AI Is Already Able To Detect Post-Traumatic Stress Disorder in Your Voice

New advances in Artificial Intelligence in the field of medicine: technology has been able to diagnose post-traumatic stress disorder through the analysis of patients' voice.

The medical industry has not stopped benefiting in the last decades of their marriage well matched with technology. In the field of AI, intelligent algorithms, robots or machine learning have led to great advances in recent years: from the recognition of heart attacks by telephone to express tests to detect cancer early, monitoring in real time patients, algorithms that can predict psychosis and much more.

Now, a group of researchers from New York University have used AI to detect post-traumatic stress disorder (PTSD) in war veterans by listening to the sound of patients' voices. His research, conducted together with SRI International, the research institute responsible for bringing Siri to iPhones, was published Monday in the journal Depression and Anxiety. For years this has been the most difficult disorders to diagnose. Traditional methods, such as individual clinical interviews, may be inaccurate due to the subjectivity of the doctor, or if the patient is suppressing their symptoms.

According to The New York Times, SRI and NYU spent five years developing a voice analysis program that includes human speech, but can also detect the signifiers and emotions of post-traumatic stress disorder. As The NYT reports, this is the same process that teaches automated customer service programs how to approach angry users: by listening to the minor variables and auditory markers that would be imperceptible to the human ear, the researchers say that the algorithm can diagnose PTSD with 89% accuracy, which is quite an achievement.

The researchers interviewed and searched 129 war veterans and collected 40,000 speech samples to study with the AI ​​software. They then used the audio to teach the algorithm which vocal changes correlated with the diagnoses of post-traumatic stress disorder (PTSD): a slower and more monotonous cadence was an indicator of suffering from this problem, as well as a shorter tonal range with less enunciation.

AI can detect minute changes in the voice, such as tension in the muscles of the throat and if the tongue touches the lips: all possible indicators of a diagnosis of PTSD that escape to the human ear. "We think that the indicator characteristics would reflect an agitated speech. In fact, when we saw the data, the features are flatter, more atonal. We were capturing the numbness that is so typical of patients with PTSD, "said Charles Marmar, a professor of psychiatry at New York University and one of the authors of the article.

Of course, there are still contraindications for the use of this algorithm. Being trained only with data from male war veterans, the scope of the program's potential is limited to men in the military, although it could be a proof of concept towards a more universal technology. As it is refined, speech analysis could become an effective biomarker to objectively identify the disorder, allowing physicians to accurately diagnose the people suffering from this problem, which causes great psychological and physical suffering and provide them with the support of mental health they need.