Nicole: An Intelligent Psychiatric Training platform for diagnosing Schizophrenia

Problem

With the dearth of practical training due to the less number of Schizophrenic subjects that the medical students get to interact with in real life, it's quite difficult for them to sharpen their diagnostic skillset.

Goal

To develop a virtual psychiatric training platform for Schizophrenia, to help medical students hone their skills on. The platform can also be used for diagnosis of the disease itself during normal interaction mode.

Role

Computer Vision

NLP/Speech processing

Machine Learning

Advisor

Prof. Daniel Thalmann

Prof. Nadia Thalmann

Nicole is an Intelligent Psychiatric Training platform that serves two purposes:

It can be used for detecting the possibility of a person having Schiophrenia during a normal conversation with the user.
It can act as a negative symptom Schizophrenic patient itself so as to provide psychiatric training to the medical students.

For the detection purposes the following different threads are taken into consideration:

Body movement signals including the speed of limbs, ankles, head (captured using Kinect).
Disorganized topic shifts seen very frequently in Shizophrenic patients.
Low level speech signals
For detecting hallucinations, a set of questions related to the current environment like “What do you see now?”, etc are asked to the user. These are then linked back to a visual dialog module and the correlation between the answers of the user and the model is calculated. If the correlation is missing for a long period of time with bizzare answers then this factor is given priority.