Logistic Regression and Elastic Net Regularization for the Diagnosis of Fibromyalgia (LEDF)

Fibromyalgia (FM) diagnosis remains a challenge for clinicians due to a lack of objective diagnostic tools. One proposed solution is the use of quantitative ultrasound (US) techniques, such as image texture analysis, which has demonstrated discriminatory capabilities with other chronic pain conditions. The investigators propose the use of US image texture variables to construct an elastic net regularized, logistic regression model, for differentiating between the trapezius muscle in the healthy and FM patients. 162 Ultrasound videos of the right and left trapezius muscle were acquired from healthy participants and participants with FM. The videos will then be put through a mutli-step processing pipe including converting them into skeletal muscle regions of interest (ROI). The ROI's will be then filtered by an algorithm utilizing the complex wavelet structural similarity index (CW-SSIM), which removes ROI's that are too similar to one another. Eighty-eight texture variables will be extracted from the ROI's, which will be used in nested cross-validation to construct a logistic regression model with and without elastic net regularization. The generalized performance accuracy of both models will be estimated and confirmed with a final validation on a holdout test set. Depending on the predicted, generalized performance accuracy it will be validated or not by the final, holdout test set (confirming the model construction is accurate). These models should then confirm or deny the hypothesis that a regularized logistic regression model built on ultrasound texture features can accurately differentiate between healthy trapezius muscle and that of patients with FM.