Researchers at New York University School of Medicine, led by Charles R. Marmar, used machine learning technology to create a special algorithm. It can recognize whether a person suffers from post-traumatic stress disorder. The algorithm is 89 percent accurate. This is reported by the journal Depression and Anxiety.
The algorithm is able to calculate how severe a person has the so-called post-traumatic syndrome caused by one or more traumatic events. For example, it occurs in victims of violence. The syndrome manifests itself in different ways. One of types is depression, which is caused by excessive anxiety when the patient recalls the traumatic event.
There is as yet no method of determining what exactly can become a trigger for post-traumatic syndrome as there is no method of determining the syndrome itself. In doing so, the patient may exaggerate symptomatic manifestations. Developers at the School of Medicine attempted to create an algorithm that could objectively diagnose post-traumatic stress disorder.
The study involved a group of people who had been in combat in Iraq or Afghanistan. These people suffer from post-traumatic stress disorder. The other group included veterans without the syndrome. Researchers transcribed recordings and lengthy interviews. They identified more than 40,000 speech features. These included tempo, intonation, and the length at which vowels are pronounced.
The researchers used software from SRI International, a nonprofit institute, to detect speech features. Recall that they were the ones who developed Siri. At the end of the study, the scientists identified 18 speech traits common to people suffering from post-traumatic syndrome. Which ones? They are monotone speech, poor intonation and almost no expressiveness of speech.
Artificial Intelligence was trained using these 18 features. Therefore, it can quite accurately recognize post-traumatic stress disorder in a person. Later, there are plans to implement more data and develop software for smartphones. However, in order to implement this, permission from the authorities for the clinical use of the algorithm is required.