Back to Annual Meeting Program
Ultrasonographic Assessment of Transversal Tendon and Nerve Dynamics in CTS Patients Versus Healthy Controls
Richard van Balen1; Jan-Wiebe H. Korstanje, MSc2; Marjan Scheltens-de Boer, MD3; Joleen H. Blok, PhD3; Harm P. Slijper, PhD2’ Henk J. Stam, MD, PhD1’ Steven E.R. Hovius, MD, PhD4; Ruud W. Selles, PhD2
1Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, Netherlands; 2Department of Rehabilitation Medicine and Physical Therapy & Department of Plastic and Reconstructive Surgery, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, Netherlands; 3Department of Clinical Neurophysiology, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, Netherlands; 4Department of Plastic and Reconstructive Surgery, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, Netherlands
Introduction: Assessment of carpal tunnel syndrome (CTS) using ultrasound is increasingly used, but still suffers from operator dependency leaving room for improvement. The objective of this study was to improve ultrasound assessment of CTS by: 1) normalizing data, 2) using measurement during forceful gripping instead of full extension of the hand, and 3) using hand flexor tendon data alongside median nerve data.
Methods: We included 51 CTS patients and 25 healthy controls in a case-control study. Each patient underwent a full clinical examination and electromyography by an experienced neurologist and filled out the Boston Carpal Tunnel Questionnaire. We made transversal ultrasound recordings of the carpal tunnel during hand movement, starting at full extension and ending with forceful gripping. From the ultrasound images, we extracted the area, perimeter, and circularity of the median nerve of the hand flexor tendons of the index finger and middle finger. Using logistic regression with LASSO variable selection, we evaluated the performance of: 1) the non-normalized area of the median nerve at full extension (a) versus the normalized area of the median nerve at full extension (b), 2) all ultrasound variables at full extension, followed by LASSO selection (c) versus all ultrasound variables at forceful gripping, followed by LASSO selection of variables (d), and 3) model b versus all ultrasound variables at both full extension and forceful gripping followed by LASSO selection (e). The confounding variables BMI, age, and gender were fixed in all models.
Results: Model b outperformed model a (correct classification rate 81% versus 77%) and showed comparable areas under (AUC) the ROC curve (86% versus 87%). Model d outperformed model c on the correct classification rate (83% versus 80%), but had a smaller AUC (82% versus 84%). Model e outperformed model b on the correct classification rate (82% versus 81%), but had a smaller AUC (82% versus 86%).
Conclusion: We have shown that: 1) normalization improved the classification, 2) using forceful gripping instead of full extension marginally improved the classification, and 3) adding the hand flexor tendons improved the classification. In conclusion, the best predictive model was model e, combining normalization of the median nerve and the hand flexor tendon variables during both full extension and forceful gripping. However, the most practical model is model b since this model only uses the normalized area of the median nerve measured in a single position.
Back to Annual Meeting Program