Parcourir les annonces Parcourir les étudiants Publier une annonce Ajouter un CV

PHILIPS HEALTHCARE – Internship 2 : Landmark detection in fetal ultrasound, with quality estimation.


Suresnes (92)

Date de début
Date de fin

Philips Healthcare is a world leader in medical imaging. Its products cover the full range of imaging modalities: X-Rays, MRI, Ultrasound, CT, etc. The company is internationally recognized for the excellence of its technology, developed within innovative research groups.

Philips Healthcare Medisys Research Lab is based in Suresnes (92) and is dedicated to medical image processing. The team, with about thirty researchers and engineers, is focused on delivering the most innovative solutions in the domain and is in close contact with famous universities and clinical sites in France and abroad.

Internship description

Ultrasound (US) imaging is the modality of choice for fetal screening. In many countries, the US exam performed between 18 and 22 weeks of pregnancy is used to assess the development of the fetus by performing some biometry measurements. One of those measurement is the femur length, corresponding to the distance between the femur end-points. In this internship, we aim at designing an algorithm for automatic detection of those landmarks.

For such an application, deep learning techniques have reached state-of-the-art performance. However, they usually fail to produce a reliable estimation of the quality of their prediction, for instance on out-of-distribution data. This would be a precious feature during the biometry exam in view of future integration into an ultrasound system.

This internship will be divided into several moments. A first step will be to implement a baseline model for landmark detection. Then, we will review the existing methods for uncertainty/quality estimation. The retained method will be implemented and evaluated, especially on non-standard femur acquisitions. A final development may be to transpose and adapt the pipeline to a more complex problem, for instance for 3-dimensional data, corresponding to a 3D ultrasound acquisition.

Candidate profile

  • Third year of engineer school / Master 2 Recherche, with specialty in machine learning, image processing or applied mathematics;
  • Solid knowledge of statistics, machine learning, deep learning, image processing;
  • Experience Python; knowledge of the Tensorflow/Keras framework;
  • English speaking, reading and writing is mandatory;
  • Good communication skills and ability to work in a team.

Vous devrez avoir ces compétences :