VRneck Solution is an R&D project that aims to introduce innovative solutions to the process of diagnosis and rehabilitation of cervical spine impairments. The project is being implemented by the University of Łódź and Edventure Research Lab – a company from the Betacom S.A. capital group.
VRneck Solution is a methodological novelty in diagnosis and rehabilitation of cervical spine — based on the analysis of the patient’s head movements, it creates the so-called grid of segment motion of cervical spine and cervicothoracic junction. An analysis of parameters of this grid should enable assessing the functioning of individual cervical spine segments. The key parameters are the accuracy and specificity of this grid, i.e. to what extent this test can detect functional impairment of individual cervical spine segments (accuracy) or not detect an impairment that does not exist (specificity). It is estimated that the minimum value of these parameters should not be lower than 70%.
The technological innovative elements of the project are:
- Diagnostics workstation
It is an element of development (not research) works, and its innovativeness consists in using elements of virtual reality (VR goggles) and the so-called headtracker for the analysis of functions of the cervical spine based on the analysis of the patient’s head movements.
- A system for automatic diagnosis and treatment based on artificial intelligence methods
The innovativeness of this element of the system consists in its hierarchical and layered structure, which, in addition to learning data, will also leverage domain-specific expert knowledge in the field of cervical spine diagnostics and treatment. There are no automatic, universal, commercially available techniques for creating hierarchical models (whether in the form of ready-to-use libraries or scientific studies). When it comes to our project, we are planning to develop models based on artificial intelligence and machine learning. It will recognize the components and other simple diagnostic elements in order to create the final diagnostic system by applying additional versions of approximate / fuzzy rules requesting components of subsystems results for the observed input data.
An interesting aspect of the project is the determination of the minimum size of data sets required to feed data to the artificial intelligence system. On the one hand, AI systems require data sets with the maximum possible or even “unlimited” amount of data. On the other, there are practical limitations regarding the ability to obtain data at a given time during the project.
The VRneck project will include clinical trials with a minimum of 400 patients. Data from clinical trials will be enhanced with historical data from the databases of other hospitals. The challenge is to determine whether the patient data will be representative enough and show the entire spectrum of cases that the VRneck Solution system should learn to diagnose and treat. The solution of this problem is the structure of the artificial intelligence system under development. It will have a layered structure, and the layers, especially the first one, will consist of several smaller subsystems (e.g. classifiers) that will solve smaller “sub-problems” while using smaller data sets.
A subsidy application has been submitted to the National Center for Research and Development. We are planning to conduct a panel meeting very soon (mid-July this year), which will bring together the experts from the National Center for Research and Development and the representatives of the VRneck Solution project applicants.