CONTEXT

Sector: Healthcare

Business Expertise: Forecasting

OBJECTIVES
  • Infer core body temperature from data collected by external sensors
  • Design a high-performance prediction model that runs on a low-powered Android platform
SOLUTION
  • Building a relevant and robust database
  • Implementation and continuous improvement of multiple prediction models
RESULTS
  • Predictive temperature model with a mean absolute accuracy of 0.2°C
  • Integration into the Thermodiag medical device, currently undergoing CE marking, is  highly praised by many healthcare professionals
When you invent a product, you want to be involved in everything. When you entrust part of your project to a service provider, you may feel like the project is slipping away from you. EURODECISION is really committed to keeping the business expert at the center of the project. For us, this was a very important point, but also very reassuring.
Benjamin MENARDCTO - F2D Medical
EURODECISION was present throughout the entire project, from R&D to industrialization. Whether from a professional standpoint (meeting deadlines and specifications) or a personal one (interactions with experts), the relationships have always been pleasant. On Trustpilot, I recommend working with EURODECISION!
Benjamin MENARDCTO - F2D Medical

The start-up F2D Medical has set itself the goal of advancing predictive and personalized medicine. Building on the unanimous observation among healthcare professionals that current core temperature measurement is inaccurate, they considered developing a non-invasive device for continuous body temperature monitoring. This is how the connected Thermodiag armband, now patented, came to be after several years of R&D.

In addition to the hardware component that continuously collects data from temperature sensors, the diagnostic aid tool required developing predictive models to infer core temperature accurately. Seeking expertise in mathematical modeling, F2D Medical engaged EURODECISION’s machine learning experts to support them from the R&D phase.

The F2D Medical and EURODECISION teams worked in close cooperation to produce, on a very small data sample, a POC whose very encouraging results confirmed the value of the chosen algorithmic method. This method indeed proved to be the most suitable technique for achieving high performance on an Android terminal such as a smartphone. EURODECISION then built the biostatistical rationale required for cohort studies, whose subjects had to have representative profiles (sex, age group…). These clinical trials with more than 300 subjects having confirmed the validity of the prediction models, F2D Medical completed an initial funding round in 2021 and embarked on product industrialization.

During the following phases, EURODECISION worked to adjust and refine the prediction models and make them even more robust. The artificial intelligence experts also collaborated with the integration teams responsible for porting to Android to ensure the prediction models worked as expected.

In its current version, which is undergoing CE marking by F2D Medical, Thermodiag integrates several prediction models that are triggered sequentially for continuous temperature monitoring, achieving a mean absolute accuracy of 0.2°C.

While awaiting this certification, the Thermodiag device has been presented in several hospitals where it was particularly well received by healthcare professionals. Initially designed for monitoring fragile patients (e.g., during chemotherapy or in the post-operative phase), discussions with doctors from various specialties are opening up additional applications.

F2D Medical indeed intends to continue improving its medical device. Whether it involves assessing the relevance of using synthetic data or considering hyper-specialization of predictive models by pathology, the start-up maintains a close relationship with EURODECISION. By keeping mathematical modeling experts informed of project progress, F2D Medical ensures they are ready to step in as soon as their expertise is needed again.

 

 

To learn more about the project: f2d-medical.io