The standing radiograph is used as a gold standard to diagnose spinal deformity including scoliosis, a medical condition defined as lateral spine curvature > 10°. However, the health concern of X-ray and large inter-observer variation of measurements on X-ray images have significantly restricted its application, particularly for scoliosis screening and close follow-up for adolescent patients. In this study, a radiation-free freehand 3-D ultrasound system was developed for scoliosis assessment using a volume projection imaging method. Based on the obtained coronal view images, two measurement methods were proposed using transverse process and spinous profile as landmarks, respectively. As a reliability study, 36 subjects (age: 30.1 ±14.5; male: 12; female: 24) with different degrees of scoliosis were scanned using the system to test the inter- and intra-observer repeatability. The intra- and inter-observer tests indicated that the new assessment methods were repeatable, with ICC larger than 0.92. Small intra- and inter-observer variations of measuring spine curvature were observed for the two measurement methods (intra-: 1.4 ±1.0° and 1.4 ±1.1°; inter-: 2.2 ±1.6° and 2.5 ±1.6°). The results also showed that the spinal curvature obtained by the new method had good linear correlations with X-ray Cobb's method (R2 = 0.8, p < 0.001, 29 subjects). These results suggested that the ultrasound volume projection imaging method can be a promising approach for the assessment of scoliosis, and further research should be followed up to demonstrate its potential clinical applications for mass screening and curve progression and treatment outcome monitoring of scoliosis patients.

SummaryBackground/ObjectiveStanding radiograph with Cobb's method is routinely used to diagnose scoliosis, a medical condition defined as a lateral spine curvature > 10° with concordant vertebral rotation. However, radiation hazard and two-dimensional (2-D) viewing of 3-D anatomy restrict the application of radiograph in scoliosis examination.MethodsIn this study, a freehand 3-D ultrasound system was developed for the radiation-free assessment of scoliosis. Bony landmarks of the spine were manually extracted from a series of ultrasound images with their spatial information recorded to form a 3-D spine model for measuring its curvature. To validate its feasibility, in vivo measurements were conducted in 28 volunteers (age: 28.0 ± 13.0 years, 9 males and 19 females). A significant linear correlation (R2 = 0.86; p < 0.001) was found between the spine curvatures as measured by Cobb's method and the 3-D ultrasound imaging with transverse process and superior articular process as landmarks. The intra- and interobserver tests indicated that the proposed method is repeatable.ResultsThe 3-D ultrasound method using bony landmarks tended to underestimate the deformity, and a proper scaling is required. Nevertheless, this study demonstrated the feasibility of the freehand 3-D ultrasound system to assess scoliosis in the standing posture with the proposed methods and 3-D spine profile.ConclusionFurther studies are required to understand the variations that exist between the ultrasound and radiograph results with a larger number of volunteers, and to demonstrate its potential clinical applications for monitoring of scoliosis patients. Through further clinical trials and development, the reported 3-D ultrasound imaging system can potentially be used for scoliosis mass screening and frequent monitoring of progress and treatment outcome because of its radiation-free and easy accessibility feature.

This paper presents an automated measurement of spine curvature by using prior knowledge on vertebral anatomical structures in ultrasound volume projection imaging (VPI). This method can be used in scoliosis assessment with free-hand 3-D ultrasound imaging. It is based on the extraction of bony features from VPI images using a newly proposed two-fold thresholding strategy, with information of the symmetric and asymmetric measures obtained from phase congruency. The spinous column profile is detected from the segmented bony regions, and it is further used to extract a curve representing spine profile. The spine curvature is then automatically calculated according to the inflection points along the curve. The algorithm was evaluated on volunteers with the different severity of scoliosis. The results obtained using the newly developed method had a good linear correlation with those by the manual method (r ≥ 0.90, p <; 0.001) and X-ray Cobb's method (r = 0.83, p <; 0.001). The bigger variations observed in the manual measurement also implied that the automatic method is more reliable. The proposed method can be a promising approach for facilitating the applications of 3-D ultrasound imaging in the diagnosis, treatment, and screening of scoliosis.

3D Ultrasound imaging has become an important means of scoliosis assessment as it is a real-time, cost-effective and radiation-free imaging technique. However, the coronal images from different depths of a 3D ultrasound image have different imaging definitions. So there is a need to select the coronal image that would give the best image definition. Also, manual selection of coronal images is time-consuming and limited to the discretion and capability of the assessor. Therefore, in this paper, we have developed a convolution learning-to-rank algorithm to select the best ultrasound images automatically using raw ultrasound images. The ranking is done based on the curve angle of the spinal cord. Firstly, we approached the image selection problem as a ranking problem; ranked based on probability of an image to be a good image. Here, we use the RankNet, a pairwise learning-to-rank method, to rank the images automatically. Secondly, we replaced the backbone of the RankNet, which is the traditional artificial neural network (ANN), with convolution neural network (CNN) to improve the feature extracting ability for the successive iterations. The experimental result shows that the proposed convolutional RankNet achieves the perfect accuracy of 100% while conventional DenseNet achieved 35% only. This proves that the convolutional RankNet is more suitable to highlight the best quality of ultrasound image from multiple mediocre ones.

No abstract