Norwegian version of this page

Public defense: Predicting Anterior Cruciate Ligament Reconstruction Outcome

A recent study used machine learning to develop personalized risk calculators for ACL surgery patients, predicting the chance of needing a revision surgery within five years. By utilizing data from Norway and Denmark, the study aims to tailor outcomes to individual risk profiles. While promising, the models highlight both potential and areas for improvement.

Portrett Kyle Martin
  • ​​​​​​Kyle Martin is a phD candidate at Oslo Sports Trauma Research Center.
  • On June 4, he will defend his dissertation: "Predicting Anterior Cruciate Ligament Reconstruction Outcome: Machine Learning Analysis of National Knee Ligament Registries".
  • Read the dissertation (nva.sikt.no)
  • Learn more about the project.
  • Watch the defence live on NIHs YouTube-channel.

About the study

When answering these questions, physicians often generalize and make estimates using their experience and knowledge of previous research studies. However, for ACL reconstruction surgery, one size does not fit all and patients may have very different risk profiles with regards to eventual outcome.

This thesis sought to overcome this limitation by using machine learning – a subset of artificial intelligence – to develop risk calculators that can estimate the risk of certain outcomes after ACL reconstruction on a patient-specific level. The primary outcome of interest was the likelihood of a patient requiring a revision surgery (a re-do operation for a failed reconstruction) within 5 years. The Norwegian and Danish national knee ligament registries, which collect data on every ACL reconstruction performed in those respective countries, served as the information source used to train the machine learning algorithms.

Results

This thesis consisted of 6 studies which developed and tested the validity of several prediction models and led to the creation of the first machine-learning-based online calculators which clinicians could use in the office setting to estimate outcomes for their patients. The accuracy of these models was moderate overall – better than random chance but also with room for improvement. Taken together, these studies demonstrate the potential of novel AI-based tools for clinical outcome prediction, while also highlighting the associated challenges and limitations.

Committee

Professor Jorunn Sundgot-Borgen, Norwegian School of Sport Sciences (NIH), chairperson

Professor Kristoffer Weisskirchner Barfod, University of Copenhagen, Denmark

Assistant Professor Jacobien Oosterhoff, University Medical Centre Groningen, Department of Orthopaedic Surgery, Groningen, the Netherlands

Supervisors

Main Supervisor: Gilbert Moatshe MD PhD, NIH, UiO, OSTRC

Co-supervisors: Roald Bahr MD PhD, NIH, OSTRC 

Lars Engebretsen MD PhD, NIH, OSTRC, UiO 

The defense will be chaired by:

Professor Truls Raastad

Program

  • 10.15-11.00: Trial lecture
  • 13.00-16.00: Dissertation lecture followed by discussion

Practical Information

The defense is open to everyone and will be streamed on NIH's YouTube channel.

Organiser

Solveig Sunde
Tags: Institutt for idrettsmedisinskfag
Published May 20, 2025 11:30 AM - Last modified Mar. 24, 2026 2:24 PM