Exploring the potential of representation and transfer learning for anatomical neuroimaging: Application to psychiatry.

The perspective of personalized medicine for brain disorders requires efficient learning models for anatomical neuroimaging-based prediction of clinical conditions. There is now a consensus on the benefit of deep learning (DL) in addressing many medical imaging tasks, such as image segmentation. However, for single-subject prediction problems, recent studies yielded contradictory results when comparing DL with Standard Machine Learning (SML) on top of classical feature extraction. Most existing comparative studies were limited in predicting phenotypes of little clinical interest, such as sex and age, and using a single dataset. Moreover, they conducted a limited analysis of the employed image pre-processing and feature selection strategies. This paper extensively compares DL and SML prediction capacity on five multi-site problems, including three increasingly complex clinical applications in psychiatry namely schizophrenia, bipolar disorder, and Autism Spectrum Disorder (ASD) diagnosis. To compensate for the relative scarcity of neuroimaging data on these clinical datasets, we also evaluate three pre-training strategies for transfer learning from brain imaging of the general healthy population: self-supervised learning, generative modeling and supervised learning with age. Overall, we find similar performance between randomly initialized DL and SML for the three clinical tasks and a similar scaling trend for sex prediction. This was replicated on an external dataset. We also show highly correlated discriminative brain regions between DL and linear ML models in all problems. Nonetheless, we demonstrate that self-supervised pre-training on large-scale healthy population imaging datasets (N≈10k), along with Deep Ensemble, allows DL to learn robust and transferable representations to smaller-scale clinical datasets (N≤1k). It largely outperforms SML on 2 out of 3 clinical tasks both in internal and external test sets. These findings suggest that the improvement of DL over SML in anatomical neuroimaging mainly comes from its capacity to learn meaningful and useful abstract representations of the brain anatomy, and it sheds light on the potential of transfer learning for personalized medicine in psychiatry.

Enhancing fetal alcohol spectrum disorders diagnosis with a classifier based on the intracerebellar gradient of volumetric undersizing.

In fetal alcohol spectrum disorders (FASD), brain growth deficiency is a hallmark of subjects both with fetal alcohol syndrome (FAS) and with non-syndromic FASD (NS-FASD, i.e., those without specific diagnostic features). However, although the cerebellum was suggested to be more severely undersized than the rest of the brain, it has not yet been given a specific place in the FASD diagnostic criteria where neuroanatomical features still count for little if anything in diagnostic specificity. We applied a combination of cerebellar segmentation tools on a 1.5 T 3DT1 brain MRI dataset from a monocentric population of 89 FASD (52 FAS, 37 NS-FASD) and 126 typically developing controls (6-20 years old), providing 8 volumes: cerebellum, vermis and 3 lobes (anterior, posterior, inferior), plus total brain volume. After adjustment of confounders, the allometric scaling relationship between these cerebellar volumes (Vi ) and the total brain or cerebellum volume (Vt ) was fitted (Vi = bVt a ), and the effect of group (FAS, control) on allometric scaling was evaluated. We then estimated for each cerebellar volume in the FAS population the deviation from the typical scaling (v DTS) learned in the controls. Lastly, we trained and tested two classifiers to discriminate FAS from controls, one based on the total cerebellum v DTS only, the other based on all the cerebellar v DTS, comparing their performance both in the FAS and the NS-FASD group. Allometric scaling was significantly different between FAS and control group for all the cerebellar volumes (p < .001). We confirmed the excess of total cerebellum volume deficit (v DTS = -10.6%) and revealed an antero-inferior-posterior gradient of volumetric undersizing in the hemispheres (-12.4%, 1.1%, 2.0%, respectively) and the vermis (-16.7%, -9.2%, -8.6%, repectively). The classifier based on the intracerebellar gradient of v DTS performed more efficiently than the one based on total cerebellum v DTS only (AUC = 92% vs. 82%, p = .001). Setting a high probability threshold for >95% specificity of the classifiers, the gradient-based classifier identified 35% of the NS-FASD to have a FAS cerebellar phenotype, compared to 11% with the cerebellum-only classifier (pFISHER = 0.027). In a large series of FASD, this study details the volumetric undersizing within the cerebellum at the lobar and vermian level using allometric scaling, revealing an anterior-inferior-posterior gradient of vulnerability to prenatal alcohol exposure. It also strongly suggests that this intracerebellar gradient of volumetric undersizing may be a reliable neuroanatomical signature of FAS that could be used to improve the specificity of the diagnosis of NS-FASD.

Asynchronous neural maturation predicts transition to psychosis.

AIM: Neuroimaging-based machine-learning predictions of psychosis onset rely on the hypothesis that structural brain anomalies may reflect the underlying pathophysiology. Yet, current predictors remain difficult to interpret in light of brain structure. Here, we combined an advanced interpretable supervised algorithm and a model of neuroanatomical age to identify the level of brain maturation of the regions most predictive of psychosis. METHODS: We used the voxel-based morphometry of a healthy control dataset (N = 2024) and a prospective longitudinal UHR cohort (N = 82), of which 27 developed psychosis after one year. In UHR, psychosis was predicted at one year using Elastic-Net-Total-Variation (Enet-TV) penalties within a five-fold cross-validation, providing an interpretable map of distinct predictive regions. Using both the whole brain and each predictive region separately, a brain age predictor was then built and validated in 1605 controls, externally tested in 419 controls from an independent cohort, and applied in UHR. Brain age gaps were computed as the difference between chronological and predicted age, providing a proxy of whole-brain and regional brain maturation. RESULTS: Psychosis prediction was performant with 80 ± 4% of area-under-curve and 69 ± 5% of balanced accuracy (P < 0.001), and mainly leveraged volumetric increases in the ventromedial prefrontal cortex and decreases in the left precentral gyrus and the right orbitofrontal cortex. These regions were predicted to have delayed and accelerated maturational patterns, respectively. CONCLUSION: By combining an interpretable supervised model of conversion to psychosis with a brain age predictor, we showed that inter-regional asynchronous brain maturation underlines the predictive signature of psychosis.

OpenBHB: a Large-Scale Multi-Site Brain MRI Data-set for Age Prediction and Debiasing.

Prediction of chronological age from neuroimaging in the healthy population is an important issue because the deviations from normal brain age may highlight abnormal trajectories towards brain disorders. As a first step, ML models have emerged to predict chronological age from brain MRI, as a proxy measure of biological age. However, there is currently no consensus w.r.t which Machine Learning (ML) model is best suited for this task, largely because of a lack of public benchmark. Furthermore, new large emerging population neuroimaging datasets are often biased by the acquisition center images are coming from. This bias heavily deteriorates models generalization capacities, especially for Deep Learning (DL) algorithms that are known to overfit rapidly on the simplest features (known as simplicity bias). Here we propose a new public benchmarking resource, namely Open Big Healthy Brains (OpenBHB), along with a challenge for both brain age prediction and site-effect removal through a representation learning framework. OpenBHB is large-scale, gathering >5K 3D T1 brain MRI from Healthy Controls (HC) and highly multi-sites, aggregating >60 centers worldwide and 10 studies. OpenBHB is expected to grow both in terms of available modalities and number of subjects. All OpenBHB datasets are uniformly preprocessed, including quality check, with container technologies that consist in: 3D Voxel-Based Morphometry maps (VBM from CAT12), quasi-raw (simple linear alignment of images), and Surface-Based Morphometry indices (SBM, from FreeSurfer). The OpenBHB challenge is permanent and we provide all tools, materials and tutorials for participants to easily submit and benchmark their model against each other on a public leaderboard.

Lithium prevents grey matter atrophy in patients with bipolar disorder: an international multicenter study.

BACKGROUND: Lithium (Li) is the gold standard treatment for bipolar disorder (BD). However, its mechanisms of action remain unknown but include neurotrophic effects. We here investigated the influence of Li on cortical and local grey matter (GM) volumes in a large international sample of patients with BD and healthy controls (HC). METHODS: We analyzed high-resolution T1-weighted structural magnetic resonance imaging scans of 271 patients with BD type I (120 undergoing Li) and 316 HC. Cortical and local GM volumes were compared using voxel-wise approaches with voxel-based morphometry and SIENAX using FSL. We used multiple linear regression models to test the influence of Li on cortical and local GM volumes, taking into account potential confounding factors such as a history of alcohol misuse. RESULTS: Patients taking Li had greater cortical GM volume than patients without. Patients undergoing Li had greater regional GM volumes in the right middle frontal gyrus, the right anterior cingulate gyrus, and the left fusiform gyrus in comparison with patients not taking Li. CONCLUSIONS: Our results in a large multicentric sample support the hypothesis that Li could exert neurotrophic and neuroprotective effects limiting pathological GM atrophy in key brain regions associated with BD.

Using structural MRI to identify bipolar disorders - 13 site machine learning study in 3020 individuals from the ENIGMA Bipolar Disorders Working Group.

Bipolar disorders (BDs) are among the leading causes of morbidity and disability. Objective biological markers, such as those based on brain imaging, could aid in clinical management of BD. Machine learning (ML) brings neuroimaging analyses to individual subject level and may potentially allow for their diagnostic use. However, fair and optimal application of ML requires large, multi-site datasets. We applied ML (support vector machines) to MRI data (regional cortical thickness, surface area, subcortical volumes) from 853 BD and 2167 control participants from 13 cohorts in the ENIGMA consortium. We attempted to differentiate BD from control participants, investigated different data handling strategies and studied the neuroimaging/clinical features most important for classification. Individual site accuracies ranged from 45.23% to 81.07%. Aggregate subject-level analyses yielded the highest accuracy (65.23%, 95% CI = 63.47-67.00, ROC-AUC = 71.49%, 95% CI = 69.39-73.59), followed by leave-one-site-out cross-validation (accuracy = 58.67%, 95% CI = 56.70-60.63). Meta-analysis of individual site accuracies did not provide above chance results. There was substantial agreement between the regions that contributed to identification of BD participants in the best performing site and in the aggregate dataset (Cohen's Kappa = 0.83, 95% CI = 0.829-0.831). Treatment with anticonvulsants and age were associated with greater odds of correct classification. Although short of the 80% clinically relevant accuracy threshold, the results are promising and provide a fair and realistic estimate of classification performance, which can be achieved in a large, ecologically valid, multi-site sample of BD participants based on regional neurostructural measures. Furthermore, the significant classification in different samples was based on plausible and similar neuroanatomical features. Future multi-site studies should move towards sharing of raw/voxelwise neuroimaging data.

Will machine learning applied to neuroimaging in bipolar disorder help the clinician? A critical review and methodological suggestions.

OBJECTIVES: The existence of anatomofunctional brain abnormalities in bipolar disorder (BD) is now well established by magnetic resonance imaging (MRI) studies. To create diagnostic and prognostic tools, as well as identifying biologically valid subtypes of BD, research has recently turned towards the use of machine learning (ML) techniques. We assessed both supervised ML and unsupervised ML studies in BD to evaluate their robustness, reproducibility and the potential need for improvement. METHOD: We systematically searched for studies using ML algorithms based on MRI data of patients with BD until February 2019. RESULT: We identified 47 studies, 45 using supervised ML techniques and 2 including unsupervised ML analyses. Among supervised studies, 43 focused on diagnostic classification. The reported accuracies for classification of BD ranged between (a) 57% and 100%, for BD vs healthy controls; (b) 49.5% and 93.1% for BD vs patients with major depressive disorder; and (c) 50% and 96.2% for BD vs patients with schizophrenia. Reported accuracies for discriminating subjects genetically at risk for BD (either from control or from patients with BD) ranged between 64.3% and 88.93%. CONCLUSIONS: Although there are strong methodological limitations in previous studies and an important need for replication in large multicentric samples, the conclusions of our review bring hope of future computer-aided diagnosis of BD and pave the way for other applications, such as treatment response prediction. To reinforce the reliability of future results we provide methodological suggestions for good practice in conducting and reporting MRI-based ML studies in BD.

Predicting the individual risk of psychosis conversion in at-risk mental state (ARMS): a multivariate model reveals the influence of nonpsychotic prodromal symptoms.

To improve the prediction of the individual risk of conversion to psychosis in UHR subjects, by considering all CAARMS' symptoms at first presentation and using a multivariate machine learning method known as logistic regression with Elastic-net shrinkage. 46 young individuals who sought help from the specialized outpatient unit at Sainte-Anne hospital and who met CAARMS criteria for UHR were assessed, among whom 27 were reassessed at follow-up (22.4 ± 6.54 months) and included in the analysis. Elastic net logistic regression was trained, using CAARMS items at baseline to predict individual evolution between converters (UHR-P) and non-converters (UHR-NP). Elastic-net was used to select the few CAARMS items that best predict the clinical evolution. All validations and significances of predictive models were computed with non-parametric re-sampling strategies that provide robust estimators even when the distributional assumption cannot be guaranteed. Among the 25 CAARMS items, the Elastic net selected 'obsessive-compulsive symptoms' and 'aggression/dangerous behavior' as risk factors for conversion while 'anhedonia' and 'mood swings/lability' were associated with non-conversion at follow-up. In the ten-fold stratified cross-validation, the classification achieved 81.8% of sensitivity (P = 0.035) and 93.7% of specificity (P = 0.0016). Non-psychotic prodromal symptoms bring valuable information to improve the prediction of conversion to psychosis. Elastic net logistic regression applied to clinical data is a promising way to switch from group prediction to an individualized prediction.

Genetic Influence on the Sulcal Pits: On the Origin of the First Cortical Folds.

The influence of genes on cortical structures has been assessed through various phenotypes. The sulcal pits, which are the putative first cortical folds, have for long been assumed to be under tight genetic control, but this was never quantified. We estimated the pit depth heritability in various brain regions using the high quality and large sample size of the Human Connectome Project pedigree cohort. Analysis of additive genetic variance indicated that their heritability ranges between 0.2 and 0.5 and displays a regional genetic control with an overall symmetric pattern between hemispheres. However, a noticeable asymmetry of heritability estimates is observed in the superior temporal sulcus and could thus be related to language lateralization. The heritability range estimated in this study reinforces the idea that cortical shape is determined primarily by nongenetic factors, which is consistent with the important increase of cortical folding from birth to adult life and thus predominantly constrained by environmental factors. Nevertheless, the genetic cues, implicated with various local levels of heritability in the formation of sulcal pits, play a fundamental role in the normal gyral pattern development. Quantifying their influence and identifying the underlying genetic variants would provide insight into neurodevelopmental disorders.

Prediction of activation patterns preceding hallucinations in patients with schizophrenia using machine learning with structured sparsity.

Despite significant progress in the field, the detection of fMRI signal changes during hallucinatory events remains difficult and time-consuming. This article first proposes a machine-learning algorithm to automatically identify resting-state fMRI periods that precede hallucinations versus periods that do not. When applied to whole-brain fMRI data, state-of-the-art classification methods, such as support vector machines (SVM), yield dense solutions that are difficult to interpret. We proposed to extend the existing sparse classification methods by taking the spatial structure of brain images into account with structured sparsity using the total variation penalty. Based on this approach, we obtained reliable classifying performances associated with interpretable predictive patterns, composed of two clearly identifiable clusters in speech-related brain regions. The variation in transition-to-hallucination functional patterns not only from one patient to another but also from one occurrence to the next (e.g., also depending on the sensory modalities involved) appeared to be the major difficulty when developing effective classifiers. Consequently, second, this article aimed to characterize the variability within the prehallucination patterns using an extension of principal component analysis with spatial constraints. The principal components (PCs) and the associated basis patterns shed light on the intrinsic structures of the variability present in the dataset. Such results are promising in the scope of innovative fMRI-guided therapy for drug-resistant hallucinations, such as fMRI-based neurofeedback.

Features and Determinants of Lacune Shape: Relationship With Fiber Tracts and Perforating Arteries.

BACKGROUND AND PURPOSE: Lacunes are a major manifestation of cerebral small vessel disease. Although still debated, the morphological features of lacunes may offer mechanistic insights. We systematically analyzed the shape of incident lacunes in cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, a genetically defined small vessel disease. METHODS: A total of 88 incident lacunes from 57 patients were segmented from 3-dimensional T1 magnetic resonance images and 3 dimensionally reconstructed. Anatomic location, diameter, volume, surface area, and compactness of lacunes were assessed. The shape was analyzed using a size, orientation, and position invariant spectral shape descriptor. We further investigated the relationship with perforating arteries and fiber tracts. RESULTS: Lacunes were most abundant in the centrum semiovale and the basal ganglia. Diameter, volume, and surface area of lacunes in the basal ganglia and centrum semiovale were larger than in other brain regions. The spectral shape descriptor revealed a continuum of shapes with no evidence for distinct classes of lacunes. Shapes varied mostly in elongation and planarity. The main axis and plane of lacunes were found to align with the orientation of perforating arteries but not with fiber tracts. CONCLUSIONS: Elongation and planarity are the primary shape principles of lacunes. Their main axis and plane align with perforating arteries. Our findings add to current concepts on the mechanisms of lacunes.

Shape of the Central Sulcus and Disability After Subcortical Stroke: A Motor Reserve Hypothesis.

BACKGROUND AND PURPOSE: Both brain and cognitive reserves modulate the clinical impact of chronic brain diseases. Whether a motor reserve also modulates the relationships between stroke and disability is unknown. We aimed to determine whether the shape of the central sulcus, a marker of the development of underlying motor connections, is independently associated with disability in patients with a positive history of small subcortical ischemic stroke. METHODS: Shapes of central sulci were reconstructed from high-resolution magnetic resonance imaging and ordered without supervision according to a validated algorithm in 166 patients with a positive history of small subcortical ischemic stroke caused by CADASIL (Cerebral Autosomal Dominant Arteriopathy With Subcortical Infarcts and Leukoencephalopathy), a severe monogenic cerebral small vessel disease affecting young patients. Ordinal logistic regression modeling was used to test the relationships between modified Rankin scale, a disability scale strongly weighted toward motor disability, and sulcal shape. RESULTS: Modified Rankin scale was strongly associated with sulcal shape, independent of age, sex, and level of education (proportional odds ratio =1.19, 95% confidence interval =1.06-1.35; P=0.002). Results remained significant after further adjustment for brain atrophy, volume of lacunes, and volume of white matter hyperintensities of presumed vascular origin. CONCLUSIONS: The severity of disability in patients with a positive history of small subcortical ischemic stroke caused by a severe cerebral small vessel disease is related to the shape of the central sulcus, independently of the main determinants of disability. These results support the concept of a motor reserve that could modulate the clinical severity in patients with a positive history of small subcortical ischemic stroke.

Loss of venous integrity in cerebral small vessel disease: a 7-T MRI study in cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL).

BACKGROUND AND PURPOSE: Previous pathological studies in humans or in animal models have shown alterations of small arteries and veins within white matter lesions in cerebral small vessel disease. We aimed to evaluate in vivo, the integrity of the cerebral venous network using high-resolution MRI both within and outside white matter hyperintensities in cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL). METHODS: High-resolution T2*-weighted images were obtained at 7-T in 13 CADASIL patients with no or only mild symptoms and 13 age- and sex-matched controls. Macroscopic veins were automatically counted in the centrum semiovale and compared between patients and controls. In addition, T2* was compared between groups in the normal-appearing white matter. RESULTS: Vein density was found lower in CADASIL patients compared with that in controls (-14.6% in patients, P<0.001). This was detected both within and outside white matter hyperintensities. Mean T2*, that is presumably inversely related to the venous density, was also found increased in normal-appearing white matter of patients (+7.2%, P=0.006). All results were independent from the extent of white matter hyperintensities. CONCLUSIONS: A significant reduction in the number of visible veins was observed in the centrum semiovale of CADASIL patients both within and outside white matter hyperintensities, together with an increase of T2* in the normal-appearing white matter. Additional studies are needed to decipher the exact implication of such vasculature changes in the appearance of white matter lesions.

Prediction of post-traumatic stress disorder in family members of ICU patients: a machine learning approach.

PURPOSE: Post-traumatic stress disorder (PTSD) can affect family members of patients admitted to the intensive care unit (ICU). Easily accessible patient's and relative's information may help develop accurate risk stratification tools to direct relatives at higher risk of PTSD toward appropriate management. METHODS: PTSD was measured 90 days after ICU discharge using validated instruments (Impact of Event Scale and Impact of Event Scale-Revised) in 2374 family members. Various supervised machine learning approaches were used to predict PTSD in family members and evaluated on an independent held-out test dataset. To better understand variables' contributions to PTSD predicted probability, we used machine learning interpretability methods on the best predictive algorithm. RESULTS: Non-linear ensemble learning tree-based methods showed better predictive performances (Random Forest-area under curve, AUC = 0.73 [0.68-0.77] and XGBoost-AUC = 0.73 [0.69-0.78]) than regularized linear models, kernel-based models, or deep learning models. In the best performing algorithm, most important features that positively contributed to PTSD's predicted probability were all non-modifiable factors, namely, lower patient's age, longer duration of ICU stay, relative's female sex, lower relative's age, relative being a spouse/child, and patient's death in ICU. A sensitivity analysis in bereaved relatives did not alter the algorithm's predictive performance. CONCLUSION: We propose a machine learning-based approach to predict PTSD in relatives of ICU patients at an individual level. In this model, PTSD is mostly influenced by non-modifiable factors.

Longitudinal MicroRNA Signature of Conversion to Psychosis.

BACKGROUND AND HYPOTHESIS: The emergence of psychosis in ultra-high-risk subjects (UHR) is influenced by gene-environment interactions that rely on epigenetic mechanisms such as microRNAs. However, whether they can be relevant pathophysiological biomarkers of psychosis' onset remains unknown. STUDY DESIGN: We present a longitudinal study of microRNA expression, measured in plasma by high-throughput sequencing at baseline and follow-up, in a prospective cohort of 81 UHR, 35 of whom developed psychosis at follow-up (converters). We combined supervised machine learning and differential graph analysis to assess the relative weighted contribution of each microRNA variation to the difference in outcome and identify outcome-specific networks. We then applied univariate models to the resulting microRNA variations common to both strategies, to interpret them as a function of demographic and clinical covariates. STUDY RESULTS: We identified 207 microRNA variations that significantly contributed to the classification. The differential network analysis found 276 network-specific correlations of microRNA variations. The combination of both strategies identified 25 microRNAs, whose gene targets were overrepresented in cognition and schizophrenia genome-wide association studies findings. Interpretable univariate models further supported the relevance of miR-150-5p and miR-3191-5p variations in psychosis onset, independent of age, sex, cannabis use, and medication. CONCLUSIONS: In this first longitudinal study of microRNA variation during conversion to psychosis, we combined 2 methodologically independent data-driven strategies to identify a dynamic epigenetic signature of the emergence of psychosis that is pathophysiologically relevant.

Attention-gated 3D CapsNet for robust hippocampal segmentation.

Purpose: The hippocampus is organized in subfields (HSF) involved in learning and memory processes and widely implicated in pathologies at different ages of life, from neonatal hypoxia to temporal lobe epilepsy or Alzheimer's disease. Getting a highly accurate and robust delineation of sub-millimetric regions such as HSF to investigate anatomo-functional hypotheses is a challenge. One of the main difficulties encountered by those methodologies is related to the small size and anatomical variability of HSF, resulting in the scarcity of manual data labeling. Recently introduced, capsule networks solve analogous problems in medical imaging, providing deep learning architectures with rotational equivariance. Nonetheless, capsule networks are still two-dimensional and unassessed for the segmentation of HSF. Approach: We released a public 3D Capsule Network (3D-AGSCaps, https://github.com/clementpoiret/3D-AGSCaps) and compared it to equivalent architectures using classical convolutions on the automatic segmentation of HSF on small and atypical datasets (incomplete hippocampal inversion, IHI). We tested 3D-AGSCaps on three datasets with manually labeled hippocampi. Results: Our main results were: (1) 3D-AGSCaps produced segmentations with a better Dice Coefficient compared to CNNs on rotated hippocampi (p=0.004, cohen's d=0.179); (2) on typical subjects, 3D-AGSCaps produced segmentations with a Dice coefficient similar to CNNs while having 15 times fewer parameters (2.285M versus 35.069M). This may greatly facilitate the study of atypical subjects, including healthy and pathological cases like those presenting an IHI. Conclusion: We expect our newly introduced 3D-AGSCaps to allow a more accurate and fully automated segmentation on atypical populations, small datasets, as well as on and large cohorts where manual segmentations are nearly intractable.

A fast and robust hippocampal subfields segmentation: HSF revealing lifespan volumetric dynamics.

The hippocampal subfields, pivotal to episodic memory, are distinct both in terms of cyto- and myeloarchitectony. Studying the structure of hippocampal subfields in vivo is crucial to understand volumetric trajectories across the lifespan, from the emergence of episodic memory during early childhood to memory impairments found in older adults. However, segmenting hippocampal subfields on conventional MRI sequences is challenging because of their small size. Furthermore, there is to date no unified segmentation protocol for the hippocampal subfields, which limits comparisons between studies. Therefore, we introduced a novel segmentation tool called HSF short for hippocampal segmentation factory, which leverages an end-to-end deep learning pipeline. First, we validated HSF against currently used tools (ASHS, HIPS, and HippUnfold). Then, we used HSF on 3,750 subjects from the HCP development, young adults, and aging datasets to study the effect of age and sex on hippocampal subfields volumes. Firstly, we showed HSF to be closer to manual segmentation than other currently used tools (p < 0.001), regarding the Dice Coefficient, Hausdorff Distance, and Volumetric Similarity. Then, we showed differential maturation and aging across subfields, with the dentate gyrus being the most affected by age. We also found faster growth and decay in men than in women for most hippocampal subfields. Thus, while we introduced a new, fast and robust end-to-end segmentation tool, our neuroanatomical results concerning the lifespan trajectories of the hippocampal subfields reconcile previous conflicting results.

Understanding the relationship between cerebellar structure and social abilities.

BACKGROUND: The cerebellum contains more than 50% of all neurons in the brain and is involved in a broad range of cognitive functions, including social communication and social cognition. Inconsistent atypicalities in the cerebellum have been reported in individuals with autism compared to controls suggesting the limits of categorical case control comparisons. Alternatively, investigating how clinical dimensions are related to neuroanatomical features, in line with the Research Domain Criteria approach, might be more relevant. We hypothesized that the volume of the "cognitive" lobules of the cerebellum would be associated with social difficulties. METHODS: We analyzed structural MRI data from a large pediatric and transdiagnostic sample (Healthy Brain Network). We performed cerebellar parcellation with a well-validated automated segmentation pipeline (CERES). We studied how social communication abilities-assessed with the social component of the Social Responsiveness Scale (SRS)-were associated with the cerebellar structure, using linear mixed models and canonical correlation analysis. RESULTS: In 850 children and teenagers (mean age 10.8 ± 3 years; range 5-18 years), we found a significant association between the cerebellum, IQ and social communication performance in our canonical correlation model. LIMITATIONS: Cerebellar parcellation relies on anatomical boundaries, which does not overlap with functional anatomy. The SRS was originally designed to identify social impairments associated with autism spectrum disorders. CONCLUSION: Our results unravel a complex relationship between cerebellar structure, social performance and IQ and provide support for the involvement of the cerebellum in social and cognitive processes.

Significant correlation between a set of genetic polymorphisms and a functional brain network revealed by feature selection and sparse Partial Least Squares.

Brain imaging is increasingly recognised as an intermediate phenotype to understand the complex path between genetics and behavioural or clinical phenotypes. In this context, a first goal is to propose methods to identify the part of genetic variability that explains some neuroimaging variability. Classical univariate approaches often ignore the potential joint effects that may exist between genes or the potential covariations between brain regions. In this paper, we propose instead to investigate an exploratory multivariate method in order to identify a set of Single Nucleotide Polymorphisms (SNPs) covarying with a set of neuroimaging phenotypes derived from functional Magnetic Resonance Imaging (fMRI). Recently, Partial Least Squares (PLS) regression or Canonical Correlation Analysis (CCA) have been proposed to analyse DNA and transcriptomics. Here, we propose to transpose this idea to the DNA vs. imaging context. However, in very high-dimensional settings like in imaging genetics studies, such multivariate methods may encounter overfitting issues. Thus we investigate the use of different strategies of regularisation and dimension reduction techniques combined with PLS or CCA to face the very high dimensionality of imaging genetics studies. We propose a comparison study of the different strategies on a simulated dataset first and then on a real dataset composed of 94 subjects, around 600,000 SNPs and 34 functional MRI lateralisation indexes computed from reading and speech comprehension contrast maps. We estimate the generalisability of the multivariate association with a cross-validation scheme and demonstrate the significance of this link, using a permutation procedure. Univariate selection appears to be necessary to reduce the dimensionality. However, the significant association uncovered by this two-step approach combining univariate filtering and L1-regularised PLS suggests that discovering meaningful genetic associations calls for a multivariate approach.

Striatal and extrastriatal dopamine transporter in cannabis and tobacco addiction: a high-resolution PET study.

The dopamine (DA) system is known to be involved in the reward and dependence mechanisms of addiction. However, modifications in dopaminergic neurotransmission associated with long-term tobacco and cannabis use have been poorly documented in vivo. In order to assess striatal and extrastriatal dopamine transporter (DAT) availability in tobacco and cannabis addiction, three groups of male age-matched subjects were compared: 11 healthy non-smoker subjects, 14 tobacco-dependent smokers (17.6 ± 5.3 cigarettes/day for 12.1 ± 8.5 years) and 13 cannabis and tobacco smokers (CTS) (4.8 ± 5.3 cannabis joints/day for 8.7 ± 3.9 years). DAT availability was examined in positron emission tomography (HRRT) with a high resolution research tomograph after injection of [11C]PE2I, a selective DAT radioligand. Region of interest and voxel-by-voxel approaches using a simplified reference tissue model were performed for the between-group comparison of DAT availability. Measurements in the dorsal striatum from both analyses were concordant and showed a mean 20% lower DAT availability in drug users compared with controls. Whole-brain analysis also revealed lower DAT availability in the ventral striatum, the midbrain, the middle cingulate and the thalamus (ranging from -15 to -30%). The DAT availability was slightly lower in all regions in CTS than in subjects who smoke tobacco only, but the difference does not reach a significant level. These results support the existence of a decrease in DAT availability associated with tobacco and cannabis addictions involving all dopaminergic brain circuits. These findings are consistent with the idea of a global decrease in cerebral DA activity in dependent subjects.