David Major
VRVis
Network
Latest external collaboration on country level. Dive into details by clicking on the dots.
Publication
Featured researches published by David Major.
Medical Image Analysis | 2013
David Major; Jiří Hladůvka; Florian Schulze; Katja Bühler
The spinal column is one of the most distinguishable structures in CT scans of the superior part of the human body. It is not necessary to segment the spinal column in order to use it as a frame of reference. It is sufficient to place landmarks and the appropriate anatomical labels at intervertebral disks and vertebrae. In this paper, we present an automated system for landmarking and labeling spinal columns in 3D CT datasets. We designed this framework with two goals in mind. First, we relaxed input data requirements found in the literature, and we label both full and partial spine scans. Secondly, we intended to fulfill the performance requirement for daily clinical use and developed a high throughput system capable of processing thousands of slices in just a few minutes. To accomplish the aforementioned goals, we encoded structural knowledge from training data in probabilistic boosting trees and used it to detect efficiently the spinal canal, intervertebral disks, and three reference regions responsible for initializing the landmarking and labeling. Final landmarks and labels are selected by Markov Random Field-based matches of newly introduced 3-disk models. The framework has been tested on 36 CT images having at least one of the regions around the thoracic first ribs, the thoracic twelfth ribs, or the sacrum. In an average time of 2 min, we achieved a correct labeling in 35 cases with precision of 99.0% and recall of 97.2%. Additionally, we present results assuming none of the three reference regions could be detected.
IEEE Transactions on Medical Imaging | 2017
Alexey A. Novikov; Dimitrios Lenis; David Major; Jiri Hladůvka; Maria Wimmer; Katja Bühler
The success of deep convolutional neural networks (NNs) on image classification and recognition tasks has led to new applications in very diversified contexts, including the field of medical imaging. In this paper, we investigate and propose NN architectures for automated multiclass segmentation of anatomical organs in chest radiographs (CXRs), namely for lungs, clavicles, and heart. We address several open challenges including model overfitting, reducing number of parameters, and handling of severely imbalanced data in CXR by fusing recent concepts in convolutional networks and adapting them to the segmentation problem task in CXR. We demonstrate that our architecture combining delayed subsampling, exponential linear units, highly restrictive regularization, and a large number of high-resolution low-level abstract features outperforms state-of-the-art methods on all considered organs, as well as the human observer on lungs and heart. The models use a multiclass configuration with three target classes and are trained and tested on the publicly available Japanese Society of Radiological Technology database, consisting of 247 X-ray images the ground-truth masks for which are available in the segmentation in CXR database. Our best performing model, trained with the loss function based on the Dice coefficient, reached mean Jaccard overlap scores of 95% for lungs, 86.8% for clavicles, and 88.2% for heart. This architecture outperformed the human observer results for lungs and heart.
international symposium on biomedical imaging | 2016
Maria Wimmer; David Major; Alexey A. Novikov; Katja Bühler
We present a novel pipeline for acquisition protocol independent spine labeling in volumetric Magnetic Resonance Imaging (MRI) data of the lumbar spine. Our learning-based system uses local Entropy-optimized Texture Models (ETMs) for reducing the intensity scale in clinical data to only a few gray levels. The task of intervertebral disc localization is then performed on the normalized data. The benefit of our method is, that we can deal with various MRI protocols, such as T1-weighted (T1w) and T2-weighted (T2w) scans. Using the entropy objective allows us furthermore to apply the algorithm to acquisition protocols which are not covered by the training set. We achieve high disc localization accuracies for both, MRI protocols which are covered and not covered by training. The approach can be easily extended to other modalities.
eurographics | 2015
David Major; Alexey A. Novikov; Maria Wimmer; Jiří Hladůvka; Katja Bühler
Automated identification of main arteries in Computed Tomography Angiography (CTA) scans plays a key role in the initialization of vessel tracking algorithms. Automated vessel tracking tools support physicians in vessel analysis and make their workflow time-efficient. We present a fully-automated framework for identification of five main arteries of three different body regions in various field-of-view CTA scans. Our method detects the two common iliac arteries, the aorta and the two common carotid arteries and delivers seed positions in them. After the field-of-view of a CTA scan is identified, artery candidate positions are regressed slice-wise and the best candidates are selected by Naive Bayes classification. Final artery seed positions are detected by picking the most optimal path over the artery classification results from slice to slice. Our method was evaluated on 20 CTA scans with various field-of-views. The high detection performance on different arteries shows its generality and future applicability for automated vessel analysis systems.
IEEE Transactions on Medical Imaging | 2017
Alexey A. Novikov; David Major; Maria Wimmer; Gert Sluiter; Katja Bühler
We propose an automated pipeline for vessel centerline extraction in 3-D computed tomography angiography (CTA) scans with arbitrary fields of view. The principal steps of the pipeline are body part detection, candidate seed selection, segment tracking, which includes centerline extraction, and vessel tree growing. The final tree-growing step can be instantiated in either a semi- or fully automated fashion. The fully automated initialization is carried out using a vessel position regression algorithm. Both semi-and fully automated methods were evaluated on 30 CTA scans comprising neck, abdominal, and leg arteries in multiple fields of view. High detection rates and centerline accuracy values for 38 distinct vessels demonstrate the effectiveness of our approach.
eurographics | 2015
Alexey A. Novikov; Maria Wimmer; David Major; Katja Bühler
We introduce a cascade classification algorithm for bifurcation detection in Computed Tomography Angiography (CTA) scans of blood vessels. The proposed algorithm analyzes the vessel surrounding by a trained classifier first, followed by an accurate segmentation of the vessel outer wall by Morphological Active Contour Without Edges and finally extracts the boundary features of the segmented object and classifies its shape by Approximate K-nearest Neighbour classifier. The algorithm shows encouraging and competitive results for blood vessels from various parts of a human body including head, neck and legs.
Archive | 2015
Jiří Hladůvka; David Major; Katja Bühler
Algorithms centered around spinal columns in CT data such as spinal canal detection, disk and vertebra localization and segmentation are known to be computationally intensive and memory demanding. The majority of these algorithms need initialization and try to reduce the search space to a minimum. In this work we introduce bone profiles as a simple means to compute a tight ROI containing the spine and seed points within the spinal canal. Bone profiles rely on the distribution of bone intensity values in axial slices. They are easy to understand, and parameters guiding the ROI and seed point detection are straight forward to derive. The method has been validated with two datasets containing 52 general and 242 spine-focused CT scans. Average runtimes of 1.5 and 0.4 s are reported on a single core. Due to its slice-wise nature, the method can be easily parallelized and fractions of the reported runtimes can be further achieved. Our memory requirements are upper bounded by a single CT slice.
Medical Image Analysis | 2014
David Major; Jiří Hladůvka; Florian Schulze; Katja Bühler
1361-8415/
international conference in central europe on computer graphics and visualization | 2011
Florian Schulze; David Major; Katja Bühler
see front matter 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.media.2014.01.008 DOI of original article: http://dx.doi.org/10.1016/j.media.2013.07.005 E-mail address: [email protected] (D. Major) Table 1 Overview of state-of-the-art approaches and our method. Data modality and requirement to contain a specific part of spine (‘‘C’’ – cervical, ‘‘T’’ – thorax, ‘‘L’’ – lumbar, ‘‘f spine scan and ‘‘T12’’ – thoracic twelfth vertebra/rib). Whether tested with pathology data or not. How many imaging vendors do the test data come from. Correct labe (correctly labeled test data/all test data and correctly labeled vertebra or disk/all tested vertebra or disk) and time performance in seconds. ‘‘–’’ means not available.
IEEE Transactions on Medical Imaging | 2018
Alexey A. Novikov; Dimitrios Lenis; David Major; Jiri Hladuvka; Maria Wimmer; Katja Bühler