Premium
Carpal tunnel syndrome diagnosis by a self‐normalization process and ultrasound compound imaging
Author(s) -
Liao YinYin,
Wu ChinChou,
Kuo TaiTzung,
Chen JiannPerng,
Hsu YenWei,
Yeh ChihKuang
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4767754
Subject(s) - ultrasound , wrist , carpal tunnel syndrome , medicine , supine position , median nerve , carpal tunnel , ultrasound imaging , speckle pattern , nuclear medicine , radiology , artificial intelligence , anatomy , computer science , surgery
Purpose: Carpal tunnel syndrome (CTS) is the common entrapment neuropathy that occurs due to compression of the median nerve at the wrist. Ultrasound images have been used to highlight anatomical variants of the median nerve, and CTS is thought to be associated to enlargement of the cross‐sectional area (CSA) of the median nerve. However, there remains controversy regarding the most appropriate cutoff values of the computer measurements including the CSA, flattening ratio, and palmar bowing of median nerve, especially given that they can be influenced by image artifacts and factors that differ between individual patients. This study proposed a modified ultrasound compound imaging technique by moving fingers to reduce image artifacts, and the estimates of the normalized CSA [i.e., CSA at the wrist (CSAw) to CSA at the midforearm] with the aim of reducing discrepancies in CSA estimates and improving the ability of CTS discrimination. Methods: The subjects were examined with their arms supine and while they were making repetitive movements of their fingers (from an open palm into a clenched fist) within 3 s. By a commercial ultrasound scanner with a 10‐MHz linear array transducer, a total of 70 images were acquired in each subject. The frame rate of ultrasound system was 25 fps. Nine frames in the acquisition sequence that had produced partial speckle decorrelation were incoherently added to form a compound image, and the inplane motion of them was corrected using the multilevel block‐sum pyramid algorithm. The manual contours outlined by ten experimenters and three physicians were used to test the performance in determining the boundary of the median nerve. The receiver operating characteristic (ROC) curve was used to evaluate the usefulness of the estimates in distinguishing healthy volunteers from CTS patients. Results: The manual contours of the median nerve in the compound images had an average area overlap exceeding 90% and relatively small area errors. The areas under the ROC curve obtained using the CSAw estimates for the original and compound images were 0.60 ± 0.09 (mean ± standard error) and 0.80 ± 0.05, respectively; that using normalized CSA estimates for the original and compound images were 0.76 ± 0.04 and 0.89 ± 0.04, respectively. The results show that variations in the CSAw values of compound images for healthy overweight and obese subjects can adversely influence CTS diagnosis, but that this can be overcome using the normalized CSA estimate of compound images. Conclusions: Compound imaging provides images of superior quality for determining the location of the median nerve boundary. Using the normalized CSA estimate would assist in eliminating problems associated with variability between populations, since the subject becomes his or her own internal control, thereby improving the ultrasound‐based diagnosis of CTS.