SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining

• Published in: Medical image analysis Vol. 86; p. 102789
• Main Authors: Billot, Benjamin, Greve, Douglas N., Puonti, Oula, Thielscher, Axel, Van Leemput, Koen, Fischl, Bruce, Dalca, Adrian V., Iglesias, Juan Eugenio
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.05.2023
• Subjects: Bayes Theorem; Brain - diagnostic imaging; CNN; Contrast and resolution invariance; Domain randomisation; Humans; Image Processing, Computer-Assisted - methods; Magnetic Resonance Imaging - methods; Neural Networks, Computer; Neuroimaging; Segmentation; CNN; Domain randomisation; Contrast and resolution invariance; Segmentation
• ISSN: 1361-8415; 1361-8423; 1361-8423
• DOI: 10.1016/j.media.2023.102789

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. [Display omitted] •A CNN to segment brain MRI scans of any contrast and resolution without retraining.•Domain-independence is achieved by training on randomised unrealistic synthetic data.•SynthSeg sustains almost the accuracy of supervised CNNs across all tested domains.•It also outperforms state-of-the-art domain adaptation, without being retrained.•The model is implemented in FreeSurfer for easy distribution and deployment.