Can Machine Learning Decode Depression in Students?
Depression and anxiety symptoms are common among university students in many regions of the world. According to the latest Center for Collegiate Mental Health report, anxiety and depression are the top reasons that college students seek counseling. The trend has been growing over the last four years.
Mental health problems like anxiety and depression can interfere with a student’s studies and hinder performance. Depression is associated with poor academic performance and dropping out of school. Traditionally clinicians have interviewed patients, asking questions about mood, lifestyle, and previous mental problems to identify whether a patient is depressed or not. That method might be something of the past. Machine learning might step in to diagnose depression in patients.
In recent years, machine learning has emerged as a possible tool to diagnose depression. Machine learning is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience without being explicitly programmed.
MIT News reports that machine-learning models have been developed that can detect words and intonations in speech that may indicate depression. Although these methods are accurate, there’s a limit to them as they depend on specific answers to specific questions to make the diagnosis.
A neural-network model
MIT researchers have developed a neural-network model that can scrutinize raw text and audio data from interviews to discover speech patterns that may point to depression. This model doesn’t need information about questions and answers to make an accurate prediction.
With further development, this model could be deployed in smartphone apps where the model would monitor a user’s text and voice for signs of emotional and mental distress and then alert someone appropriate. This could be a boon to people who are suffering from depression but don’t realize that what they are going through is depression and that they need treatment for it.
Machine learning predicts severity and length of depression
One study analyzed baseline data from over a thousand people with Major Depressive Disorder (MDD) to predict the severity and length of the participants’ depression. The researchers compared the use of traditional analytics and a machine learning approach and found that the machine learning approach was superior. Machine learning could predict the characteristics of a person’s depression more effectively using less information than traditional analytics.
Machine learning links clinical depression with biomarkers
Machine learning has also been employed to link clinical depression with biomarkers. In a study published in PLoS One, researchers used machine learning tools and traditional statistics to analyze the relationship between 67 biomarkers in 5,227 research subjects. Three biomarkers for depression were found: red cell distribution of width, serum glucose, and total bilirubin.
Machine learning and the detection of suicide
There is a close link between depression and suicide and it is notoriously difficult to predict suicide. Is it possible that machine learning can help out here as well? One pilot study used machine learning to look at the clinical data from 144 patients with mood disorders.
The researchers use clinical variables associated with suicide attempts among patients with mood disorders and other variables to ‘train’ a machine learning algorithm. The resulting algorithm was then used on ‘new’ subjects to identify them as either suicidal or not. The researchers came up with three machine learning algorithms that could distinguish between people who had attempted suicide and those that had not. The accuracy rate varied between 65% and 72%.
Research into the application of machine learning to detect the presence of depression in people is ongoing and shows much promise.