Does restrictive anorexia nervosa impact brain aging? A machine learning approach to estimate age based on brain...

Area of Science:

• Neuroscience
• Radiology
• Psychiatry

Background:

• Anorexia nervosa (AN) is a severe eating disorder causing significant brain structure alterations.
• Brain aging patterns in AN are not fully understood, necessitating advanced analytical methods.

Purpose of the Study:

• To investigate brain-predicted age difference (brain-PAD) as a potential biomarker in anorexia nervosa using machine learning.
• To explore the relationship between brain aging, AN status, and clinical factors.

Main Methods:

• A machine learning model was trained on T1-weighted MRI scans from healthy individuals to predict brain age.
• Structural MRI scans from female participants with acute AN (acAN), weight-restored AN (wrAN), and healthy controls (HC) were analyzed.
• Brain-predicted age difference (brain-PAD) was calculated and compared across groups, with associations to BMI and symptom severity examined.

Main Results:

• The machine learning model demonstrated high accuracy in predicting brain age (MAE = 1.93 years, r = 0.88).
• Acute AN patients exhibited an overestimation of brain age (+2.25 years), indicating accelerated brain aging.
• Weight-restored AN patients showed normalized brain-PAD, similar to healthy controls, and brain-PAD in acAN was negatively correlated with BMI.

Conclusions:

• Acute anorexia nervosa is associated with accelerated brain aging, detectable via brain-predicted age difference.
• Brain aging may partially normalize following weight restoration in AN patients.
• Brain-PAD shows potential as a neuroimaging biomarker for AN, particularly in its acute phase.