Hierarchical parsing and semantic navigation of full body CT data

Seifert, Sascha; Barbu, Adrian; Zhou, S. Kevin; Liu, David; Feulner, Johannes; Huber, Martin; Suehling, Michael; Cavallaro, Alexander; Comanicìu, Dorin

doi:10.1117/12.812214

Cited by 121 publications

(121 citation statements)

References 13 publications

Supporting

Mentioning

116

Contrasting

Unclassified

Order By: Relevance

“…Machine learning approaches as introduced by Seifert et.al. [11] and Zheng et.al. [12] for the detection of body landmarks in combination with vessel segmentation algorithms [13,14] allow a fully automatic segmentation of vessels and visualization in preparation for reading.…”

Section: Introductionmentioning

confidence: 93%

Evaluation of Computer-Assisted Quantification of Carotid Artery Stenosis

et al. 2011

View full text Add to dashboard Cite

The purpose of this study was to evaluate the influence of advanced software assistance on the assessment of carotid artery stenosis; particularly, the inter-observer variability of readers with different level of experience is to be investigated. Forty patients with suspected carotid artery stenosis received head and neck dual-energy CT angiography as part of their pre-interventional workup. Four blinded readers with different levels of experience performed standard imaging interpretation. At least 1 day later, they performed quantification using an advanced vessel analysis software including automatic dual-energy bone and hard plaque removal, automatic and semiautomatic vessel segmentation, as well as creation of curved planar reformation. Results were evaluated for the reproducibility of stenosis quantification of different readers by calculating the kappa and correlation values. Consensus reading of the two most experienced readers was used as the standard of reference. For standard imaging interpretation, experienced readers reached very good (k=0.85) and good (k=0.78) inter-observer variability. Inexperienced readers achieved moderate (k=0.6) and fair (k=0.24) results. Sensitivity values 80%, 91%, 83%, 77% and specificity values 100%, 84%, 82%, 53% were achieved for significant area stenosis >70%. For grading using advanced vessel analysis software, all readers achieved good inter-observer variability (k=0. 77, 0.72, 0.71, and 0.77). Specificity values of 97%, 95%, 95%, 93% and sensitivity values of 84%, 78%, 86%, 92% were achieved. In conclusion, when supported by advanced vessel analysis software, experienced readers are able to achieve good reproducibility. Even inexperienced readers are able to achieve good results in the assessment of carotid artery stenosis when using advanced vessel analysis software.

show abstract

Section: Introductionmentioning

confidence: 93%

Evaluation of Computer-Assisted Quantification of Carotid Artery Stenosis

et al. 2011

View full text Add to dashboard Cite

show abstract

“…The detected bounding boxes could, for example, be used for initializing a detailed vertebra segmentation algorithm with subsequent analysis. Furthermore, the proposed ssystem can support semantic body parsing and semantic annotation to automatically generate semantic location description frequently used by physicians for reporting [12,11]. Both applications will be considered in future work.…”

Section: Discussionmentioning

confidence: 99%

Detection of 3D Spinal Geometry Using Iterated Marginal Space Learning

Kelm

Zhou

Suehling

et al. 2011

Medical Computer Vision. Recognition Techniques and Applications in Medical Imaging

Self Cite

View full text Add to dashboard Cite

Abstract. Determining spinal geometry and in particular the position and orientation of the intervertebral disks is an integral part of nearly every spinal examination with Computed Tomography (CT) and Magnetic Resonance (MR) imaging. It is particularly important for the standardized alignment of the scan geometry with the spine. In this paper, we present a novel method that combines Marginal Space Learning (MSL), a recently introduced concept for efficient discriminative object detection, with a generative anatomical network that incorporates relative pose information for the detection of multiple objects. It is used to simultaneously detect and label the intervertebral disks in a given spinal image volume. While a novel iterative version of MSL is used to quickly generate candidate detections comprising position, orientation, and scale of the disks with high sensitivity, the anatomical network selects the most likely candidates using a learned prior on the individual nine dimensional transformation spaces. Since the proposed approach is learning-based it can be trained for MR or CT alike. Experimental results based on 42 MR volumes show that our system not only achieves superior accuracy but also is the fastest system of its kind in the literature -on average, the spinal disks of a whole spine are detected in 11.5s with 98.6% sensitivity and 7.3% false positive detections. An average position error of 2.4mm and angular error of 3.9 • is achieved.

show abstract

“…In [8], the pairwise spatial context is used to compute the information gain that guides an active scheduling scheme for detecting multiple landmarks. Seifert et al [9] encoded pairwise spatial contexts into a discriminative anatomical network. The holistic context goes beyond the relationships among a cohort of landmarks and refers to the whole relationship between all voxels and the landmarks; in other words, regarding the image as a whole.…”

Section: Related Work and Context Exploitationmentioning

confidence: 99%

Anatomical Landmark Detection Using Nearest Neighbor Matching and Submodular Optimization

Liu

Zhou

2012

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

Self Cite

View full text Add to dashboard Cite

Abstract. We present a two-stage method for effective and efficient detection of one or multiple anatomical landmarks in an arbitrary 3D volume. The first stage of nearest neighbor matching is to roughly estimate the landmark locations. It searches out of 100,000 volumes for the closest to an input volume and then transfers landmark annotations to the input. The second stage of submodular optimization is to refine the landmark locations by running discriminative landmark detectors within the search ranges constrained by the first stage results. Further it coordinates multiple detectors with a search strategy optimized on the fly to reduce the overall computation cost arising in a submodular formulation. We validate the accuracy, speed and robustness of our approach by detecting body regions and landmarks in a dataset of 2,500 CT scans.

show abstract

Hierarchical parsing and semantic navigation of full body CT data

Cited by 121 publications

References 13 publications

Evaluation of Computer-Assisted Quantification of Carotid Artery Stenosis

Evaluation of Computer-Assisted Quantification of Carotid Artery Stenosis

Detection of 3D Spinal Geometry Using Iterated Marginal Space Learning

Anatomical Landmark Detection Using Nearest Neighbor Matching and Submodular Optimization

Contact Info

Product

Resources

About