Spring-Charged Particles Model to Improved Shape Recovery: An Application for X-Ray Spinal Segmentation
Tri A. Sardjono, Peter M. A. Van Ooijen, I. Ketut Eddy Purnama and 6 more
IEEE Access · 2022
doi:10.1109/access.2022.3212755 (opens in a new tab)
We email you when a new notice or a question about this paper appears.
Abstract
Deformable models are widely used in medical image segmentation methods, to find not only single but also multiple objects within an image. They have the ability to follow the contours of an object of interest, define the boundary of ROI (Region Of Interest) and improve shape recovery. However, these methods still have limitations in cases of low image quality or clutter. This paper presents a new deformable model, the Spring-Charged Particles Model (SCPM). It simulates the movement of positively charged particles connected by springs, attracted towards the contour of objects of interest which is charged negatively, according to the gradient-magnitude image. Springs prevent the particles from moving away and keep the particles at appropriate distances without reducing their flexibility. SCPM was tested on simple shape images and on frontal X-ray images of scoliosis patients. Artificial noise was added to the simple images to examine the robustness of the method. Several configurations of springs and positively charged-particles were evaluated by determining the best spinal segmentation result. The performance of SCPM was compared to the Charged Fluid Model (CFM), Active Contours, and a convolutional neural network (CNN) with U-Net architecture to measure its ability for determining the curvature of the spinal column from frontal X-Ray images. The results show that SCPM is better at segmenting the spine and determining its curvature, as indicated by the highest Area Score value of 0.837, and the lowest standard deviation value of 0.028.
Abstract licence: CC BY-NC-ND
Reads
Read by people on this page
None yet.
Checked or read by AI through naulon
None yet.
A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.
Are you an author?
Is this your paper?
Claim it to add your note and be paid when agents cite it. Free.