How was FaceScan trained & validated?
To capture human data for training and validation of FaceScan, 2,348 subjects predominantly aged 18+ were recruited over a course of two years, at the University of Toronto (Canada), and the physical examination center of the Affiliated Hospital of Hangzhou Normal University (China).
The use of human subjects in this study was approved by the institutional review committees at both institutions. Subjects participated in a series of events capturing the data needed to train the machine learning models, as well as participating in the use of FaceScan’s technology for validation purposes.
To ensure accuracy in measurements, the computational models were also trained and validated against already well-established scientific instruments found in labs and clinics. An example of this was a proof-of-concept study involving the comparison of our technology against a traditional automated blood pressure monitor, to which accuracy was comparable.
Further information from the various clinical studies regarding subjects, data collection, signal processing, training of BP prediction methods, results and more can be provided following the joint execution of an NDA.