DEV Community

Harshith Sai Tunuguntla
Harshith Sai Tunuguntla

Posted on

Machine Learning in Mental Health

The Problem

"Mental Health"
Mental Health has been one of the major problems people face in day-to-day lives. Starting from stress to suicidal thoughts people are affected in many ways. 3 out of 10 people have been found to get impacted with Mental Health Disorders. People need to take care about Mental Health along with Physical Health.

"Into the Machine Learning"
Since we are in the new era of computers with great technologies like Machine Learning, Artificial Intelligence. This blog mainly deals about how Machine Learning helps people in resolving their Mental Health problems. We will also look at different other methodologies, methods, applications, questionnaires that have been developed. We will also look at their impact on people.

Lets start off by looking at the need for Mental Health

The World Health Report 2001 : One in Four People are affected by mental Health


Mental Health issues triggered by Pandemic pose a challenge : Doctors


Covid-19 impact on mental health un-realized : Times of India


Approximately 9.5% of Americans adults ages 18 and over suffering from mental illness : Study


Approximately 18% of people of ages 18-45 in a given year, suffer from anxiety disorder : Research

These statements clearly explain us the need for mental health and its importance. Researchers have started looking at many different methods in order to tackle the mental health. The most regular and normal way of treating mental health is by visiting a doctor who takes different assessment tests and analyze the core problem and the level of damage mental health done on particular person and starts therapy.

Several research papers have emerged in the last century which dealt with the various ideologies about mental health. Here is the quick graph showing the emerge of research in this area

Image description

Left: Graph showing an increasing trend in the number of ML mental health publications over time. Right: Frequency distribution of the different research contribution types of the papers in the review corpus.

This is a quick look at how the research has developed in this area.

The Traditional way
The traditional way of dealing with mental health is with the help of psychiatrists. Psychiatrists specialize in helping people come out of what they go through. These professionals are medical doctors who specialize in the treatment of mental, emotional, or behavioral problems. A psychiatrist can prescribe medications. They may hold therapy sessions or work with non-medical therapists to treat you. These professionals follow Sigmund Freud's theories and other more modern theories that are based on the idea that painful childhood memories in your unconscious mind are the cause of emotional troubles. They also use talk therapy and find out how you are impacted with that.

Thereby they determine the problem of what people are facing and then they start recommending medication required.

The new methodologies emerging
There have been introduction of new methodologies and ways of finding the underlying problems people are facing related to Mental Health.

Some of them include identifying the underlying problem using speech analysis. The variation in tone is fed in Machine Learning Algorithm and the underlying problems are identified.

There are several methods that are being proposed by various researchers which are not yet implemented but have a larger scope in the future. Some of them are as follows:

  • To assist in the early diagnosis and longitudinal monitoring of mental illness symptoms in everyday speech conversation. By Chang et al. Data Domain : Audio, targeting depression.
  • To support efficient treatment of PTSD, which requires objective understanding of patients’ emotional distress. By Broek et al. Data Domain : Audio, targeting PTSD.
  • To accurately detect depression from very easy to obtain motor activity. By Frogner et al. Data Domain : Accelerometer, targeting depression.
  • To assist mental health and well-being self-managing by developing a stress-detection application as part of a mobile app. By Gjoreski et al. Data Domain : Multiple (body), targeting stress.
  • To explore if mental well-being can be inferred from smartphone behavioral data and automatically tracked over time. By DeMasi and Recht. Data Domain : Mobile phone (GPS), targeting depression.
  • To effectively predict (normal, atypical, and suicidal) mental states of patients with mental health conditions to monitor suicide risk. By Alam et al. Data Domain : Multiple (body), targeting suicidal thoughts.
  • To understand deviant behaviors on online mental health communities. By Chancellor. Data Domain : Social media(Reddit), targeting eating disorder, suicide.
  • To better understand the kinds of factors that affect mental health patients who have thoughts of death or suicide. By Galiatsatos et al. Data Domain : Questionnaire (from Health record), targeting depression.
  • To help early diagnosis of mental illness to facilitate help seeking, share health progression, and optimize treatments. By Jain and Agarwal. Data Domain : Chatbot, web media activity + wearables, targeting Mental illness (generic).

The Introduction of Machine Learning in Mental Health

From the above methodologies it can be observed that researchers started to propose methodologies that include Machine Learning technology such as predicting using Natural Language Processing, detecting using sensors, social media posts etc.

So Machine Learning in Mental Health started to emerge when researchers proposed these methodologies. Speaking about machine learning, various researchers proposed different ways of using Machine Learning for resolving mental health issues.

All the research papers are categorized into (i) supervised, (ii) unsupervised, (iii) semi-supervised learning; and (v) novel techniques.

In supervised learning most often described the application of one or more of these techniques: Support Vector Machines, Random Forest, Decision Trees, k-Nearest neighbors, supervised LDA, Lasso, and Logistic Regression. Unsupervised learning uses mathematical techniques to cluster data to provide new insights. Here, the dataset only contains inputs, but no desired output labels. To discover patterns and help structure the data, clustering methods respond to the presence or absence of commonalities in each piece of data. Analysis of natural language (NLP), speech and text, presents a specialized area of ML that mostly utilizes unsupervised techniques.

The Case Study

I started my case study with finding answers to various questions that will ultimately lead to finding the outcome of Machine Learning in Mental Health.

In this process I reached out to various research papers to find out how they are solving the problem. I tried many online applications that are up and running, these applications use the methods of images, questionnaires, feedback system. I ran across chatbots, interactive sessions.

I had to find a dataset which should portray as real solution so I had to reach out to counselchat.com who had real experts answering the questions posted over internet. With this dataset we now have the questions asked by various users and the real time answers that are given by the counselors. This data can be incorporated in a chatbot to give them the most appropriate answer as the output.

Top comments (1)

Collapse
 
harshithtunuguntla profile image
Harshith Sai Tunuguntla

No not yet, there have been Machine Learning models which are trying to predict the illness, will improve a lot in upcoming days