Predicting Mental and Neurological Illnesses Based on Cerebellar Normative Features

Area of Science:

• Neuroimaging
• Machine Learning
• Developmental Neuroscience

Background:

• Structural brain variations are linked to mental and neurological conditions.
• Predictive value of brain scans for these conditions is limited due to heterogeneity.
• The cerebellum, a less-studied region, shows potential links to these disorders.

Purpose of the Study:

• To evaluate the predictive power of cerebellar developmental deviations for various neurological and psychiatric disorders using machine learning.
• To assess the utility of normative cerebellar modeling across the lifespan.

Main Methods:

• Machine learning models were trained on data from over 27,000 participants.
• Individual deviations from normative cerebellar development were analyzed for prediction of autism spectrum disorder (ASD), bipolar disorder, schizophrenia (SZ), mild cognitive impairment, and Alzheimer's disease.
• Multiple cerebellar atlases were used to derive regional metrics including median, variance, and extreme deviations.

Main Results:

• Cerebellar lobular and voxelwise data demonstrated moderate accuracy in discriminating individuals with ASD and SZ from controls (AUC 0.56-0.65).
• Both anterior and posterior cerebellar regions contributed to the predictive models.
• The findings suggest specific cerebellar regions are implicated in ASD and SZ.

Conclusions:

• Cerebellar normative modeling is a valuable tool for predicting ASD and SZ.
• The use of multiple cerebellar atlases improved the interpretability of the predictive findings.
• This study underscores the importance of the cerebellum in neurodevelopmental and psychiatric disorders.