Rofo 2018; 190(S 01): S55
DOI: 10.1055/s-0038-1641403
Vortrag (Wissenschaft)
Muskuloskelettale Radiologie
Georg Thieme Verlag KG Stuttgart · New York

Combining fractal- and entropy-based bone texture analysis for the prediction of Osteoarthritis: data from the Multicenter Osteoarthritis study (MOST)

R Ljuhar
1   Technische Universität Wien, Wien
,
Z Bertalan
2   ImageBiopsy Lab, R&D, Wien
,
D Ljuhar
3   Braincon Technologies, R&D, Wien
,
A Fahrleitner-Pammer
4   Department of Internal Medicine, Division of Endocrinology and Diabetology, Medical University of Graz, Graz
,
S Nehrer
5   Center for Regenerative Medicine & Orthopedics, Danube University, Krems, Krems
,
H Dimai
4   Department of Internal Medicine, Division of Endocrinology and Diabetology, Medical University of Graz, Graz
› Author Affiliations
Further Information

Publication History

Publication Date:
17 April 2018 (online)

Also available at
 

Zielsetzung:

Osteoarthritis (OA) is one of the leading causes of long-term pain and disabilities associated with musculoskeletal disorders. Effective treatment and disease-progression slowdown depend on early detection and quantification of risk. However, current disease parameters, like joint space width (JSW), have proven to be insufficient for the prediction of OA. The purpose of this study was to investigate if combining bone texture analyses with joint space width (JSW) and joint space area (JSA) may improve prediction of OA.

Material und Methoden:

Conventional posterior-anterior (PA) knee radiographs were obtained from the Multicenter Osteoarthritis Study (MOST) database. Oriented fractal- and entropy based texture algorithms were developed, using state-of-the-art machine-learning algorithms. The selected subchondral area used for textural analyses included 4 regions of interest (ROI) in the proximal tibia and one on each condyle of the distal femur. JSW, JSA were assessed using newly developed and fully automated software.

Ergebnisse:

1092 radiographs from one study center were screened for eligibility. Of these, 574 (230 women, 344 men) met the inclusion criteria (Kellgren & Lawrence (KL) score of 0 at baseline). At month 84, 41 female and 79 male patients had developed KL≥1, and 189 female and 265 male patients remained at KL0. AUC for incident OA using JSW/JSA/clinical features was 0.67 ± 0.08 for women, and 0.61 ± 0.1 for men. In contrast, combining fractal/entropy-based texture, JSW/A and clinical features resulted in significantly improved AUC (0.80 ± 0.07 for women and 0.69 ± 0.1 for men, respectively).

Schlussfolgerungen:

This study provides strong evidence, that a combination of fractal- and entropy-based textural analyses of plain subchondral bone radiographs together with JSW/A and clinical features is superior to JSW/A and clinical features alone in predicting incident OA in men and women.