Evaluating Machine Learning Biomarkers for Major Depressive Disorder Across Neuroimaging Modalities

The study conducted a systematic evaluation of machine learning-based biomarkers for Major Depressive Disorder (MDD) across various modalities, including neuroimaging and genetic data. It involved training and testing over 2.4 million machine learning models on data from 1,801 participants. The findings revealed that accuracies for diagnostic classification ranged between 48.1% and 62.0%, indicating that integrating data from multiple neuroimaging modalities does not significantly improve model performance. The study also highlighted the challenges in achieving clinically relevant predictive accuracy for MDD diagnosis at the individual level, suggesting a need for further research and methodological advancements in precision psychiatry. For more detailed insights, you can access the full article here.