S. Nehrer (Krems, AT)

Donau University Krems Center for Regenerative Medicine

Presenter Of 2 Presentations

Extended Abstract (for invited Faculty only) Joint Specific Cartilage Repair

13.3.6 - Case Presentation 6

Presentation Number
13.3.6
Presentation Topic
Joint Specific Cartilage Repair
Lecture Time
08:20 - 08:30
Session Type
Instructional Course / Workshop
Corresponding Author
Podium Presentation Osteoarthritis

16.3.4 - Machine learning predicts rate of cartilage loss: data from the Osteoarthritis Initiative

Presentation Number
16.3.4
Presentation Topic
Osteoarthritis
Lecture Time
11:42 - 11:51
Session Type
Free Paper Session
Corresponding Author
Disclosure
Tiago Paixao, Imagebiopsy Lab, employee Christoph Goetz, Imagebiopsy Lab, employee Zsolt Bertalan, Imagebiopsy Lab, employee Richard Ljuhar, Imagebiopsy Lab, Shareholder

Abstract

Purpose

The rate of cartilage loss can vary widely between patients at risk or suffering from knee osteoarthritis (OA) but its causes remain unknown. We investigate whether quantitative and semi-quantitative radiographic features can be used to predict the rate of Joint Space Width (JSW) loss.

Methods and Materials

We collected bilateral knee radiographs, obtained by the OAI study, from 4100 patients (2383 female, 1717 male). Each patient was imaged up to 7 times, separated by at least 12 months, across a time span of 8 years. Each radiograph was analyzed by software to obtain Kellgren-Lawrence (KL) and OARSI grades for Osteophytes, Sclerosis and Joint Space Narrowing (JSN) readings, as well as JSW measurements for each individual knee. Individual knees with rate of JSW loss above 0.072 mm/year (the average yearly loss within JSN grade) were classified as progressors (956 knees). From these, knees in the top 10% of JSW loss rate were classified as fast progressors (91 knees). A logistic regression model was used to predict the fast progressor phenotype with KL and OARSI grades at baseline as independent variables. Model performance was estimated using 10-fold cross-validation dataset splits and used Area Under the Curve (AUC) as performance criteria.

Results

The logistic regression classifiers achieved AUCs of 0.71 (SE 0.015) and 0.66 (SE 0.013) at classifying individual knees as fast progressors for medial and lateral compartments, respectively. Analysis of the individual coefficients of the classifiers reveals that JSN and Sclerosis grades are the main predictors of rapid cartilage loss.

panel.jpg

Conclusion

Our results show that it is possible to predict future rapid cartilage loss from readings from a single plain radiograph. Interestingly, neither KL grade nor Osteophytes OARSI grade contributed greatly to this prediction. Instead, Sclerosis and JSN grade seem to be the major predictors of rapid cartilage loss, suggesting a non-canonical mode of OA progression.

Collapse

Moderator Of 2 Sessions

Turner - ICRS Board Room (15) ICRS Committee Meeting

Communication Committee

Regency Free Paper Session
Session Type
Free Paper Session
Date
06.10.2019
Time
13:30 - 15:00
Location
Regency

Meeting Participant of

Turner - ICRS Board Room (15) ICRS Committee Meeting

Communication Committee