Automated Detection and Grading of Hip Osteoartritis

Automated Detection and Grading of Hip Osteoartritis

This is an AI for Health MSc project. Students are eligible to receive a monthly reimbursement of €500,- for a period of six months. For more information please read the requirements.

Clinical problem

Osteartritis (OA) of the hip joint is a common condition which can lead to severe health problems. According to the RIVM, around 1.5 million people in the Netherlands suffer from OA (500 million worldwide), which is diagnosed using plain radiographs. On the radiographs, radiologists recognize OA by noticing joint space narrowing, extra bone formation (called ‘osteofytes’) and increased bone density. In a clinical setting, the number of radiographs that need OA screening is very high, and the interobserver reliability in grading the severity of OA is low to moderate.

Solution

In this project we plan to develop a deep learning algorithm for automated detection of joint space narrowing, extra bone formation and increased bone density. Besides automated detection, the algorithm must also be trained to grade hip OA severity. The algorithm will be externally validated in an orthopedic surgery department. We will build further on a previous deep learning algorithms developed for this purpose but which were developed with only smaller subsets of radiographs and with older deep learning algorithms.

Data

The World COACH consortium, a worldwide collaboration of 8 prospective cohorts on hip osteoarthritis include 40,555 participants (aged 35 to 80 years at baseline), of which 34,018 participants have annotated baseline pelvic radiographs available. This consortium is hosted by Erasmus MC and this project will therefore be carried out in close collaboration with Erasmus MC.

Results

When the algorithm has sufficient sensitivity and specificity it will be evaluated in practice.

Embedding

The student will be supervised by a team with expertise in deep learning and musculoskeletal radiology and orthopedics at Radboudumc. For model development, we provide access to our deep learning GPU cluster Sol.

Requirements

  • Students Artificial Intelligence, Data Science, Computer Science, Bioinformatics, Biomedical Engineering or similar in the final stages of their Master education.
  • You should be proficient in Python programming and have a theoretical understanding of deep learning architectures.
  • Experience with medical images is beneficial.

Information

  • Project duration: 6 months
  • Location: Radboud University Medical Center
  • For more information or to apply for this project, please contact Bram van Ginneken.

People

Walter van der Weegen

Walter van der Weegen

 Rintje Agricola

Rintje Agricola

Bram van Ginneken

Bram van Ginneken

Professor

Diagnostic Image Analysis Group

 Matthieu Rutte

Matthieu Rutte

Silvan Quax

Silvan Quax

Research Scientist RTC Deep Learning

RTC Deep Learning