Reliable AI for better surgical decisions

The challenge

Personalised medical care offers enormous potential for better patient treatment. Our customer set out to use AI to provide reliable individual surgery recommendations that lead to better surgical outcomes and lower risks for patients. The challenge was to enable the use of neural networks for surgery recommendations in line with the high quality standards of the Medical Device Regulation (MDR).

Our customer

A leading manufacturer of medical devices with strong innovative capability and high standards for the quality of its own products.

How we helped

Together with the customer, XITASO developed and implemented a development and quality assurance process for the use of neural networks in surgery recommendations, enabling certification of the product under MDR. Multi-stage analysis and testing procedures were applied at both the data and model level.

The training processes of the neural networks were optimised for deterministic prediction performance in order to achieve reliable results. By using a suitable training architecture together with a comprehensive continuous delivery pipeline, full traceability was ensured throughout the ongoing development of the system.

On this basis, the documentation required for successful certification of the product could also be generated.

Technologies

Python C# Neural Networks PyTorch scipy numpy Azure Functions Azure Blob Storage Azure LogAnalytics MongoDB Terraform SBOM