Researchers are using machine learning to discover patterns in patients’ speech and voice that allow diagnosis of diseases

Briefing

Researchers are using machine learning to discover patterns in patients’ speech and voice that allow diagnosis of diseases

February 28, 2017

Briefing

  • Heart Disease Identification – Doctors from Mayo Clinic and Beyond Verbal identified 13 vocal characteristics associated with coronary artery disease, discovering patterns were more visible when patients talked about negative experiences
  • ADHD Diagnosis – Audioprofiling diagnosed Attention Deficit Hyperactivity Disorder (ADHD) in children with 90% accuracy, discovering that speaking syllables in less equal length can be good indicator for disease
  • PTSD Monitoring – Department of Veterans Affairs and Cogito used voice analysis app to monitor military members with Post-Traumatic Stress Disorder (PTSD), with app also used at Massachusetts General Hospital to track patients with bipolar disorder and depression
  • Traumatic Brain Injury Tracking – U.S. Army along with MIT’s Lincoln Lab created algorithm to diagnose mild traumatic brain injury by analyzing voice patterns, such as prolonged syllable and vowel sounds, as well as difficulty pronouncing phrases
  • Parkinson’s Disease Analysis – Parkinson’s Voice Initiative achieved near perfect (98.6%) accuracy in diagnosing Parkinson’s disease by studying 30-second voice recordings of patients
  • Challenges Remain – Including privacy concerns, if voice becomes identifiable to personal diseases, and how algorithms will work with non-English speakers

Accelerator

Sector

Healthcare/Health Sciences

Organization

AudioProfiling, Beyond Verbal, Cogito Corp., MIT Lincoln Laboratory, Mayo Clinic, U.S. Department of Veteran Affairs, United States Army (USA)

Source

Original Publication Date

February 13, 2017

Leave a comment