REPEATABILITY OF THE CT OSTEOARTHRITIS KNEE SCORE (COAKS) AND A PROTOTYPE CT-GENERATED KELLGREN AND LAWRENCE GRADE

Abstract

INTRODUCTION

The CT Osteoarthritis Knee Score (COAKS) is a semiquantitative system for grading structural disease features of knee OA from weight bearing CT (WBCT). Previous work has demonstrated substantial to near-perfect inter- and intra-observer reliability of COAKS with the aid of a feature scoring atlas, but test-retest repeatability has not yet been evaluated. Given that x-ray and CT rely on the same fundamental physical properties, COAKS could also be harnessed to provide a CT-generated KLG and avoid the need for radiographic imaging.

OBJECTIVE

(1) To evaluate test-retest repeatability of COAKS; (2) to develop a CT-generated KLG (ctKLG) and evaluate its test-retest repeatability; and (3) to compare this prototype ctKLG against radiographic KLG (rKLG).

METHODS

14 individuals recruited and consented at the University of Kansas Medical Center had baseline and follow-up WBCT imaging suitable for analysis. Participant demographics were: mean ± SD age 61.3 ± 8.4 years, BMI 30.7 ± 4.3 kg/m² and male:female ratio 8:6. All scanning was performed on the same XFI WBCT scanner (Planmed Oy, Helsinki, Finland) with the mean ± SD interval between baseline and follow-up attendances 14.9 ± 8.1 days. A Synaflexer^TM device was used to standardize knee positioning during scanning. Imaging acquisition parameters were 96 kV tube voltage, 51.4 mA tube current, 3.5 s exposure time. A standard bone algorithm was applied for reconstruction with 0.3 mm isotropic voxels and a 21 cm vertical scan range. All scans were anonymized prior to analysis both according to the individual and imaging attendance. All knees were reviewed for their COAKS by an experienced musculoskeletal radiologist (T.D.T.). Scores were recorded in a cloud-based file on Google Sheets (alongside the feature atlas in Google Docs) and read by a custom MATLAB script to generate structural heat maps. Test-retest repeatability weighted Kappa (Kw) scores were calculated for each feature (J = JSW; O = osteophytes; C = subchondral cysts; S = subchondral sclerosis) at each compartment (MTF = medial tibiofemoral; LTF = lateral tibiofemoral; PF = patellofemoral; PTF = proximal tibiofibular). A custom MATLAB script applied a decision tree based on recognized KLG verbal definitions to generate ctKLGs for each knee, including a combined score for the MTF and LTF compartments to mimic single-view AP radiographic conditions. A second experienced musculoskeletal radiologist (Z.A.) read study inclusion radiographs for rKLG likewise blinded. Kw was also calculated for ctKLG and rKLG.

RESULTS

Structural heatmaps are shown in Figure 1 for participants with ctKLGs of 1 (study ID 117, right knee) and 4 (study ID 101, right knee) alongside difference maps (follow-up minus baseline). These maps give examples of minimal difference in baseline and follow-up grading at the extremes of structural disease. Best repeatability by feature and compartment was for JSW at the MTF compartment with a Kw (95% CI) of 0.94 (0.93-0.96) and for osteophytes at the PF compartment with a Kw of 0.91 (0.90-0.93). Repeatability for each feature across all compartments was near-perfect (0.82 and above), except being substantial for subchondral sclerosis (0.72, 0.69-0.74). Repeatability for ctKLG was substantial for all individual compartments and near-perfect for the combined tibiofemoral score (0.83,0.80-0.86). Kw was likewise near perfect for rKLG (0.90, 0.88-0.93). Full repeatability Kw (95% CI) results for COAKS, prototype ctKLG and rKLG are given in Tables 1 & 2. It was noted that ctKLG scores were sensitive to small changes in the decision tree based on verbal rKLG interpretation.

CONCLUSION

COAKS repeatability results were similar to those previously demonstrated for intra-observer rating, suggesting that scan factors are consistent enough to have little effect on reader performance. A ctKLG derived from COAKS was similarly repeatable to rKLG with near-perfect performance. A ctKLG model offers a means to stratify structural disease from WBCT without the need for radiographs, however further development is needed to establish a robust decision tree in deriving this from COAKS.