Associations between the meniscal spatial-specific signal indexes of T2-weighted images and the presence of radiographic knee osteoarthritis

Abstract

Rationale and objective

Volume and signal intensity of meniscus in T2-weighted images are typical manifestations of meniscal injuries, which are risk factors for the presence of radiographic knee osteoarthritis (RKOA). The objective of this study was to predict the presence of RKOA by proposing six meniscal spatial-specific signal (MSS) indexes.

Materials and methods

Ninety subjects with symptomatic KOA were divided into non-RKOA and RKOA groups with a cut-off of Kellgren and Lawrence grade $\ge$2. Lateral and medial menisci were automatically segmented, and meniscus pixels were categorized into three classes based on signal intensity to yield six MSS indexes, e.g., M-MSS-Ci denoting the MSS indexes with the ith class for the medial meniscus. Reproducibility of the MSS indexes, and their relationships with knee cartilage volume and WOMAC were evaluated by linear regression analysis. Among the six MSS indexes and their combinations, the best index for predicting the presence of RKOA was identified using the area under the curve (AUC).

Results

The MSS indexes by the automatic segmentation demonstrated high reproducibility with the manual segmentation. The medial MSS indexes were different across the two groups ($P<0.05$), and the RKOA group showed lower M-MSS-C1, higher M-MSS-C2 and M-MSS-C3. In addition, M-MSS-C1 and M-MSS-C3 were correlated with knee cartilage volume, and M-MSS-C1 and M-MSS-C2 were correlated with WOMAC ($P<0.05$). The M-MSS (combination of three medial MSS indexes) outperformed all other indexes with an AUC of 0.84 for predicting the presence of RKOA.

Conclusions

The medial MSS indexes are more associated with the presence of RKOA, and the M-MSS, as a combination of three medial MSS indexes, demonstrates its effectiveness in predicting the presence of RKOA.