[Early prognostic of ASD: A challenge]. / Pronostiquer tôt les troubles du spectre autistique : Un défi ?

Autism Spectrum Disorders (ASD) are born in the womb generated by intrauterine genetic or environmental insult. ASD diagnostic is made at the age of 3-5 years in Europe and in the US. Relying on this, we have tested the hypothesis of identifying already at birth babies who might be diagnosed later with ASD, thereby facilitating an early use of psychoeducative techniques to attenuate the severity of the symptoms. Here, we discuss the various approaches that have been used to enable an early diagnosis. We have ourselves used an approach based on a "without a priori" machine learning analysis of all maternity biological and ultrasound data available in French maternities (around 116) in utero and after birth. This program made it possible to identify at birth almost all (96%) of babies who will be later neurotypical and around half of those who will be diagnosed with ASD. Some of the parameters allowing this identification were largely unexpected with no known links with ASD. This approach will enable an early identification of babies at risk, but also might be used to diagnose ASD later on, and perhaps could help to get a better understanding of the heterogeneity of ASD.

Title: Pronostiquer tôt les troubles du spectre autistique : Un défi ? Abstract: Les troubles du spectre de l'autisme (TSA) « naissent ¼ in utero à la suite d'évènements pathologiques génétiques ou environnementaux. Le diagnostic des TSA n'est cependant effectué que vers l'âge de 3-5 ans en Europe et aux États-Unis. Un pronostic précoce permettrait pourtant d'atténuer la sévérité des atteintes cognitives, grâce à des approches psycho-éducatives. Une large panoplie d'approches a été suggérée pour établir un pronostic précoce des TSA, se fondant sur l'imagerie cérébrale, sur des enregistrements EEG, sur des biomarqueurs sanguins ou sur l'analyse des contacts visuels. Nous avons développé une approche fondée sur l'analyse par machine learning des données biologiques et échographiques recueillies en routine, du début de la grossesse au lendemain de la naissance, dans les maternités françaises. Ce programme qui permet d'identifier la presque totalité des bébés neurotypiques et la moitié des bébés qui auront un diagnostic de TSA quelques années plus tard, permet aussi d'identifier les paramètres ayant un impact sur le pronostic. Si quelques-uns d'entre eux étaient attendus, d'autres n'ont aucun lien avec les TSA. L'étude sans a priori des données de maternité devrait ainsi permettre un pronostic des TSA dès la naissance, ainsi que de mieux comprendre la pathogenèse de ces syndromes et de les traiter plus tôt.

Machine learning analysis of pregnancy data enables early identification of a subpopulation of newborns with ASD.

To identify newborns at risk of developing ASD and to detect ASD biomarkers early after birth, we compared retrospectively ultrasound and biological measurements of babies diagnosed later with ASD or neurotypical (NT) that are collected routinely during pregnancy and birth. We used a supervised machine learning algorithm with a cross-validation technique to classify NT and ASD babies and performed various statistical tests. With a minimization of the false positive rate, 96% of NT and 41% of ASD babies were identified with a positive predictive value of 77%. We identified the following biomarkers related to ASD: sex, maternal familial history of auto-immune diseases, maternal immunization to CMV, IgG CMV level, timing of fetal rotation on head, femur length in the 3rd trimester, white blood cell count in the 3rd trimester, fetal heart rate during labor, newborn feeding and temperature difference between birth and one day after. Furthermore, statistical models revealed that a subpopulation of 38% of babies at risk of ASD had significantly larger fetal head circumference than age-matched NT ones, suggesting an in utero origin of the reported bigger brains of toddlers with ASD. Our results suggest that pregnancy follow-up measurements might provide an early prognosis of ASD enabling pre-symptomatic behavioral interventions to attenuate efficiently ASD developmental sequels.

Brain Volumes in Mice are Smaller at Birth After Term or Preterm Cesarean Section Delivery.

The rate of cesarean section (CS) delivery has steadily increased over the past decades despite epidemiological studies reporting higher risks of neonatal morbidity and neurodevelopmental disorders. Yet, little is known about the immediate impact of CS birth on the brain, hence the need of experimental studies to evaluate brain parameters following this mode of delivery. Using the solvent clearing method iDISCO and 3D imaging technique, we report that on the day of birth, whole-brain, hippocampus, and striatum volumes are reduced in CS-delivered as compared to vaginally-born mice, with a stronger effect observed in preterm CS pups. These results stress the impact of CS delivery, at term or preterm, during parturition and at birth. In contrast, cellular activity and apoptosis are reduced in mice born by CS preterm but not term, suggesting that these early-life processes are only impacted by the combination of preterm birth and CS delivery.

Treating Autism With Bumetanide: Are Large Multicentric and Monocentric Trials on Selected Populations Complementary?

In their article in the Journal, Sprengers et al.1 conclude that bumetanide does not attenuate autism spectrum disorder (ASD) despite a nominally significant treatment effect in repetitive behaviors, which is a core symptom of ASD but was defined as a secondary measure in this trial. Four earlier studies performed by 3 independent institutes, including 2 studies2,3 not mentioned by Sprengers et al., testing a total of 169 children (versus 122 placebo) showed a significant amelioration of ASD symptoms. Bumetanide also significantly attenuated behavioral features of patients with tuberous sclerosis according to another study by Sprengers' same group.4.

Surface Regularity via the Estimation of Fractional Brownian Motion Index.

The recent definition of fractional Brownian motions on surfaces has raised the statistical issue of estimating the Hurst index characterizing these models. To deal with this open issue, we propose a method which is based on a spectral representation of surfaces built upon their Laplace-Beltrami operator. This method includes a first step where the surface supporting the motion is recovered using a mean curvature flow, and a second one where the Hurst index is estimated by linear regression on the motion spectrum. The method is evaluated on synthetic surfaces. The interest of the method is further illustrated on some fetal cortical surfaces extracted from magnetic resonance images as a means to quantify the brain complexity during the gestational age.

Local Spectral Analysis of the Cerebral Cortex: New Gyrification Indices.

Gyrification index (GI) is an appropriate measure to quantify the complexity of the cerebral cortex. There is, however, no universal agreement on the notion of surface complexity and there are various methods in literature that evaluate different aspects of cortical folding. In this paper, we give two intuitive interpretations on folding quantification based on the magnitude and variation of the mean curvature of the cortical surface. We then present a local spectral analysis of the mean curvature to introduce two local gyrification indices that satisfy our interpretations. For this purpose, the graph windowed Fourier transform is extended to the framework of surfaces discretized with triangular meshes. An adaptive window function is also proposed to deal with the intersubject cortical size variability. The intrinsic nature of the method allows us to compute the degree of folding at different spatial scales. Our experiments show that while more classical surface area-based GIs may fail at differentiating deep folds from very convoluted ones, our spectral GIs overcome this issue. The method is applied to the cortical surfaces of 124 healthy adult subjects of OASIS database and average gyrification maps are computed and compared with other GI definitions. In order to illustrate the capacity of our method to capture and quantify important aspects of gyrification, we study the relationship between brain volume and cortical complexity, and design a scaling analysis with a power law model. Results indicate an allometric relation and confirm the well-known observations that larger brains are more folded. We also perform the scaling analysis at the vertex level to investigate how the degree of folding varies locally with the brain volume. Results reveal that in our healthy adult brain database, cortical regions which are the least folded on average show an increased folding complexity when brain size increases.