Alcohol-related neurodevelopmental disorders: imaging and modelling markers from the brain anatomy – ANDIMBA

The ANDIMBA project will thus: (WP #1) set up a FASD research database from a large clinical monocentric recruitment with acurate clinical and cognitive phenotyping, and high resolution anatomical (T1 and Diffusion weighted) MRI data; (WP #2) work out appropriate innovative computational tools for brain segmentation and morphological variance modelling to improve to improve the object-based computational analysis of the “undergrown” brain anatomy; (WP #3) merge these clinical and methodological researches to tackle unbiased developmentally and individually-relevant neuroanatomical markers, combining them with machine learning into classifiers that could reveal a neuroanatomical FASD signature even in non-specific forms, or help predict functional prognosis.

The work is still largely in progress but early conference communications (2022) include:
- a combination of radiologist-accessible measurements of the brain, corpus callosum and cerebellar vermis could be proposed to improve the diagnosis of non-SAF FASD, the diagnosis of which poses a problem of strength of the probabilistic link between fetal alcohol and neurodevelopmental disorder
- the cerebellum as a whole shows excess volume reduction in FASD patients, which is detectable at the individual level, and there is an intracerebellar hemisphero-vermian and postero-inferior severity gradient. This represents a significant advance in the detail and completeness of the description of cerebellar involvement in FASD.
- the proposal of a new tool for the segmentation of the medial sagittal section of the corpus callosum adapted to callosal dysgenesis observed in FASD.

Apart from the clinical interest in FASD, the ANDIMBA project could help understand the developmental toxicity of alcohol revealing neuroanatomical differences in sensitivity of resilience. It will also shed light on the frequent but complex link between brain undergrowth and NDD.

In progress.

Alcohol-related neurodevelopmental disorders (NDD) are a leading cause of cognitive and behavioral disability, still under recognized in our country, partly due to the lack of specificity of the fetal alcohol spectrum disorder (FASD) clinical diagnosis in the absence the whole malformative features of fetal alcohol syndrome (FAS). Besides, their prognosis remains largely unpredictable from clinical and exposure features. Brain size is steadily affected in FASD with occasional macroscopic damage of anatomical structures, prompting the search for neuroanatomical markers of the disease. MRI-based computational neuroanatomy studies recently extended the scope of FASD-related brain abnormalities but results remains equivocal and no marker has emerged yet that could be useful for the diagnosis of non-specific forms of FASD (NS-FASD) or for the functional prognosis. We propose at least 3 potential ways to increase the odds of unveiling such good neuroanatomical markers with object-based morphometry (automatic segmentation, identification and measurement of anatomical region and structures). The main one would be to take into account that non-homothetic scaling (allometry) is rather the rule than the exception, even in case of typical development, meaning that a linear account of brain size in comparisons is potentially misleading when dealing with undergrown brains. We also believe that the FASD dysgenetic brain is prone to segmentation bias that could be induced by the normalisation process on atlas template for instance. Eventually, we expect machine learning strategies to combine markers into classifiers would help reach clinical relevance. The ANDIMBA project will thus: (WP #1) set up a FASD research database from a large clinical monocentric recruitment with comprehensive clinical and cognitive phenotyping, and hight quality anatomical (T1 and Diffusion weighted) MRI data; (WP #2) work out innovative spectral and sulcal-based hemispheric segmentation tools and allometry-sensitive power-law-based models to improve the object-based computational analysis of the undergrown brain anatomy; (WP #3) merge these clinical and methodological researches to tackle unbiased (cross-checked between two really different segmentation methods) developmentally-relevant (accounting for morphological allometries) neuroanatomical markers, combining them with machine learning into classifiers that could reveal a neuroanatomical syndromic signature even in some NS-FASD or predict functional prognosis. Apart from the potential clinical interest in FASD, the ANDIMBA project could help understand the developmental toxicity of alcohol revealing neuroanatomical differences in sensitivity of resilience. It will also shed light on the frequent but complex link between brain undergrowth and NDD.