Cervical vertebral corner detection using haar-like features and modified hough forest

Abstract

This is the accepted version of the paper.This version of the publication may differ from the final published version. Abstract-The neck (cervical spine) is a flexible part of the human body and is particularly vulnerable to injury. Patients suspected of cervical spine injuries are often imaged using lateral view radiographs. Incorrect diagnosis based on these images may lead to serious long-term consequences. Our overarching goal is to develop a computer-aided detection system to help an emergency room physician correctly diagnose a patient's injury. In this paper, we present a method to localize the corners of cervical vertebrae in a set of 90 lateral cervical radiographs. Haar-like features are computed using intensity and gradient image patches, each of which votes for possible corner position using a modified Hough forest regression technique. Votes are aggregated using two dimensional kernel density estimation, to find the location of the corner. Our method demonstrates promising results, identifying corners with an average median error of 2.08 mm. Permanent repository link

Improving an Active Shape Model with Random Classification Forest for Segmentation of Cervical Vertebrae

“…As our overarching goal is to develop an injury detection system, the segmentation still requires further work to apply morphometric analysis to detect injuries especially for the vertebrae with conditions like osteoporosis, fractures and implants. Our work is currently focused on improving the segmentation accuracy by experimenting with better and automatic initialization using vertebrae corners [18,19] and/or endplates [5].…”

“…When segmenting a vertebra in a new image, the mean shape is approximately initialized near the vertebra using manually clicked vertebra centers [18,19]. The model then looks for displacement of the mean shape towards the actual vertebra based on the extracted profiles perpendicular to the mean shape in the image (see Fig.…”

Vertebral corners detection on sagittal X-rays based on shape modelling, random forest classifiers and dedicated visual features

“…Overview of state of the art approaches for vertebra corner detection on planar X-ray images. NA: Not available, C: Cervical, L: Lumbar Among the literature that focuses on vertebrae corner detection and reports detection precision, (Al Arif et al, 2017, 2015a, 2015b provided these values on the cervical region (C3-C7). In (Al Arif et al, 2015b) they reported average RMSE values of 3.033.08 mm.…”

“…NA: Not available, C: Cervical, L: Lumbar Among the literature that focuses on vertebrae corner detection and reports detection precision, (Al Arif et al, 2017, 2015a, 2015b provided these values on the cervical region (C3-C7). In (Al Arif et al, 2015b) they reported average RMSE values of 3.033.08 mm. In (Al Arif et al, 2017), a mean corner localization error of 2.01 mm is reported using a Hough forest classifier with mixed intensity and contrast Haar-like features and a rectangular ROI.…”

“…Focusing exclusively on the cervical vertebrae (C3-C7), (Al Arif et al, 2015a) developed a semi-automated multiscale patch-based Hough forest-based corner detection method, which provides both anterior and posterior cervical vertebrae corners. They improved the performance of their algorithm by additional intensity and gradient information through tailored Haar-like features to a learned spatial prior probability distribution (Al Arif et al, 2015b). In their recent work (Al Arif et al, 2017), they further improved their algorithm performance by exploiting trapezoidal and rectangular ROIs, and a refined vote aggregation procedure from multiple patches.…”