Summary

This study presents a novel method for the assessment of joint space width and subchondral bone micro architecture for the diagnosis of knee osteoarthritis.

Abstract

Objective

Assessment of osteoarthritis (OA) of the knee usually involves AP and lateral radiographs to evaluate medial and lateral joint spaces, but perspective errors and low reproducibility are limiting factors. In addition to joint space width, subchondral bone area may provide important information on the status of OA. However, no adequate standard has been developed so far to quantify subchondral changes. The method described here combines assessment of joint space width (JSW) and texture analysis of the adjacent subchondral bone micro architecture to discriminate between patients with and without OA.

Methods

The study included 274 standardized knee radiographs from 110 patients with OA, and 164 controls. Knee joint space analysis was performed at the medial and lateral compartment, applying an entropy based algorithm for automated detection of critical landmarks and joint space contour. Furthermore, subchondral bone texture was assessed by using fractal analysis at predefined regions of the proximal tibia. A matrix of 3x8 ROIs was used to gain sufficient textural information (FIG.). Self-similarity of the texture, reflecting 2D projection of the 3D trabecular structure, has been used to calculate the Bone Structure Value (BSV) which provides indirect information on bone quality.

Results

Comparing mean BSVs of the control and the OA group of selected 89 female patients, a statistical significant deviation of 7.04% (p<0.01) in mean values was determined with an odds-ratio of 2.89 (95% CI, 1.2-6.89). A combination of JSW & BSV showed a further increase in discriminative power between the controls and OA patients. Differences in BSV were found between left/right knee and male/female. Furthermore, a rising BMI was identified to be linked to lower BSV values.

Conclusions

The novel method described here is sufficient to discriminate between subjects with and without OA. Furthermore, fractal analysis alone may provide information on bone quality aspects. Future work should therefore focus on the potential role of bone micro architecture algorithms to serve as a fracture risk assessment tool. Such algorithms could serve as early disease predictor of not just OA but also additional degenerative bone diseases like osteoporosis.