Multivariate use of MRI biomarkers to classify histologically confirmed necrosis in symptomatic total hip arthroplasty.

The failure of total hip arthroplasty (THA) is commonly associated with the necrosis of the periprosthetic tissue. To date, there is no established method to noninvasively quantify the progression of such necrosis. Magnetic resonance imaging (MRI) of soft tissues near implants has undergone a recent renaissance due to the development of multispectral metal-artifact reduction techniques. Advanced analysis of multispectral MRI has been shown capable of detecting small magnetism effects of metallic debris in periprosthetic tissue. The purpose of this study is to demonstrate the diagnostic utility of these MRI-based tissue-magnetism signatures. Together with morphological MRI metrics, such as synovial volume and thickness, these measurements are utilized as biomarkers to noninvasively detect soft-tissue necrosis in symptomatic THA patients ( N = 78 ). All subjects underwent an advanced MRI scan before revision surgery and tissue biopsies utilized for necrosis grading. Statistical analyses demonstrated a weak, but significant positive correlation (P = .04) between MRI magnetism signatures and necrosis scores, while indicating no meaningful association between the latter and serum cobalt and chromium ion levels. Receiver-operating characteristic (ROC) analyses were then performed based on uni- and multivariate logistic regression models utilizing the measured MRI biomarkers as predictors of severe necrosis. The area under the curve of the ROC plots for MRI biomarkers as combined predictors were found to be 0.70 and 0.84 for cross-validation and precision-recall tests, respectively.