Scientists Have Developed an Interpretive Approach for Recognizing Depressive disorder

Scientists have developed an interpretive approach that makes it possible to diagnose depressive disorder with 82.6% accuracy based on certain indicators of patients' nervous system. For this purpose, the authors used brain activity images to construct functional networks reflecting the interactions of different brain parts in sick and healthy people. Then, by evaluating the characteristics of each node and the connections between them, the researchers made a disease classification scale. The new technology could potentially be used to diagnose depressive disorders at early stage. The results of the study, supported by a grant from the Russian Science Foundation (RSF), are published in Chaos: An Interdisciplinary Journal of Nonlinear Science.

Major depressive disorder is a common mental illness that affects about 280 million people worldwide. Patients lose interest in activities, experience insomnia, sleepiness, weakness, guilt and self-deprecation. At the same time, subjective scales, patient reports and physician observations are used to identify depressive disorder, which is not always possible to accurately determine the severity of the illness. In recent years, to improve diagnosis of mental disorders, doctors have begun to use machine learning algorithms based on artificial neural networks similar to those that function in the brain. However, the results obtained with their help are difficult to interpret. As a result, it is impossible to identify the main parameters on the basis of which the neural network makes a decision.

Researchers from Immanuel Kant Baltic Federal University and Plovdiv Medical University (Bulgaria) proposed an interpretive approach to effectively identify depressive disorder. For this purpose, they used images obtained by functional magnetic resonance imaging, a method based on measuring the intensity of blood flow: when an area of the brain was excited, blood flow to it increased. Then, based on the images obtained, the researchers built graphs — complex networks in which nodes (points) simulated brain areas, and the connections between the nodes — the edges of the graph — corresponded to the interactions between the areas. Based on the resulting graphs, the scientists compared brain function in 35 patients with depressive disorder and in 50 healthy people, and then, using machine learning techniques, tried to divide the feature space of both groups.

The proposed approach made it possible to distinguish between sick and healthy people with an accuracy of 82.6%. At the same time, the groups differed better if the authors took into account a certain set of network metrics in their analysis: node strength corresponding to brain area activity, number of edges, that is, number of interactions between areas, and clustering coefficient, showing the extent to which nodes tended to cluster together. If the researchers applied the features separately or added other insignificant network characteristics, however, the algorithm did not work correctly.

Thus, the method identified those characteristics of the functional brain network that differed in sick and healthy subjects. At the same time, the algorithm recognized differences not in individual local connections between brain regions (as most methods do), but in global characteristics of the entire neuronal network. The new technique will make it possible to track global changes in the brain structure in people with depressive disorder, as well as to provide clinical practice with a reliable way to diagnose this disease.

Andrey Andreev, Ph.D. in Physics and Mathematics, Senior Researcher at IKBFU:
In the future, we plan to highlight the characteristic features of functional brain networks of healthy people and patients with major depressive disorder at different levels of the brain based on the obtained data. This will allow us to select the most important biomarkers that should be analyzed to determine the disease on the basis of magnetic resonance imaging.