Artificial Intelligence for Automatic Measurement of Sagittal Vertical Axis Using ResUNet Framework

Abstract

We present an automated method for measuring the sagittal vertical axis (SVA) from lateral radiography of whole spine using a convolutional neural network for keypoint detection (ResUNet) with our improved localization method. The algorithm is robust to various clinical conditions, such as degenerative changes or deformities. The ResUNet was trained and evaluated on 990 standing lateral radiographs taken at Chang Gung Memorial Hospital, Linkou and performs SVA measurement with median absolute error of 1.183 ± 0.166 mm. The 5-mm detection rate of the C7 body and the sacrum are 91% and 87%, respectively. The SVA calculation takes approximately 0.2 s per image. The intra-class correlation coefficient of the SVA estimates between the algorithm and physicians of different years of experience ranges from 0.946 to 0.993, indicating an excellent consistency. The superior performance of the proposed method and its high consistency with physicians proved its usefulness for automatic measurement of SVA in clinical settings.

Deep learning approach for automatic landmark detection and alignment analysis in whole-spine lateral radiographs

“…Our model differs from the original CPN since (1) we added Differentiable Spatial to Numerical Transform (DSNT) 12 layers so that the landmark coordinates can be regressed directly; (2) similar to the original CPN, we used 2D heatmaps (probability density maps) to indicate the probable locations of landmarks. However, we added an additional regularisation loss on heatmaps (as illustrated in the Methods section) so that our model can predict heatmaps of landmarks with arbitrary shapes and sizes at the first stage and with small splotches of constraint shapes (narrow Gaussian or narrow exponential 6 ) at the second stage. Our deep learning model can thus localise the anatomic landmarks in a two-stage, coarse-to-fine manner.…”

“…Additional loss for heatmap regularisation (Jenson–Shannon entropy) is then considered to encourage that . In our paper, we experiment with the following forms of 6 : where is a positive real number, is a Gaussian, is an exponential function, and and are the normalisation constants ensuring that and . These two functions are designed such that they reach the same half maxima at and .…”

Moderate Degeneration of the Lumbar Paravertebral Muscle in Patients With Dynamic Sagittal Imbalance,A retrospective cohort study

“…Sagittal balance is a primary issue for clinical assessment of spine. [1]The larger the C7-SVA, the worse the HRQOL(Health related quality of life). [2]Skeletal structure and paravertebral muscles of spine play an important role in maintaining spinal stability.Among them, ES and MF of lumbar spine were considered important in paravertebral muscles [20]Their function is the extension of the spine in reaction to gravity and body weight as well as the maintainance of lumbar spine stability,degeneration of them can cause instability of the spine.…”

Degeneration of the Lumbar Paravertebral Muscle in Patients With Dynamic Sagittal Imbalance, A Retrospective Cohort Study

“…Sagittal balance is a primary issue for clinical assessment of spine. [1]The value of SVA has obvious correlation with HRQOL(Health related quality of life).The larger the C7-SVA, the worse the HRQOL.…”