Internal validation was performed based on data from Oggway ltd, Tel Aviv, Israel. This involved short sequential, targeted behavioural tracking aimed to train the device and include split-sample validation.
vHive performed external validation was performed as fully independent of University of Surrey and Zoetis Centre for Digital Innovation with a different population and breed mix including such variables as;
- UK vs Israel
- Long sequences
- Natural behaviour
- Continuous tagging
- Types of validation
vHive performed video recordings of behaviour and observer interpretation which was painstakingly corresponded to the device outputs. In total 11.5 hours of tagged footage of 51 dogs was created and assessed.
Our diagnostic effectiveness (the proportion of records correctly classified), was very high.
>95% for walk, trot, canter/gallop, eat, drink and headshake
>90% for static/inactive and sleep
The positive predictive values were all >95%, meaning that when the device classified a certain behaviour, the dog was usually displaying it.
After our internal and external, independent validation research, the vHive team has concluded that not only does the device detect various kinds of behaviour that are correctly linked to the eight predefined behavioural states, but it promises to point to early indications of health and welfare issues in a dog that wears it such as normal itching behaviour during nighttime versus behaviour that can be indicative of an ear infection. Also, its portability makes it feasible for mass commercialisation.