Skip to main content

Über dieses Buch

This book constitutes the refereed proceedings of the 4th International Workshop and Challenge on Computational Methods and Clinical Applications for Spine Imaging, CSI 2016, held in conjunction with MICCAI 2016, in Athens, Greece, in October 2016.

The 13 workshop papers were carefully reviewed and selected for inclusion in this volume. They aim at reviewing the state-of-the-art techniques, sharing the novel and emerging analysis and visualization techniques and discussing the clinical challenges and open problems in this rapidly growing field - including all major aspects of problems related to spine imaging, including clinical applications of spine imaging, computer aided diagnosis of spine conditions, computer aided detection of spine-related diseases, emerging computational imaging techniques for spinal diseases, fast 3D reconstruction of spine, feature extraction, multiscale analysis, pattern recognition, image enhancement of spine imaging, image-guided spine intervention and treatment, multimodal image registration and fusion for spine imaging, novel visualization techniques, segmentation techniques for spine imaging, statistical and geometric modeling for spine and vertebra, spine and vertebra localization.





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

X-ray is a common modality for diagnosing cervical vertebrae injuries. Many injuries are missed by emergency physicians which later causes life threatening complications. Computer aided analysis of X-ray images has the potential to detect missed injuries. Segmentation of the vertebrae is a crucial step towards automatic injury detection system. Active shape model (ASM) is one of the most successful and popular method for vertebrae segmentation. In this work, we propose a new ASM search method based on random classification forest and a kernel density estimation-based prediction technique. The proposed method have been tested on a dataset of 90 emergency room X-ray images containing 450 vertebrae and outperformed the classical Mahalanobis distance-based ASM search and also the regression forest-based method.

S. M. Masudur Rahman Al Arif, Michael Gundry, Karen Knapp, Greg Slabaugh

Machine Learning Based Bone Segmentation in Ultrasound

Ultrasound (US) guidance is of increasing interest for minimally invasive procedures in orthopedics due to its safety and cost benefits. However, bone segmentation from US images remains a challenge due to the low signal to noise ratio and artifacts that hamper US images. We propose to learn the appearance of bone-soft tissue interfaces from annotated training data, and present results with two classifiers, structured forest and a cascaded logistic classifier. We evaluated the proposed methods on 143 spinal images from two datasets acquired at different sites. We achieved a segmentation recall of 0.9 and precision 0.91 for the better dataset, and a recall and precision of 0.87 and 0.81 for the combined dataset, demonstrating the potential of the framework.

Nora Baka, Sieger Leenstra, Theo van Walsum

Variational Segmentation of the White and Gray Matter in the Spinal Cord Using a Shape Prior

Segmenting the inner structure of the spinal cord on magnetic resonance (MR) images is difficult because of poor contrast between white and gray matter (WM/GM). We present a variational formulation to automatically detect cerebrospinal fluid and WM/GM. The segmentation results are obtained by continuous cuts combined with a shape prior. Intensity-based segmentation guarantees high accuracy while the shape prior aims at precision. We tested the algorithm on a set of MR images with visual WM/GM contrast and evaluated it w.r.t. manual GM segmentations. The automated GM segmentations are on a par with the manual results.

Antal Horváth, Simon Pezold, Matthias Weigel, Katrin Parmar, Oliver Bieri, Philippe Cattin

Automated Intervertebral Disc Segmentation Using Deep Convolutional Neural Networks

In this paper, we propose to use deep convolutional neural networks to solve the challenging Intervertebral Disc (IVD) segmentation problem. We investigated the influence of four different patch sampling strategies on the performance of the deep convolutional neural networks. Evaluated on the MICCAI 2015 IVD segmentation challenge datasets, our method achieved a mean Dice overlap coefficient of 89.2% and a mean average absolute surface distance of 1.3 mm. The results achieved by our method are comparable with those achieved by the state-of-the-art methods.

Xing Ji, Guoyan Zheng, Daniel Belavy, Dong Ni



Fully Automatic Localisation of Vertebrae in CT Images Using Random Forest Regression Voting

We describe a system for fully automatic vertebra localisation and segmentation in 3D CT volumes containing arbitrary regions of the spine, with the aim of detecting osteoporotic fractures. To avoid the difficulties of high-resolution manual annotation on overlapping structures in 3D, the system consists of several 2D operations. First, a Random Forest regressor is used to localise the spinal midplane in a coronal maximum intensity projection. A 2D sagittal image showing the midplane is then produced. A second set of regressors are used to localise each vertebral body in this image. Finally, a Random Forest Regression Voting Constrained Local Model is used to segment each detected vertebra.The system was evaluated on 402 CT volumes. 83% of vertebrae between T4 and L4 were detected and, of these, 97% were segmented with a mean error of less than or equal to $$1\,mm$$. A simple classifier was applied to perform a fracture/non-fracture classification for each image, achieving 69% recall at 70% precision.

Paul A. Bromiley, Eleni P. Kariki, Judith E. Adams, Timothy F. Cootes

Global Localization and Orientation of the Cervical Spine in X-ray Images

Injuries in cervical spine X-ray images are often missed by emergency physicians. Many of these missing injuries cause further complications. Automated analysis of the images has the potential to reduce the chance of missing injuries. Towards this goal, this paper proposes an automatic localization of the spinal column in cervical spine X-ray images. The framework employs a random classification forest algorithm with a kernel density estimation-based voting accumulation method to localize the spinal column and to detect the orientation. The algorithm has been evaluated with 90 emergency room X-ray images and has achieved an average detection accuracy of 91% and an orientation error of 3.6$$^{\circ }$$. The framework can be used to narrow the search area for other advanced injury detection systems.

S. M. Masudur Rahman Al Arif, Michael Gundry, Karen Knapp, Greg Slabaugh

Accurate Intervertebral Disc Localisation and Segmentation in MRI Using Vantage Point Hough Forests and Multi-atlas Fusion

An accurate method for localising and segmenting intervertebral discs in magnetic resonance (MR) spine imaging is presented. Atlas-based labelling of discs in MRI is challenging due to the small field of view and repetitive structures, which may cause the image registration to converge to a local minimum. To tackle this initialisation problem, our approach uses Vantage Point Hough Forests to automatically and robustly regress landmark positions, which are used to initialise a discrete deformable registration of all training images. An image-adaptive fusion of propagated segmentation labels is obtained by non-negative least-squares regression. Despite its simplicity and without using specific domain knowledge, our approach achieves sub-voxel localisation accuracy of 0.61 mm, Dice segmentation overlaps of nearly 90% (for the training data) and takes less than ten minutes to process a new scan.

Mattias P. Heinrich, Ozan Oktay

Multi-scale and Modality Dropout Learning for Intervertebral Disc Localization and Segmentation

Automatic localization and segmentation of intervertebral discs (IVDs) from volumetric magnetic resonance (MR) images is important for spine disease diagnosis. It dramatically alleviates the workload of radiologists given that the traditional manual annotation is time-consuming and error-prone with limited reproducibility. Compared with single modality data, multi-modality MR images are able to provide complementary information. However, how to effectively integrate them to generate more accurate segmentation results still remains open for studies. In this paper, we introduce a multi-scale and modality dropout learning framework to segment IVDs from four-modality MR images. Specifically, we design a 3D fully convolutional network which takes multiple scales of images as input and merges these pathways at higher layers to jointly integrate multi-scale information. Furthermore, in order to harness the complementary information from different modalities, we propose a modality dropout strategy to alleviate the co-adaption issue during the training. We evaluated our method on the MICCAI 2016 Challenge on Automatic Intervertebral Disc Localization and Segmentation from 3D Multi-modality MR Images. Our method achieved the best overall performance with the mean segmentation Dice as 91.2% and localization error as 0.62 mm, which demonstrated the superiority of our proposed framework.

Xiaomeng Li, Qi Dou, Hao Chen, Chi-Wing Fu, Pheng-Ann Heng

Fully Automatic Localization and Segmentation of Intervertebral Disc from 3D Multi-modality MR Images by Regression Forest and CNN

In this paper, we propose a fully automatic framework to localize and segment intervertebral discs (IVDs) from 3D Multi-modality MR Images. Random forest regression is employed to coarsely localize the IVD. Then IVDs are segmented sequentially by training the specific convolutional neural network (CNN) classifier for each IVD. We compared the performance using single- and multi-modality images. Evaluated on the MICCAI 2016 IVD on-site challenge datasets, our method achieved a mean localization distance of 0.64 mm and a mean Dice overlap coefficient of 90.8%. The results show that our method is robust and comparable with state-of-the-art methods.

Xing Ji, Guoyan Zheng, Li Liu, Dong Ni

Computer Aided Diagnosis and Intervention


Manual and Computer-Assisted Pedicle Screw Placement Plans: A Quantitative Comparison

In this paper, we present a quantitative comparison of manual and computer-assisted preoperative pedicle screw placement plans, obtained from three-dimensional (3D) computed tomography (CT) images of 17 patients with thoracic spinal deformities. Manual planning was performed by two spine surgeons by means of a dedicated software for planning of surgical procedures, while computer-assisted planning was based on automated 3D segmentation and modeling of vertebral structures from CT images, and automated modeling of the pedicle screw in 3D with maximization of the screw fastening strength. The analysis of the size (diameter and length) and insertion trajectory (pedicle crossing point, sagittal and axial inclinations) for 316 pedicle screws revealed a statistically significant difference in the screw size and insertion trajectory. However, computer-assisted planning did not propose narrower and shorter screws, which was reflected through a higher normalized screw fastening strength.

Dejan Knez, Janez Mohar, Robert J. Cirman, Boštjan Likar, Franjo Pernuš, Tomaž Vrtovec

Detection of Degenerative Osteophytes of the Spine on PET/CT Using Region-Based Convolutional Neural Networks

The identification and detection of degenerative osteophytes of the spine is a challenging and time-consuming task that is important for the diagnosis of many spine diseases. Previous attempts to automate this task have been focused on using image features derived from radiographic diagnostic expertise rather than directly learning features. In this paper, we present a bottom-up approach to generate features for classification using a region-based convolutional neural network with unwrapped cortical shell maps from 18F-NaF positron emission tomography and computed tomography scans of the vertebral bodies of the thoracic and lumbar spine. We evaluated osteophyte detection performance on 45 individuals with 5-fold cross validation and achieved state-of-the-art performance with 85% sensitivity at 2 false positive detections per patient.

Yinong Wang, Jianhua Yao, Joseph E. Burns, Jiamin Liu, Ronald M. Summers

Reconstruction of 3D Lumvar Vertebra from Two X-ray Images Based on 2D/3D Registration

Constructing a 3D bone from two X-ray images is a challenging task, especially when we would like to build a complicated structure like spine. This paper presents a novel method for reconstructing lumbar vertebra by building correspondence of two X-ray images with a prior model. First, the pose between X-ray images and the vertebra model was estimated; second, the correspondences between the Digitally Reconstructed Radiographies (DRRs) and vertebra model were built; third, the deformation field from DRRs to X-ray images was calculated; last, deformation field was applied to vertebra model to generate the patient’s specified 3D model. This method just needs one prior model for 3D reconstruction. The experiments on nine vertebrae of three patients show the average reconstruction error is 1.2 mm (1.0 mm–1.3 mm) which is comparable to the state of the art.

Longwei Fang, Zuowei Wang, Zhiqiang Chen, Fengzeng Jian, Huiguang He

Classification of Progressive and Non-progressive Scoliosis Patients Using Discriminant Manifolds

Adolescent idiopathic scoliosis (AIS) is a 3-D deformation of the spine. Identifying curve progression in AIS at the first visit is a clinically relevant problem but remains challenging due to lack of relevant descriptors. We present here a classification framework to identify patients whose spine deformity will progress from those who will remain stable. The method uses personalized 3-D spine reconstructions at baseline from progressive (P) and non-progressive (NP) patients to train a predictive model. Morphological changes between groups are detected using a manifold learning algorithm based on Grassmannian kernels in order to assess the similarity between shape topology and inter-vertebral poses in both groups (P, NP). We test the method to classify 52 progressive and 81 non-progressive patients enrolled in a prospective clinical study, yielding classification rates comparing favorably to standard classification methods.

William Mandel, Robert Korez, Marie-Lyne Nault, Stefan Parent, Samuel Kadoury


Weitere Informationen

Premium Partner

Neuer Inhalt

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.



Best Practices für die Mitarbeiter-Partizipation in der Produktentwicklung

Unternehmen haben das Innovationspotenzial der eigenen Mitarbeiter auch außerhalb der F&E-Abteilung erkannt. Viele Initiativen zur Partizipation scheitern in der Praxis jedoch häufig. Lesen Sie hier  - basierend auf einer qualitativ-explorativen Expertenstudie - mehr über die wesentlichen Problemfelder der mitarbeiterzentrierten Produktentwicklung und profitieren Sie von konkreten Handlungsempfehlungen aus der Praxis.
Jetzt gratis downloaden!