Summary
Objective
To create a scalable and feasible retrospective consecutive knee osteoarthritis (OA) radiographic database with limited human labor using commercial and custom-built artificial intelligence (AI) tools.
Methods
We applied four AI tools, two commercially available and two custom-built tools, to analyze 6 years of clinical consecutive knee radiographs from patients aged 35–79 at the University of Copenhagen Hospital, Bispebjerg-Frederiksberg Hospital, Denmark. The tools provided Kellgren-Lawrence (KL) grades, joint space widths, patella osteophyte detection, radiographic view detection, knee joint implant detection, and radiographic marker detection.
Commercial tools
The commercial knee OA tool (COAT), RBknee v2.1, Radiobotics ApS, Copenhagen, Denmark, analyzed all knee radiographs. The knee OA tool has previously been externally validated for knee OA grading with the Kellgren-Lawrence (KL) grade scale on Danish clinical data, and the performance was in strong agreement7 with radiologic senior MSK consultants experienced in performing the KL grade.5 In addition, the knee OA tool’s performance on the presence of patella osteophytes on lateral view has been internally validated by the company that developed the tool and had a receiver operating characteristic (ROC) area under the curve (AUC) of 0.91, a sensitivity of 78%, and a specificity of 87% using a combination of international and local Danish data (Radiobotics. Instruction for Use Specification MDR – RBknee2, RBknee v.2.1 (MDR), 3rd revision. 2021). The knee OA tool is CE-marked for KL grading, and we used the FDA-cleared 1.0.1 version for minimum joint space width (JSW) measurements in mm inspired by Neumann et al.8 and Duryea et al.9 with internal validation by the company: orthogonal regression slope of 1.00 for both medial and lateral compartments, orthogonal regression intercept of −0.07 mm for medial and −0.10 mm for lateral compartments, and mean absolute error of 0.28 mm for medial and 0.39 mm for lateral compartments (Radiobotics. Instructions for Use – RBknee – U.S. Only, Revision: 06, Software v. RBknee v.1.0. 2021). COAT did not detect markers on the radiographs, and therefore, left and right knee findings could be transposed for patients with an unconventional bilateral frontal radiographic with the left knee on the left side and the right knee on the right side. So, we used an additional tool, MDT, from Radiobotics ApS, RBmarker v0.1.0, to detect digital and lead markers on knee radiographs to ensure a conventional position of the bilateral frontal image (right knee on the left side and left knee on the right side) with correct marking, and we followed up with a review of the flagged radiographs to confirm the findings. Overlayed markers were burned into the radiographs before analysis using Python v3.9.16, numpy v1.22.3, and pydicom v2.3.1. Radiographs with missing markers and inconsistent markings (left marker on the left side or right marker on the right side) were excluded. The tools were used off-label.
Results
In total, 25,778 knee radiographs from 8575 patients were included in the database after excluding inapplicable radiographs, and 92.5% of the knees had a complete OA dataset. Using the four AI tools, we saved about 800 hours of radiologist reading time and only manually reviewed 16.0% of the images in the database.
Conclusions
This study shows that clinical knee OA databases can be built using AI with limited human reading time for uniform grading and measurements. The concept is scalable temporally and across geographic regions and could help diversify further OA research by efficiently including radiographic knee OA data from different populations globally. We can prevent data dredging and overfitting OA theories on existing trite cohorts by including various gene pools and continuous expansion of new clinical cohorts. Furthermore, the suggested tools and applied approaches provide an ability to retest previous hypotheses and test new hypotheses on real-life clinical data with current disease prevalence and trends.
Open access
This research is open access.
Osteoarthritis and Cartilage is the official journal of the Osteoarthritis Research Society International, an international, multidisciplinary journal for the many kinds of specialists and practitioners concerned with osteoarthritis.
