What do NVIDIA, a crack team of MIT and German researchers, and depression sufferers have in common? Well, they’re the ingredients of a new AI tool designed to detect the onset of depression. 

NVIDIA’s AI research lab recently lent its tech to Tuka Alhanai, an MIT research assistant who sought to produce an effective and cost-efficient diagnostic tool to detect depression. Alhanai’s inspiration came from the results of another study made by German researchers that proved depression can be detected by looking for clues in a patient’s speech patterns and vocabulary.

While most psychotherapists use “talk therapy” to treat depression, this joint effort could make “talk diagnosis” – detecting symptoms of depression by analyzing a patient’s speech via deep learning – a valuable tool in detecting and treating this mental health issue.

According the National Institute of Mental Health, depression hits 1 in 15 adults in America at any given year. Depression remains a crippling condition that can result in major bouts, sometimes costing patients’ relationships and livelihood (Pixnio.com/people/male-men/man-depression-dark#).

To create the deep learning based-diagnostic tool, Alhanai patterned her model from the datasets of participants in a study conducted by the University of Southern California, which were in turn used by German researchers to prove that speech patterns and vocabulary could be analyzed for signs of depression. Alhanai’s team then assigned the work of training their model to a cluster of machines fitted with over 40 NVIDIA Titan X GPUs attached to deep learning libraries.

Understanding depression via a diagnostic tool of this nature holds some promise, since initial testing showed that the model could identify depression in participants via normal conversation with over 70% accuracy.

To know more about this story, read it here: https://blogs.nvidia.com/blog/2019/02/19/deep-learning-identifies-depression/?tm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+nvidiablog+%28The+NVIDIA+Blog%29

Cover photo source: www.Nvidia.com/en-au/titan/titan-xp/

Leave a Reply