Automation of Spine Curve Assessment in Frontal Radiographs Using Deep Learning of Vertebral-Tilt Vector

Abstract

In this paper, an automated and visually explainable system is proposed for a scoliosis assessment from spinal radiographs, which deals with the drawback of manual measurements, which are known to be time-consuming, cumbersome, and operator dependent. Deep learning techniques have been successfully applied in the accurate extraction of Cobb angle measurements, which is the gold standard for a scoliosis assessment. Such deep learning methods directly estimate the Cobb angle without providing structural information of the spine which can be used for diagnosis. Although conventional segmentation-based methods can provide the spine structure, they still have limitations in the accurate measurement of the Cobb angle. It would be desirable to build a clinician-friendly diagnostic system for scoliosis that provides not only an automated Cobb angle assessment but also local and global structural information of the spine. This paper addresses this need through the development of a hierarchical method which consisting of three major parts. (1) A confidence map is used to selectively localize and identify all vertebrae in an accurate and robust manner, (2) vertebral-tilt field is used to estimate the slope of an individual vertebra, and (3) the Cobb angle is determined by combining the vertebral centroids with the previously obtained vertebral-tilt field. The performance of the proposed method was validated, resulting in circular mean absolute error of 3.51 • and symmetric mean absolute percentage error of 7.84% for the Cobb angle.

Seg4Reg+: Consistency Learning Between Spine Segmentation and Cobb Angle Regression

“…Table 2 shows that applying the ROIE gate to the origin segmentation model promotes JA from 75.47% to 77.86%, and the visualization results in Figure 4 show that the ROIE gate is helpful for false Comparison with State-of-the-art. We compare the proposed method with state-of-the-art methods including A-Net [3], L-Net [3], BoostNet [13], PFA [10], VF [6], and Seg4Reg [7], which won the 1st place in AASCE challenge. We follow the same experiment setting with VF [6], and the performance of all competing methods is adopted from the original publications for a fair comparison.…”

“…We compare the proposed method with state-of-the-art methods including A-Net [3], L-Net [3], BoostNet [13], PFA [10], VF [6], and Seg4Reg [7], which won the 1st place in AASCE challenge. We follow the same experiment setting with VF [6], and the performance of all competing methods is adopted from the original publications for a fair comparison. As depicted in Table 3, promising results are observed in predicting Cobb angle using the proposed framework.…”

“…Conventional methods utilized active contour model [2], customized filter [1] and charged-particle models [8] for spine segmentation to calculate the Cobb angle, which are computationally expensive and the unclear spine boundary will result in inaccurate estimations. Recently, deep learning based methods [3,6,9,13,15,17] have been proposed to consolidate the tasks of vertebral landmark detection with Cobb angle estimation to improve the robustness of spinal curvature assessment. And Seg4Reg [7], which won the 1st place in the AASCE 1 challenge, regarded the segmentation results of spine as the input of the regression network for Cobb angle estimation.…”

“…Although the above methods have achieved great success, their applications to Cobb angle estimation suffer from three limitations: 1) The methods [3,6,9,13,15,17] relying on the landmark coordinates are susceptible since a small error in landmark coordinates may cause a huge mistake in angle estimation; 2) the two-stage frameworks [3,6,7,17] often suffers from the error accumulation; and 3) the cascaded networks [3,6,7] cannot guarantee a global optimum.…”

Localization and Edge-Based Segmentation of Lumbar Spine Vertebrae to Identify the Deformities Using Deep Learning Models

“…Kang Cheol Kim et al,in [39], presented an approach to identify scoliosis from X-ray images; they explained the drawbacks of manual measurements which are laborious and time-consuming. The method consists of three major parts: in the first part, a confidence map is utilized for localization.…”