"Machine learning (ML) is a powerful and cost-efficient method for establishing the relevant parameters of a noisy signal, and a further development of such concepts within the context of wound care could enable early detection of complications, shorten healing time and facilitate individualized care approaches. This project has expanded our knowledge about the pitfalls and potentials of the technology. It has advanced our understanding of the challenges related to this specific application of ML considerably, and the lessons learned will have important bearings on our future development efforts, ultimately for the benefit of patient safety," says Patrik Byhmer, CEO of Redsense Medical AB.

Project work

Over the course of the development project, Redsense processed noisy sensor data with AI/ML algorithms to evaluate the potential of using ML to identify relevant data within the signal. The system was trained on extensive simulation data from hemodialysis treatments.

Outcome

The task to separate noise from relevant information turned out to be more extensive than expected and a proof-of-concept (PoC) software could not be finalized within the scope of the project. Nevertheless, it was concluded that a product where a PoC software performs evaluation is a feasible and promising method which could provide the basis for a new product. Furthermore, the project provided valuable experience for the Company's wound care project as well as for future applications of AI technology to evaluate noisy signals in this context.

Background

Redsense has developed a patented technology based on fibre optic sensors attached to the vascular access sites to detect blood loss during hemodialysis. A further development of the technology can be used to measure physiological parameters in wounds by means of a microprocessor-controlled sensor layer integrated in wound dressings.

The project was funded by a grant of SEK 500,000 from Vinnova, Sweden's government agency for innovation, within the "Start your AI journey" initiative.