How Does a Neural Network Recognize Depression? | SBERMED AI
About company
Our solutions
Resource center

Contact us to find out more and start working with us

    * Required field

    SberMedAI LLC uses Cookies (files with information about past visits to the website) for service personalization and user convenience. To continue browsing the website, you need to enable cookies and accept our Cookies policy. You may disable cookies, but the website will become unavailable. Cookies policy


    How Does a Neural Network Recognize Depression?

    December 21 2022


    Reading time 5 minutes


    Depression is a mental disorder. Five percent of adults all over the world suffer from it1. A person with depression has a range of negative feelings: guilt, a lack of motivation, annoyance, a sense of hopelessness 2. He or she has difficulty in performing daily activities, and even sleeping, eating and working3.

    A machine which recognizes human feelings and helps to control them is like a scene from a fantasy film. Doctors, scientists and developers cooperate on implementing this. They have already achieved some results.

    Obstacles to Mental Health

    According to the World Health Organization (WHO), depression is not diagnosed and treated in a timely manner due to a lack of resources. Moreover, people with mental disorders are stigmatized in society1.

    Depression is a widespread disorder. However, many patients do not get proper treatment. In a research conducted in Japan the following difficulties of people with depression were identified4:

    • a lack of caretakers or health professionals to take care of such people and their condition

    • difficulty in accessing places, where care and help can be provided, by transport

    • confidence that treatment is ineffective

    • a desire to solve the problem on one’s own.

    Patients refused referral to psychiatrists in 63.9% of cases because they thought they did not need that4.

    The United States Preventive Services Taskforce (USPSTF) recommends depression screening as part of adult primary care. Effective diagnosis requires a large amount of data. Screening should be available to patients, and its results should be clear5.

    In an effort to solve these problems researchers use artificial intelligence (AI).

    Let’s Talk: an Understanding Companion in a Smartphone

    “Hi! How are you?” a chatbot asks a question and suggests responding with a word or a smiley. A mobile app talks to its user and determines the user’s psychological and emotional state based on answers. Then it suggests suitable content to control emotions in the form of a video or exercises6.

    Woebot7, Tess X2AI8 or Wysa9 are examples of such hi-tech companions. A neural network looks like a nice robot, a smiling girl or a penguin.

    Developers aim to ensure that a user perceives the app as a companion capable of expressing sympathy and empathy. For this purpose the program is taught the following6:

    • determining a user’s emotions and selecting suitable utterances

    • creating mood graphs with a brief comment

    • suggesting professional assistance in cases of severe depression.

    The application software combines psychology and AI. It is based on cognitive behavioral therapy, interpersonal psychotherapy and dialectical behavior therapy methods. The application uses natural language processing (NLP) 9, 10 to recognize messages from a person.

    A chatbot based on neural networks is not a substitute for professional assistance. It is designed for monitoring a person’s emotional state 24/7. The chatbots achieved initial success by improving mental health indicators of teenagers, women with postnatal depression and persons with psychoactive drug abuse11.

    The apps help overcome barriers related to stigmatization. Developers guarantee anonymity and safety: all data is encrypted, and services can be accessed using any username12.

    How Does AI Recognize Signs of Depression?

    Depression alters activity of speech areas in the brain. This causes specific changes in a person’s voice. It becomes muffled, low and weak13.

    Developers made use of this fact. Kintsugi Voice platform is integrated in call centers, telemedicine systems and patient monitoring apps. AI diagnoses depression based on short pieces of recorded voice14.

    AI is taught how to analyze activity on social networking services to identify the first signs of depression15:

    • frequency of writing posts

    • emotionally laden words in messages

    • linguistic features: prepositions, personal pronouns, word forms

    • smileys

    • participation in communities.

    Neural network conclusions often depend on language. Social networking services are used mainly by young people. Therefore, monitoring has limitations15.

    AI Diagnoses Clinical Depression

    Researchers from the University of Sheffield (UK) used AI to improve stages of diagnosis and select individual treatment16.

    AI diagnosed depression as a standard or a complicated case. For that purpose it collected and processed various types of data16:

    • negative emotions

    • personality traits

    • employment

    • race and ethnicity.

    Combining a doctor’s effort with neural network functions improved treatment outcomes. A personalized approach saved time. Many patients immediately received intensive treatment instead of standard therapy16.


    1. Depression [an online source]: the World Health Organization. URL: (accessed on: 20/11/2022).

    2. Depression [an online source]: NHS. URL: (accessed on: 20/11/2022).

    3. Depression [an online source]: National institute of Mental Health. URL: (accessed on: 20/11/2022).

    4. Kanehara A, Umeda M, Kawakami N; World Mental Health Japan Survey Group. Barriers to mental health care in Japan: Results from the World Mental Health Japan Survey. Psychiatry Clin Neurosci. 2015 Sep;69(9):523-33. doi: 10.1111/pcn.12267. Epub 2015 Feb 9. PMID: 25523280; PMCID: PMC4472610.

    5. Screening for Depression in Adults [An online source]: USPSTF. URL: (accessed on: 20/11/2022).

    6. A relational agent for mental health [An online source]: Woebot Health. URL: (accessed on: 20/11/2022).

    7. Meet Wysa [An online source]: Wysa. URL: (accessed on: 20/11/2022).

    8. A chatbot for mental health [An online source]: X2 AI. URL: (accessed on: 20/11/2022).

    9. Fitzpatrick KK, Darcy A, Vierhile M. Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial. JMIR Ment Health. 2017 Jun 6;4(2):e19. doi: 10.2196/mental.7785. PMID: 28588005; PMCID: PMC5478797.

    10. What Powers Woebot [An online source]: Woebot Health. URL: (accessed on: 20/11/2022).

    11. Darcy A, Beaudette A, Chiauzzi E, Daniels J, Goodwin K, Mariano TY, Wicks P, Robinson A. Anatomy of a Woebot® (WB001): agent guided CBT for women with postpartum depression. Expert Rev Med Devices. 2022 Apr;19(4):287-301. doi: 10.1080/17434440.2022.2075726. PMID: 35748029.

    12. FAQs [An online source]: Wysa. URL: (accessed on: 20/11/2022).

    13. Wang J, Zhang L, Liu T, Pan W, Hu B, Zhu T. Acoustic differences between healthy and depressed people: a cross-situation study. BMC Psychiatry. 2019 Oct 15;19(1):300. doi: 10.1186/s12888-019-2300-7. PMID: 31615470; PMCID: PMC6794822.

    14. Kintsugi [An online source]: Kintsugi Health. URL: (accessed on: 20/11/2022).

    15. Liu D, Feng XL, Ahmed F, Shahid M, Guo J. Detecting and Measuring Depression on Social Media Using a Machine Learning Approach: Systematic Review. JMIR Ment Health. 2022 Mar 1;9(3):e27244. doi: 10.2196/27244. PMID: 35230252; PMCID: PMC8924784.

    16. Delgadillo J, Ali S, Fleck K, Agnew C, Southgate A, Parkhouse L, Cohen ZD, DeRubeis RJ, Barkham M. Stratified Care vs Stepped Care for Depression: A Cluster Randomized Clinical Trial. JAMA Psychiatry. 2022 Feb 1;79(2):101-108. doi: 10.1001/jamapsychiatry.2021.3539. PMID: 34878526; PMCID: PMC8655665.

    Похожие статьи