University of Alberta researchers have skilled a machine learning model to identify people with post-traumatic stress dysfunction with 80 per cent accuracy by analyzing text data. The model may at some point function an accessible and cheap screening device to assist well being professionals in detecting and diagnosing PTSD or different psychological well being problems by means of telehealth platforms.
Psychiatry PhD candidate Jeff Sawalha, who led the challenge, carried out a sentiment evaluation of text from a dataset created by Jonathan Gratch at USC’s Institute for Creative Technologies. Sentiment evaluation includes taking a big physique of data, such because the contents of a sequence of tweets, and categorizing them -; for instance, seeing what number of are expressing optimistic ideas and what number of are expressing detrimental ideas.
We needed to strictly have a look at the sentiment evaluation from this dataset to see if we may correctly identify or distinguish people with PTSD simply utilizing the emotional content material of those interviews.”
Jeff Sawalha, psychiatry PhD candidate
The text within the USC dataset was gathered by means of 250 semi-structured interviews performed by a man-made character, Ellie, over video conferencing calls with 188 people with out PTSD and 87 with PTSD.
Sawalha and his group have been capable of identify people with PTSD by means of scores indicating that their speech featured primarily impartial or detrimental responses.
“This is in line with a number of the literature round emotion and PTSD. Some people are typically impartial, numbing their feelings and perhaps not saying an excessive amount of. And then there are others who categorical their detrimental feelings.”
The course of is undoubtedly advanced. For instance, even a easy phrase like “I did not hate that” may show difficult to categorize, defined Russ Greiner, examine co-author, professor within the Department of Computing Science and founding scientific director of the Alberta Machine Intelligence Institute. However, the truth that Sawalha was capable of glean details about which people had PTSD from the text data alone opens the door to the potential of making use of related fashions to different datasets with different psychological well being problems in thoughts.
“Text data is so ubiquitous, it is so out there, you will have a lot of it,” Sawalha stated. “From a machine learning perspective, with this a lot data, it could be higher capable of study among the intricate patterns that assist differentiate people who’ve a selected psychological sickness.”
Next steps contain partnering with collaborators on the U of A to see whether or not integrating different forms of data, comparable to speech or movement, may assist enrich the model. Additionally, some neurological problems like Alzheimer’s in addition to some psychological well being problems like schizophrenia have a powerful language element, Sawalha defined, making them one other potential space to research.Source:Journal reference:Sawalha, J., et al. (2022) Detecting Presence of PTSD Using Sentiment Analysis From Text Data. Frontiers in Psychiatry. doi.org/10.3389/fpsyt.2021.811392.
https://www.news-medical.net/news/20220407/Machine-learning-model-can-identify-people-with-PTSD-by-analyzing-text-data.aspx
