About
Key words
Digital twin, neuromuscular system, location intelligence, data structuring, feature learning, context representation, spatiotemporal regression, multi-objective dynamic optimization
Abstract
Increasing investments into Internet of things (IoT), big data analytics, and artificial intelligence propelled the development of digital replicas of real-world entities in a form of digital twins. These cyber-physical systems offer advanced monitoring, analytical, and predictive capabilities and have become a new major trend in computer science. Gartner ranks digital twins amongst ten key technologies of 2019, with expected 37% annual rate from the current 2 billion USD to 15 billion USD in 2023 and 26 billion USD by 2025. In this context, a particular attention is directed towards medical and health domains, due to the significance of their potential impacts. Digital twins can provide significant support in treatments of patients by predicting problems before they occur, help finding optimal solution, and, thus, reduce the risks as well as increase the effectiveness of rehabilitation.
Today, the use of digital twins is limited to highly controlled environments and smart machines, whereas development of technologies to replicate more complex systems, such as those related to functions of human body, still faces significant challenges. These include:
- Handling a variety of heterogeneous data streams that is required for learning the behaviour of monitored individuals requires significant improvements in automatic data alignment and structuring methodologies.
- The existing medical data fusion and feature learning methods are still mainly focused on extracting features from individual data source and, thus, need to be significantly improved in order to be able to fully exploit complementarities in heterogeneous data streams.
- Linking biomedical measurements with environmental and lifestyle factors that is required to bring laboratory studies into real-world environments calls for substantial advances in contextual feature extraction.
- Approaches to monitoring microhabitats in which we live need to be improved as high scattering of environmental sensors results in large spatial and temporal information gaps.
- The need for personalization of digital twins requires dynamic models to be optimized and adapted to monitored subjects, which is still beyond the capacity of contemporary optimisation algorithms.
Within the proposed work program, we plan to build on our extensive past research in order to address these challenges and introduce a digital twin, capable of replicating functional parameters of the human neuromuscular system in the actual environment. With the aging society, neuromuscular diseases are becoming a major health related risk and the leading cause of work incapability with the total attributed costs in Slovenia reaching as high as 2% of its GDP.
Our programme group joins leading experts in processing neuromuscular data streams with experts in spatiotemporal data analytics, semantic data processing and optimizations that will bring about this ambitious agenda through a streamlined iterative work plan in a co-creative manner.