KERANOVA Service Center - Medical device IA

Project

Development and integration of an artificial intelligence model for the detection and segmentation of medical images within an ophthalmic surgical device (e.g. cataract), to assist surgeons in making incisions.

#Python #PyTorch #ONNX #C++ #Eclipse #Qt #UML #Git #Linux (Debian)

Challenges & objectives

Development and continuous improvement of AI models (single- and multi-class detection/segmentation)

Integration into an embedded medical device, in a real-time environment

Compliance with memory constraints and medical tolerances (6 µm)

Reduce processing time and improve model performance

UML documentation, model evaluation and post-processing

Unit testing and compliance with medical standards

Requirements

Compliance with medical standards

Real-time constrained coding and embedded resources

Process automation

Technical documentation (UML, comments, validation)

Agile collaboration with regular reporting

Processing time

Detection rate ≥ 85

Results

Accuracy: 97.5% (anterior corneal interface) and 97% (posterior)

Treatment: 500 ms/patient, tolerance 6 µm

Integration into medical device software

More ergonomic and reliable surgical guidance of the cornea

Our achievements

Each project reflects our ability to transform a critical requirement into a concrete solution. From the embedded software layer to functional safety, Médiane Système mobilizes all its expertise to deliver innovative, robust solutions ready for industrialization.