Comparative intra- and inter-observer reliability of two methods for evaluating intraoperative ultrasonography-based spinal cord hyperechogenicity intensity in degenerative cervical myelopathy

Abstract Objectives During French-door laminoplasty, a linear array transducer of IOUS was used to observe and record the spinal cord decompression. To acquire a higher-reliability method, and compare the in-observer and inter-observer reliability of two methods in evaluating the hyperechoic intensity of spinal cord ultrasound in degenerative cervical myelopathy (DCM). Background The intensity of spinal cord hyperechogenicity is considered as a potential predictor of neurological recovery in DCM after decompression, but the accuracy of gray value ratio (GVR) is affected by many factors. Methods Totally 28 patients (20 males and 8 females) who had been followed up for 12 months were included. Their mean age at surgery was 61.2 ± 10.8 years and the average symptom duration was 23.36 ± 22.11 months. The gray values of circles 1, 2 and 3 were recorded as Gcompression, Gnorml and Gsac, respectively. Circle 1 was drawn with the maximum brightness point within the spinal cord as the center, circle 2 with the same area was plotted on the spinal cord with uniform echogenicity, without compression and at least 1 cm away from the circle 1, and circle 3 was drawn on the dorsal dural sac at the same segment as circle 1. GVR was calculated as follows: GVR-A = Gcompression/Gnorml (method A), and GVR-B = Gcompression/Gsac (method B). The in-observer and inter-observer reliabilities of the two methods were compared. It is generally believed a reliability coefficient 0.75 indicate poor and good reliability respectively. The images-based GVR-B using this protocol demonstrates higher inter- and intraobserver reliabilities than GVR-A, and can be used as the basis for prognostic prediction and future studies. Results All examination acquisitions were successfully completed. GVR-A averaged 2.043 (0.318–5.56), and GVR-B averaged 0.578(0.06–1.41). GVR-B has better repeatability of gray value measurement, smaller relative standard deviation (RSD%) (0.298 vs. 0.32) and larger inter-group correlation coefficient compared with GVR-A. The mean value (MD) of the GVR difference calculated by GVR-B between the two clinicians was closer to 0. Conclusions For DCM patients routinely using ultrasound for real-time cord visualization during spinal cord decompression by French-door laminoplasty, the images-based GVR-B using this protocol demonstrates better inter- and intraobserver reliabilities compared with GVR-A.