Machine learning is better than surgeons at assessing unicompartmental knee replacement radiographs

Journal article

Background: Poor results occasionally occur after unicompartmental 

Methods: 924 one-year anterior-posterior radiographs post-UKR were used to train a machine learning model (ResNet50v2) with a 

Results: The ResNet50v2 model correctly identified 71% (n = 10) of the patients with a poor score and 46 (82%) of those with an excellent score. In contrast, one surgeon could not identify patients with Poor scores (0%) and the other identified one (7%). Both misidentified 3 of those with Excellent scores. The model visualisation method suggested that estimated classifications were made from image features around the implants.

Conclusion: The results suggest that there are radiographical features that relate to poor outcomes, which the surgeons are unaware of. Those the model did not identify may have an extra-articular cause for their poor outcome. Further analysis to identify the features associated with poor outcomes could potentially suggest ways that indications or techniques could be improved so as to decrease the incidence of poor results.

Authors: Tu, SJ; Kendrick, S; Saravanan, K; Dodd, C; Murray, DW

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Elsevier
• Journal: Knee
• Volume: 52
• Pages: 212-219
• Publication date: 2024-11-30
• Acceptance date: 2024-11-08
• DOI: 10.1016/j.knee.2024.11.007
• EISSN: 1873-5800
• ISSN: 0968-0160
• Pmid: 39615060
• Language: English
• Keywords: convolutional neural network (CNN); transfer learning; radiograph; clinical outcomes