Artificial Intelligence Shows Promise in Detection of Anxiety Disorders, Depression


Investigators establish that audio and/or facial video features have been most analyzed by machine learning, followed by electroencephalography signals.

Artificial intelligence (AI) tools show promise in overcoming the limitations of traditional anxiety disorders and/or depression, according to the results of a study published in Springer.

Investigators established that audio and/or facial video features have been most analyzed, followed by electroencephalography (EEG) signals, to detect anxiety disorders and/or depression.

Traditional screening tools include the Columbia Suicide Screen, Risk of Suicide Questionnaire, Suicidal Ideation Questionnaire, and more.

These screening programs are often used in schools to assess suicide risk, according to investigators.

However, these traditional screening tools have limitations, such as a high prevalence of false positives, a lack of resources because of funding for the assessment programs in schools, others demands on educators and school counselors.

Many individuals identified may be of white ethnicity, so the tools may not be as effective in identifying at-risk individuals of different ethnicities, investigators said.

Additionally, there are no proven accurate biological markers for suicide that were integrated into clinical practice. However, it has been reported that Apolipoprotein E and interleukin-6 show promise at being promising biomarkers. Changed sleep architecture is a biomarker for suicidal behavior and thoughts.

Adolescents often express risk factors for suicide on social media instead of sharing the information with their health care providers, investigators said.

Traditional machine learning tools are a sub-field of AI, which involved a sequence of steps to “train” the machine. Machine learning tools work well with small input data so they can “learn” from it.

Typically, it works automatically with the data and without human intervention. However, investigators still must extract and select features for the machine learning.

Machine learning models used for the classification of diseases include the support vector machine, decision tree, probabilistic neural networks, and artificial neural networks. They have been used for the classification of mental illnesses, such as Alzheimer disease, depression, Parkinson disease, and schizophrenia.

However, investigators noted that deep learning models often have several layers between the input and output that are used for classification, which allows these tools to learn large input data before predicting an outcome.

Advanced learning models can also extract and select data automatically.

Some of the most used tools for classification are the convolutional neural network, long short-term memory, and autoencoders.

Investigators found that publicly available databases were used most often for machine learning models, followed by data obtained from hospitals or research centers.

Furthermore, investigators observed that audio and/or facial features were the most common forms of data, followed by EEG signals and texts from social media, to diagnose anxiety disorders and/or depression.

They also found that there were more studies on the detection of depression than anxiety disorder or depression and anxiety disorder together.

There was also a spike in machine learning, peaking in 2020, for the use of detection of anxiety disorders and/or depression, which could be because of the increase in anxiety disorders and depression during the COVID-19 pandemic, especially among adolescents and children, investigators said.


Barua PD, Vicnesh J, Lih OS, et al. Artificial intelligence assisted tools for the detection of anxiety and depression leading to suicidal ideation in adolescents: a review. Cogn Neurodyn. 2022;1-22. doi: 10.1007/s11571-022-09904-0

Related Videos
© 2024 MJH Life Sciences

All rights reserved.