FetMRQC: A robust quality control system for multi-centric fetal brain MRI.

Fetal brain MRI is becoming an increasingly relevant complement to neurosonography for perinatal diagnosis, allowing fundamental insights into fetal brain development throughout gestation. However, uncontrolled fetal motion and heterogeneity in acquisition protocols lead to data of variable quality, potentially biasing the outcome of subsequent studies. We present FetMRQC, an open-source machine-learning framework for automated image quality assessment and quality control that is robust to domain shifts induced by the heterogeneity of clinical data. FetMRQC extracts an ensemble of quality metrics from unprocessed anatomical MRI and combines them to predict experts' ratings using random forests. We validate our framework on a pioneeringly large and diverse dataset of more than 1600 manually rated fetal brain T2-weighted images from four clinical centers and 13 different scanners. Our study shows that FetMRQC's predictions generalize well to unseen data while being interpretable. FetMRQC is a step towards more robust fetal brain neuroimaging, which has the potential to shed new insights on the developing human brain.

U-shape short-range extrinsic connectivity organisation around the human central sulcus.

The central sulcus is probably one of the most studied folds in the human brain, owing to its clear relationship with primary sensory-motor functional areas. However, due to the difficulty of estimating the trajectories of the U-shape fibres from diffusion MRI, the short structural connectivity of this sulcus remains relatively unknown. In this context, we studied the spatial organization of these U-shape fibres along the central sulcus. Based on high quality diffusion MRI data of 100 right-handed subjects and state-of-the-art pre-processing pipeline, we first define a connectivity space that provides a comprehensive and continuous description of the short-range anatomical connectivity around the central sulcus at both the individual and group levels. We then infer the presence of five major U-shape fibre bundles at the group level in both hemispheres by applying unsupervised clustering in the connectivity space. We propose a quantitative investigation of their position and number of streamlines as a function of hemisphere, sex and functional scores such as handedness and manual dexterity. Main findings of this study are twofold: a description of U-shape short-range connectivity along the central sulcus at group level and the evidence of a significant relationship between the position of three hand related U-shape fibre bundles and the handedness score of subjects.

Diffusion MRI: Assessment of the Impact of Acquisition and Preprocessing Methods Using the BrainVISA-Diffuse Toolbox.

Diffusion MR images are prone to severe geometric distortions induced by head movement, eddy-current and inhomogeneity of magnetic susceptibility. Various correction methods have been proposed that depend on the choice of the acquisition settings and potentially provide highly different data quality. However, the impact of this choice has not been evaluated in terms of the ratio between scan time and preprocessed data quality. This study aims at investigating the impact of six well-known preprocessing methods, each associated to specific acquisition settings, on the outcome of diffusion analyses. For this purpose, we developed a comprehensive toolbox called Diffuse which automatically guides the user to the best preprocessing pipeline according to the input data. Using MR images of 20 subjects from the HCP dataset, we compared the six pre-processing pipelines regarding the following criteria: the ability to recover brain's true geometry, the tensor model estimation and derived indices in the white matter, and finally the spatial dispersion of six well known connectivity pathways. As expected the pipeline associated to the longer acquisition fully repeated with reversed phase-encoding (RPE) yielded the higher data quality and was used as a reference to evaluate the other pipelines. In this way, we highlighted several significant aspects of other pre-processing pipelines. Our results first established that eddy-current correction improves the tensor-fitting performance with a localized impact especially in the corpus callosum. Concerning susceptibility distortions, we showed that the use of a field map is not sufficient and involves additional smoothing, yielding to an artificial decrease of tensor-fitting error. Of most importance, our findings demonstrate that, for an equivalent scan time, the acquisition of a b0 volume with RPE ensures a better brain's geometry reconstruction and local improvement of tensor quality, without any smoothing of the image. This was found to be the best scan time/data quality compromise. To conclude, this study highlights and attempts to quantify the strong dependence of diffusion metrics on acquisition settings and preprocessing methods.