VRneck SOLUTION – automatic diagnosis module

VRneck SOLUTION – automatic diagnosis module
20 May 2020

Other important components of the VRneck SOLUTION system include automatic diagnosis module and automatic treatment program selection module. Both modules will be created using artificial intelligence (AI) methods. Those methods will leverage intelligent data analysis algorithms which learn to detect dependencies between data captured by VRneck devices (and other available sources of information) and functional impairment in various segments.

The AI system design will use all diagnostics data. This will create appropriate training data where multimodal information sources may be used by the models being developed in a mutually supportive manner. It is currently a popular approach in many areas of analyzing healthcare data. A generic functionality of both modules will consist in identifying, based on signals from VRneck and data from other sources, segments in which functional impairment has been detected. Due to highly varied nature of healthcare data and VRneck data, and their frequently complex structure, the AI system should be a multi-layer system. The lowest layer will be a kind of feature extraction system. Parameters with a more complex structure will have separate classifiers (e.g. classifier of the VRneck motion error grid), whose results will be parameters in the next layer. The second layer will consist in a complex reasoning system using machine learning, whose results may still be blurred. The last layer will be responsible for delivering the expected diagnosis. A characteristic feature of the system being developed is multi-parameter diagnosis. It implies the necessity of creating separate models for intelligent reasoning for each of the expected diagnosis parameters. The research stage will specify the most effective AI approaches which will provide generalization abilities for the purpose of delivering diagnosis.

The layered, and thus staged, nature of this system will correspond to the clinical analysis and reasoning conducted by doctors and physiotherapists, which means that the decisions of the AI system will be more understandable for them. The division into smaller reasoning/classification subsystems will also result in smaller requirements in terms of size of learning collections. The third benefit consists in easier control and modification of such system. It is therefore assumed that artificial intelligence in the VRneck SOLUTION will not be a closed system. It will be possible to systematically correct and teach the system, which will be supervised by a group of experts (doctors and physiotherapists). The purpose of such solution is limiting the number of incorrect diagnoses. The automatic treatment program selection module will be based on similar rules as the automatic diagnosis module. In such case, the treatment program will not be limited to the VRneck method. Based on other diagnostic parameters and a base of expert knowledge in methods of treating patients with cervical spine disorders, the system should also suggest appropriate treatment methods beyond the VRneck system.