Machine Learning Techniques in Neuroimaging for Psychiatric Disorders

Machine learning applied to neuroimaging has transformed the study of psychiatric disorders by enabling the identification of subtle patterns in brain structure and function that elude traditional univariate analyses. Supervised approaches such as support vector machines and deep neural networks have been used to distinguish individuals with conditions such as schizophrenia, bipolar disorder and major depressive disorder from healthy controls, achieving diagnostic accuracies typically in the 70–90% range. Unsupervised and semi-supervised methods facilitate discovery of subtypes and trajectories by clustering high-dimensional imaging features. Advances in feature engineering, including extraction of regional volumes, cortical thickness and connectivity matrices from structural and functional MRI, underpin these classifiers. More recently, deep learning architectures—convolutional and belief networks—have provided automated hierarchical feature learning, reducing reliance on manual region-of-interest selection. Ensemble strategies further enhance robustness by combining multiple model predictions or multimodal data such as genetic markers and resting-state signals. Rigorous validation through cross-validation and independent cohorts has become essential to ensure generalisability. Collectively, these methods offer the promise of objective biomarkers for diagnosis, prognosis and treatment stratification across diverse clinical settings.

Research from Nature Portfolio

Deep learning has been shown to extract invariant neuromorphometric features in schizophrenia. A study employing a deep belief network to learn hierarchical representations from structural MRI data demonstrated improved discrimination between patients and controls, highlighting alterations in frontal, temporal and insular cortices as well as subcortical structures. The deep belief network outperformed a conventional support vector machine classifier, illustrating the potential of layered neural architectures to capture complex, distributed brain alterations in psychiatric conditions.

Machine Learning Techniques in Neuroimaging for Psychiatric Disorders publication trend

The graph below shows the total number of articles in machine learning techniques in neuroimaging for psychiatric disorders across all publications each year (not limited to Nature Index journals).

Technical terms

Support Vector Machine (SVM): Supervised algorithm that identifies an optimal hyperplane separating classes in a high-dimensional feature space.

Deep Learning: Set of methods using neural networks with multiple layers to learn hierarchical feature representations directly from raw data.

Multivariate Pattern Analysis (MVPA): Statistical approach analysing the joint information across multiple voxels or regions to distinguish between experimental conditions.

Ensemble Learning: Technique combining predictions from several models or modalities to improve classification accuracy and generalisability.

Cross-validation: Resampling procedure dividing data into subsets to train and test a model iteratively, providing an estimate of its predictive performance on unseen data.

References

1. Magnetic resonance imaging–based machine learning classification of schizophrenia spectrum disorders: a meta‐analysis. Psychiatry and Clinical Neurosciences (2024).
2. Using deep belief network modelling to characterize differences in brain morphometry in schizophrenia. Scientific Reports (2016).
3. A Hybrid Machine Learning Method for Fusing fMRI and Genetic Data: Combining both Improves Classification of Schizophrenia. Frontiers in Human Neuroscience (2010).
4. Identifying Schizophrenia Using Structural MRI With a Deep Learning Algorithm. Frontiers in Psychiatry (2020).
5. Towards artificial intelligence in mental health by improving schizophrenia prediction with multiple brain parcellation ensemble-learning. Schizophrenia (2019).