AI Model Diagnoses Depression With 97% Accuracy

Researchers at Kaunas University of Technology (KTU) have developed an artificial intelligence (AI) model capable of diagnosing depression with high accuracy. Unlike traditional methods, which often rely on a single type of data, this multimodal approach integrates speech analysis and brain activity data, offering a more nuanced view of emotional health.

The AI system utilizes electroencephalogram (EEG) data and vocal characteristics to assess mental health. This combination enables more precise diagnoses, with the model achieving a 97.53% accuracy rate. This marks a significant improvement over alternative approaches, which often lack the capability to capture the depth of information provided by combining these data types.

“Depression is one of the most common mental disorders, with devastating consequences for both the individual and society, so we are developing a new, more objective diagnostic method that could become accessible to everyone in the future”

Dr. Rytis Maskeliūnas

Methodology and data insights

The KTU research utilized the Multimodal Open Dataset for Mental Disorder Analysis (MODMA) to train its model. EEG recordings were taken while participants were awake, relaxed and with their eyes closed, ensuring consistency. The audio data were collected during tasks designed to elicit natural language, such as reading and describing images.

Both data sources were converted into spectrograms, which visualized changes in signals over time. The EEG data captured patterns of brainwave activity, while the speech data reflected frequency and intensity distributions. A modified DenseNet-121 deep-learning model processed these images to classify individuals as healthy or experiencing depression.

To enhance the robustness of the system, noise filters and pre-processing techniques were applied, ensuring clean and comparable datasets.

Ethical and practical considerations

While promising, the model presents certain challenges. Data scarcity, partly due to the stigma around mental health, limits the training of AI systems. Additionally, ethical considerations, including privacy concerns and the explainability of AI-driven diagnoses, remain critical.

“The main problem with these studies is the lack of data because people tend to remain private about their mental health matters." 

Dr. Rytis Maskeliūnas

The team at KTU is focused on advancing the system to provide not only accurate results but also explanations for its decisions. This aligns with the principles of explainable AI (XAI), which aim to improve user trust by making decision-making processes transparent.

Future applications

The AI model has the potential to revolutionize mental health diagnostics by enabling quicker, remote evaluations. However, additional clinical trials and refinements are necessary before widespread deployment. Researchers emphasize that ensuring the model’s reliability and interpretability will be crucial for integrating it into healthcare practices.

Reference: Yousufi M, Damaševičius R, Maskeliūnas R. Multimodal fusion of EEG and audio spectrogram for major depressive disorder recognition using modified DenseNet121. Brain Sciences. 2024;14(10):1018. doi: 10.3390/brainsci14101018

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.

This content includes text that has been generated with the assistance of AI. Technology Networks' AI policy can be found here.