs, state-of-the-art domain adaptation, and Bayesian segmentation. Finally, we demonstrate the generalisability of SynthSeg by applying it to cardiac MRI and CT scans.