Next generation blood analytics boosted by AI

AI and machine learning at the core of fast and user-friendly point-of-care testing

The testing solution consists of a mobile application, small potentiostat and disposable sampling sensors.

Our role was to help Fepod to develop their artificial intelligence and machine learning capability for handling and analysing the blood sample data.

Together with Fepod, we designed and implemented a cloud-based machine learning operations platform (MLOps) using Azure and Databricks technologies. On top of that, we created a machine learning solution that receives patient measurement data from the blood sample through the mobile application and estimates the concentration of paracetamol or other drugs within that sample as accurately as possible. The doctor or nurse will get the results immediately from the application on their mobile phones.

“These capabilities we developed with Solita are at the core of our product. Fast, reliable and secure analysis of blood sample data is crucial for utilising the huge potential this affordable and user-friendly diagnostics method opens to modern health care. Our solution benefits both patients, health care professionals and even the whole society when resulting in streamlined testing processes, faster diagnosing and more accurate treatments”, says Jussi Pyysalo, CEO, Fepod.

Results in a tight timeline

There was a tight deadline for Fepod to be able to showcase and demo their end-to-end working solution already during a medical expo organised in July 2023 when we started our work in the beginning of April.

“A partner with extensive knowledge of AI/ML and experience in developing solutions within the healthcare industry was crucial to us. As the timeline was only 2,5 months and we had a limited dataset to work with, we needed a good team and seamless cooperation to make things happen and get a full working solution ready fast”, says Antti Rasi, CIO, from Fepod.

“A cloud-based machine learning operations (MLOps) platform provides scalability and flexibility that is needed for all of the varying future needs. Using machine learning is a must in order to be able to provide fast and actionable results for medical professionals from complex blood sample data. It has been great to work with the Fepod team and their cutting-edge solutions bringing point-of-care testing to a new era”, says Mikael Ruohonen from Solita.

Fepod’s technology is based on years of research at Aalto University. Fepod continues developing and commercialising the solution. They will, for example, perform more clinical testing and go through the Medical Device Certification process before being ready for the market entry. Scalable architecture and platform enables expanding the solution to new application areas.

Technologies we used:

• Microsoft Azure
• Databricks
• MLFlow
• Databricks Unity Catalog
• Databricks Workflow
• Databricks Realtime Endpoints
• Delta Lake