diseases

by D. Scott Peterson, Constructor University

Credit: Pixabay/CC0 Public Domain

Major depressive disorder (MDD) is a serious mental health condition that impacts individuals of all ages, including children and adolescents. Early detection and diagnosis, especially at an earlier age, is crucial for effective prevention and treatment. But current methods remain challenging.

Dr. Amir Jahanian-Najafabadi from Constructor University has been conducting a series of studies exploring the use of advanced computational models applied to electroencephalography (EEG), a non-invasive brain monitoring technique, to improve the detection of MDD in children and adolescent patients.

"The primary goal of this line of research is to enhance early diagnosis of neuropsychiatric and neurological disorders in children and adolescents by analyzing brain activity termed electroencephalography (EEG) as a non-invasive method, " said Dr. Jahanian-Najafabadi.

"Additionally, in collaboration with international partners, such as neurologists, we have been investigating the effects of specific medications such as fampridine on multiple sclerosis patients, and assessing their impact on symptoms, as well as on neuropsychological, physiological and structural aspects."

By analyzing brain activity through functional connectivity and a graph-based network approach, Dr. Jahanian-Najafabadi and his team aim to better understand the brain connectivity patterns associated with the disorder.

"For the last six years, we have applied machine learning and deep learning models to classify various disorders in comparison to healthy individuals, " he said.

To process EEG data, the researchers then developed a structured method that prepares the data, removes noise and extracts key connectivity and associated measures. These measures, which capture the strength and direction of interactions between different brain regions, were subsequently analyzed across various frequency bands.

The study, now published in the proceedings of the 2024 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), used data from 214 children and adolescents, 44 of whom had MDD. Machine learning models, such as Convolutional Neural Network and Random Forest, were utilized to train and classify MDD cases based on these brain connectivity patterns.

The research, however, was not limited to data and diagnostics modeling: "We also sought to contribute to the development of personalized treatment approaches, " Dr. Jahanian-Najafabadi explained.

"Several of our studies have already been published as scientific papers or book chapters, and our work continues to deepen our understanding of how brain activity can support increased accuracy in more clinical diagnoses, and track patient improvement across different age groups."

Results showed that certain connectivity measures, particularly those focusing on direct brain interactions, were highly effective in distinguishing individuals with MDD from healthy control subjects. The best-performing measure, known as the partial directed coherence factor, achieved an accuracy score close to perfect.

These findings suggest that some brain connectivity features are more useful than others in identifying MDD, which could lead to improved diagnostic tools. However, some methods, such as those involving indirect influences, did not perform as well, indicating areas for future refinement.

Overall, these studies highlight the potential of EEG-based machine learning and deep learning models for early MDD detection in young individuals.

"Ultimately, we hope that these efforts will complement existing clinical assessments and interviews conducted by medical specialists, while also enhancing the overall diagnostic and therapeutic process, " said Dr. Jahanian-Najafabadi.

More information: A. Jahanian Najafabadi et al, Resting-State Functional Connectivity in Children and Adolescents with Major Depressive Disorder: A Deep Learning Approach Using High-density EEG, 2024 IEEE Signal Processing in Medicine and Biology Symposium (SPMB) (2025). DOI: 10.1109/SPMB62441.2024.10842259