Ludwig Boltzmann Institute for Clinical Forensic Imaging

Graz, Austria

Ludwig Boltzmann Institute for Clinical Forensic Imaging

Graz, Austria

Time filter

Source Type

PubMed | University of Western Ontario, University of California at Irvine, University of Ljubljana, University of Sheffield and 5 more.
Type: | Journal: Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society | Year: 2016

A multiple center milestone study of clinical vertebra segmentation is presented in this paper. Vertebra segmentation is a fundamental step for spinal image analysis and intervention. The first half of the study was conducted in the spine segmentation challenge in 2014 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) Workshop on Computational Spine Imaging (CSI 2014). The objective was to evaluate the performance of several state-of-the-art vertebra segmentation algorithms on computed tomography (CT) scans using ten training and five testing dataset, all healthy cases; the second half of the study was conducted after the challenge, where additional 5 abnormal cases are used for testing to evaluate the performance under abnormal cases. Dice coefficients and absolute surface distances were used as evaluation metrics. Segmentation of each vertebra as a single geometric unit, as well as separate segmentation of vertebra substructures, was evaluated. Five teams participated in the comparative study. The top performers in the study achieved Dice coefficient of 0.93 in the upper thoracic, 0.95 in the lower thoracic and 0.96 in the lumbar spine for healthy cases, and 0.88 in the upper thoracic, 0.89 in the lower thoracic and 0.92 in the lumbar spine for osteoporotic and fractured cases. The strengths and weaknesses of each method as well as future suggestion for improvement are discussed. This is the first multi-center comparative study for vertebra segmentation methods, which will provide an up-to-date performance milestone for the fast growing spinal image analysis and intervention.

PubMed | University of Western Ontario, is Center for Virtual Reality and Visualization, University of Queensland, Ludwig Boltzmann Institute for Clinical Forensic Imaging and 11 more.
Type: | Journal: Medical image analysis | Year: 2016

The evaluation of changes in Intervertebral Discs (IVDs) with 3D Magnetic Resonance (MR) Imaging (MRI) can be of interest for many clinical applications. This paper presents the evaluation of both IVD localization and IVD segmentation methods submitted to the Automatic 3D MRI IVD Localization and Segmentation challenge, held at the 2015 International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI2015) with an on-site competition. With the construction of a manually annotated reference data set composed of 25 3D T2-weighted MR images acquired from two different studies and the establishment of a standard validation framework, quantitative evaluation was performed to compare the results of methods submitted to the challenge. Experimental results show that overall the best localization method achieves a mean localization distance of 0.8mm and the best segmentation method achieves a mean Dice of 91.8%, a mean average absolute distance of 1.1mm and a mean Hausdorff distance of 4.3mm, respectively. The strengths and drawbacks of each method are discussed, which provides insights into the performance of different IVD localization and segmentation methods.

Langkammer C.,Medical University of Graz | Langkammer C.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Krebs N.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Goessler W.,University of Graz | And 5 more authors.
Radiology | Year: 2010

Purpose: To investigate the relationship between transverse relaxation rates R2 and R2*, the most frequently used surrogate markers for iron in brain tissue, and chemically determined iron concentrations. Materials and Methods: This study was approved by the local ethics committee, and informed consent was obtained from each individual's next of kin. Quantitative magnetic resonance (MR) imaging was performed at 3.0 T in seven human postmortem brains in situ (age range at death, 38-81 years). Following brain extraction, iron concentrations were determined with inductively coupled plasma mass spectrometry in prespecified gray and white matter regions and correlated with R2 and R2* by using linear regression analysis. Hemispheric differences were tested with paired t tests. Results: The highest iron concentrations were found in the globus pallidus (mean ± standard deviation, 205 mg/kg wet mass ± 32), followed by the putamen (mean, 153 mg/kg wet mass ± 29), caudate nucleus (mean, 92 mg/kg wet mass ± 15), thalamus (mean, 49 mg/kg wet mass ± 11), and white matter regions. When all tissue samples were considered, transverse relaxation rates showed a strong linear correlation with iron concentration throughout the brain(r2 = 0.67 for R2, r 2 = 0.90 for R2*; P < .001). In white matter structures, only R2* showed a linear correlation with iron concentration. Chemical analysis revealed significantly higher iron concentrations in the left hemisphere than in the right hemisphere, a finding that was not reflected in the relaxation rates. Conclusion: Because of their strong linear correlation with iron concentration, both R2 and R2* can be used to measure iron deposition in the brain. Because R2* is more sensitive than R2 to variations in brain iron concentration and can detect differences in white matter, it is the preferred parameter for the assessment of iron concentration in vivo. © RSNA, 2010.

Petrovic A.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Petrovic A.,University of Graz | Scheurer E.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Scheurer E.,Medical University of Graz | Stollberger R.,University of Graz
Magnetic Resonance in Medicine | Year: 2015

Purpose: T2 quantification with multiecho sequences is typically impaired by the contribution of stimulated echoes to the echo decay due to B1+ inhomogeneity and slice profile effects. In this work, a compact signal model based on the generating functions approach, which accounts for both sources of error, is presented. Methods: The generating functions (GF) approach is used to obtain a closed solution to the evolution of the transverse magnetization in an echo train, however, not in the time domain, but in the transformed z-domain. The approach is generalized by the incorporation of flip angle distribution across the refocusing slice profiles. The approach is tested by fitting the model to simulated data as well as to phantom and in vivo measurements, followed by a comparison with the common monoexponential fitting approach. Results: The fitting simulations indicate that T2 errors of up to 30% can be commonplace in a clinical setting using the monoexponential method. Conversely, the GF approach produced accurate results. Phantom and in vivo experiments showed a good agreement of the GF values with spectroscopic measurements and single-echo spin-echo sequences. Conclusion: A correction for stimulated echoes is necessary to compute comparable T2 values. The presented approach provides a solution to this issue. © 2014 Wiley Periodicals, Inc.

Riegler G.,Graz University of Technology | Urschler M.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Ruther M.,Graz University of Technology | Bischof H.,Graz University of Technology | Stern D.,Graz University of Technology
Proceedings of the IEEE International Conference on Computer Vision | Year: 2016

An important initial step in many medical image analysis applications is the accurate detection of anatomical landmarks. Most successful methods for this task rely on data-driven machine learning algorithms. However, modern machine learning techniques, e.g. convolutional neural networks, need a large corpus of training data, which is often an unrealistic setting for medical datasets. In this work, we investigate how to adapt synthetic image datasets from other computer vision tasks to overcome the under-representation of the anatomical pose and shape variations in medical image datasets. We transform both data domains to a common one in such a way that a convolutional neural network can be trained on the larger synthetic image dataset and fine-tuned on the smaller medical image dataset. Our evaluations on data of MR hand and whole body CT images demonstrate that this approach improves the detection results compared to training a convolutional neural network only on the medical data. The proposed approach may also be usable in other medical applications, where training data is scarce. © 2015 IEEE.

Sumpf T.J.,Max Planck Institute for Biophysical Chemistry | Petrovic A.,University of Graz | Petrovic A.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Uecker M.,University of California at Berkeley | And 3 more authors.
IEEE Transactions on Medical Imaging | Year: 2014

A model-based reconstruction technique for accelerated T2 mapping with improved accuracy is proposed using undersampled Cartesian spin-echo magnetic resonance imaging (MRI) data. The technique employs an advanced signal model for T2 relaxation that accounts for contributions from indirect echoes in a train of multiple spin echoes. An iterative solution of the nonlinear inverse reconstruction problem directly estimates spin-density and T2 maps from undersampled raw data. The algorithm is validated for simulated data as well as phantom and human brain MRI at 3T. The performance of the advanced model is compared to conventional pixel-based fitting of echo-time images from fully sampled data. The proposed method yields more accurate T2 values than the mono-exponential model and allows for retrospective undersampling factors of at least 6. Although limitations are observed for very long T2 relaxation times, respective reconstruction problems may be overcome by a gradient dampening approach. The analytical gradient of the utilized cost function is included as Appendix. The source code is made available to the community. © 1982-2012 IEEE.

Urschler M.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Bornik A.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Donoser M.,Graz University of Technology
Proceedings of the IEEE International Conference on Computer Vision | Year: 2013

Integral image data structures are very useful in computer vision applications that involve machine learning approaches based on ensembles of weak learners. The weak learners often are simply several regional sums of intensities subtracted from each other. In this work we present a memory efficient integral volume data structure, that allows reduction of required RAM storage size in such a supervised learning framework using 3D training data. We evaluate our proposed data structure in terms of the tradeoff between computational effort and storage, and show an application for 3D object detection of liver CT data. © 2013 IEEE.

Ith M.,University of Bern | Scheurer E.,University of Bern | Scheurer E.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Kreis R.,University of Bern | And 3 more authors.
NMR in Biomedicine | Year: 2011

Standard methods for the estimation of the postmortem interval (PMI, time since death), based on the cooling of the corpse, are limited to about 48h after death. As an alternative, noninvasive postmortem observation of alterations of brain metabolites by means of 1H MRS has been suggested for an estimation of the PMI at room temperature, so far without including the effect of other ambient temperatures. In order to study the temperature effect, localized 1H MRS was used to follow brain decomposition in a sheep brain model at four different temperatures between 4 and 26°C with repeated measurements up to 2100h postmortem. The simultaneous determination of 25 different biochemical compounds at each measurement allowed the time courses of concentration changes to be followed. A sudden and almost simultaneous change of the concentrations of seven compounds was observed after a time span that decreased exponentially from 700h at 4°C to 30h at 26°C ambient temperature. As this represents, most probably, the onset of highly variable bacterial decomposition, and thus defines the upper limit for a reliable PMI estimation, data were analyzed only up to this start of bacterial decomposition. As 13 compounds showed unequivocal, reproducible concentration changes during this period while eight showed a linear increase with a slope that was unambiguously related to ambient temperature. Therefore, a single analytical function with PMI and temperature as variables can describe the time courses of metabolite concentrations. Using the inverse of this function, metabolite concentrations determined from a single MR spectrum can be used, together with known ambient temperatures, to calculate the PMI of a corpse. It is concluded that the effect of ambient temperature can be reliably included in the PMI determination by 1H MRS. © 2011 John Wiley & Sons, Ltd.

Stern D.,University of Graz | Stern D.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Ebner T.,University of Graz | Urschler M.,Ludwig Boltzmann Institute for Clinical Forensic Imaging
Proceedings - International Symposium on Biomedical Imaging | Year: 2016

Selection of set of training pixels and feature range show to be critical scale-related parameters with high impact on results in localization methods based on random regression forests (RRF). Trained on pixels randomly selected from images with long range features, RRF captures the variation in landmark location but often without reaching satisfying accuracy. Conversely, training an RRF with short range features in a landmark's close surroundings enables accurate localization, but at the cost of ambiguous localization results in the presence of locally similar structures. We present a scale-widening RRF method that effectively handles such ambiguities. On a challenging hand radiography image data set, we achieve median and 90th percentile localization errors of 0.81 and 2.64mm, respectively, outperforming related state-of-the-art methods. © 2016 IEEE.

Stern D.,University of Graz | Stern D.,Ludwig Boltzmann Institute for Clinical Forensic Imaging | Ursekter M.,University of Graz | Ursekter M.,Ludwig Boltzmann Institute for Clinical Forensic Imaging
Proceedings - International Symposium on Biomedical Imaging | Year: 2016

Increasingly important for both clinical and forensic medicine, radiological age estimation is performed by fusing independent bone age estimates from hand images. In this work, we show that the artificial separation into bone independent age estimates as used in established fusion techniques can be overcome. Thus, we treat aging as a global developmental process, by implicitly fusing developmental information from different bones in a dedicated regression algorithm. With 0.82 ± 0.56 years absolute deviation from chronological age on a database of 132 3D MR hand images, the results of this novel automatic algorithm are inline with radiologists performing visual examinations. © 2016 IEEE.

Loading Ludwig Boltzmann Institute for Clinical Forensic Imaging collaborators
Loading Ludwig Boltzmann Institute for Clinical Forensic Imaging collaborators