AI-based recognition of facial features and symmetry axes to build biometric databases for the development of respiratory masks

Context

In non-invasive ventilation and sleep therapy, masks are used that rest on the patient's face and enclose either the nose or the nose and mouth. A good mask fit is a key success factor for these therapies. The therapy pressure can only be applied effectively with a tight mask. Leaks or excessive pressure exerted by the mask to reduce leaks significantly impair wearing comfort and therefore acceptance of the therapy.

Biometric information on facial shapes is essential for developing well-fitting masks for people with different face shapes, body sizes and medical conditions. For this purpose, facial features are marked in 3D scans of faces, symmetry axes and planes are determined, distances and angles are measured and stored in databases. This process is time-consuming and should be accelerated by developing suitable automation tools

Tasks

As part of the final thesis, existing methods for 'facial landmark detection' are to be compared, examined for suitability in the present context and extended so that all features relevant to the given application can be extracted from 3D face scans.
The algorithms are to be integrated into an application in such a way that large volumes of facial scans can be processed by people without specialist knowledge in the field of machine learning. The quality of the results should be able to be continuously improved through the future availability of a larger volume of training data.

Requirements

• Knowledge of basic terms and methods from the fields of AI, machine learning, neural networks
• Experience with programming environments for the development of AI algorithms (Python, Matlab)