Objective To develop a computer vision evaluation method to automatically measure the degree of scoliosis based on the machine learning algorithm.Methods For 129 X-ray images of the coronal plane of spine from patients with idiopathic scoliosis,after histogram equalization of original image,we use flipping method to magnify asymmetric elements,search for the global maximum pixel value in each line,and then,scan local maximum pixel value with neighborhood 3-5.The intersection set of the two point sets can be regarded as candidate anchor points.In following filtering step,we begin with the highest point,connecting second line's candidate anchor points.If the slope absolute value of the line larger or e-qual to 1,then the second line's candidate anchor can be entitled as a fine anchor.All fine anchors were fitted with cubic spline algorithm to obtain the approximate curve of the spine,and the degree of scoliosis was measured by the standardized integral area.The correlation between the measured Cobb angle and the integrated area algorithm was evaluated.The area under the curve measured by the integrated area algorithm was analyzed by receiver operating characteristic curve(ROC).Results The Cobb angle measured by manual was positively correlated with the integrated area algorithm(r =0.658,P<0.001),and the AUC of the integrated area algorithm for moderate and severe lateral bending was0.889 and0.862.Con-clusion The integrated area algorithm method can quickly and efficiently assess the degree of scoliosis from a global per-spective.It is suitable for screening the degree of large area scoliosis assessment,and can be a useful supplement to the fine measurement of scoliosis angle.
Measuring scoliosisX-ray imageMachine learningExhaustive search algorithmsanchor pointsAn-chor pointNormalized area
NETL
NSTL
万方数据
