Pathology hinting as the combination of automatic segmentation with a statistical shape model

With improvements in acquisition speed and quality, the amount of medical image data to be screened by clinicians is starting to become challenging in the daily clinical practice. To quickly visualize and find abnormalities in medical images, we propose a new method combining segmentation algorithms with statistical shape models. A statistical shape model built from a healthy population will have a close fit in healthy regions. The model will however not fit to morphological abnormalities often present in the areas of pathologies. Using the residual fitting error of the statistical shape model, pathologies can be visualized very quickly. This idea is applied to finding drusen in the retinal pigment epithelium (RPE) of optical coherence tomography (OCT) volumes. A segmentation technique able to accurately segment drusen in patients with age-related macular degeneration (AMD) is applied. The segmentation is then analyzed with a statistical shape model to visualize potentially pathological areas. An extensive evaluation is performed to validate the segmentation algorithm, as well as the quality and sensitivity of the hinting system. Most of the drusen with a height of 85.5 microm were detected, and all drusen at least 93.6 microm high were detected.

Distance measures for image segmentation evaluation

The task considered in this paper is performance evaluation of region segmentation algorithms in the ground-truth-based paradigm. Given a machine segmentation and a ground-truth segmentation, performance measures are needed. We propose to consider the image segmentation problem as one of data clustering and, as a consequence, to use measures for comparing clusterings developed in statistics and machine learning. By doing so, we obtain a variety of performance measures which have not been used before in image processing. In particular, some of these measures have the highly desired property of being a metric. Experimental results are reported on both synthetic and real data to validate the measures and compare them with others.

The Multimodal Brain TumorImage Segmentation Benchmark (BRATS)

In this paper we report the set-up and results of the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) organized in conjunction with the MICCAI 2012 and 2013 conferences. Twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low- and high-grade glioma patients - manually annotated by up to four raters - and to 65 comparable scans generated using tumor image simulation software. Quantitative evaluations revealed considerable disagreement between the human raters in segmenting various tumor sub-regions (Dice scores in the range 74-85%), illustrating the difficulty of this task. We found that different algorithms worked best for different sub-regions (reaching performance comparable to human inter-rater variability), but that no single algorithm ranked in the top for all subregions simultaneously. Fusing several good algorithms using a hierarchical majority vote yielded segmentations that consistently ranked above all individual algorithms, indicating remaining opportunities for further methodological improvements. The BRATS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource.

Quantitative Analysis of Mouse Retinal Layers Using Automated Segmentation of Spectral Domain Optical Coherence Tomography Images.

PURPOSE Quantification of retinal layers using automated segmentation of optical coherence tomography (OCT) images allows for longitudinal studies of retinal and neurological disorders in mice. The purpose of this study was to compare the performance of automated retinal layer segmentation algorithms with data from manual segmentation in mice using the Spectralis OCT. METHODS Spectral domain OCT images from 55 mice from three different mouse strains were analyzed in total. The OCT scans from 22 C57Bl/6, 22 BALBc, and 11 C3A.Cg-Pde6b(+)Prph2(Rd2) /J mice were automatically segmented using three commercially available automated retinal segmentation algorithms and compared to manual segmentation. RESULTS Fully automated segmentation performed well in mice and showed coefficients of variation (CV) of below 5% for the total retinal volume. However, all three automated segmentation algorithms yielded much thicker total retinal thickness values compared to manual segmentation data (P < 0.0001) due to segmentation errors in the basement membrane. CONCLUSIONS Whereas the automated retinal segmentation algorithms performed well for the inner layers, the retinal pigmentation epithelium (RPE) was delineated within the sclera, leading to consistently thicker measurements of the photoreceptor layer and the total retina. TRANSLATIONAL RELEVANCE The introduction of spectral domain OCT allows for accurate imaging of the mouse retina. Exact quantification of retinal layer thicknesses in mice is important to study layers of interest under various pathological conditions.

The Virtual Skeleton Database: An Open Access Repository for Biomedical Research and Collaboration

Background: Statistical shape models are widely used in biomedical research. They are routinely implemented for automatic image segmentation or object identification in medical images. In these fields, however, the acquisition of the large training datasets, required to develop these models, is usually a time-consuming process. Even after this effort, the collections of datasets are often lost or mishandled resulting in replication of work. Objective: To solve these problems, the Virtual Skeleton Database (VSD) is proposed as a centralized storage system where the data necessary to build statistical shape models can be stored and shared. Methods: The VSD provides an online repository system tailored to the needs of the medical research community. The processing of the most common image file types, a statistical shape model framework, and an ontology-based search provide the generic tools to store, exchange, and retrieve digital medical datasets. The hosted data are accessible to the community, and collaborative research catalyzes their productivity. Results: To illustrate the need for an online repository for medical research, three exemplary projects of the VSD are presented: (1) an international collaboration to achieve improvement in cochlear surgery and implant optimization, (2) a population-based analysis of femoral fracture risk between genders, and (3) an online application developed for the evaluation and comparison of the segmentation of brain tumors. Conclusions: The VSD is a novel system for scientific collaboration for the medical image community with a data-centric concept and semantically driven search option for anatomical structures. The repository has been proven to be a useful tool for collaborative model building, as a resource for biomechanical population studies, or to enhance segmentation algorithms.

Integrated Segmentation of Brain Tumor Images for Radiotherapy and Neurosurgery

Image-based modeling of tumor growth combines methods from cancer simulation and medical imaging. In this context, we present a novel approach to adapt a healthy brain atlas to MR images of tumor patients. In order to establish correspondence between a healthy atlas and a pathologic patient image, tumor growth modeling in combination with registration algorithms is employed. In a first step, the tumor is grown in the atlas based on a new multi-scale, multi-physics model including growth simulation from the cellular level up to the biomechanical level, accounting for cell proliferation and tissue deformations. Large-scale deformations are handled with an Eulerian approach for finite element computations, which can operate directly on the image voxel mesh. Subsequently, dense correspondence between the modified atlas and patient image is established using nonrigid registration. The method offers opportunities in atlasbased segmentation of tumor-bearing brain images as well as for improved patient-specific simulation and prognosis of tumor progression.

3D morphometry using automated aortic segmentation in native MR angiography: an alternative to contrast enhanced MRA?

INTRODUCTION Native-MR angiography (N-MRA) is considered an imaging alternative to contrast enhanced MR angiography (CE-MRA) for patients with renal insufficiency. Lower intraluminal contrast in N-MRA often leads to failure of the segmentation process in commercial algorithms. This study introduces an in-house 3D model-based segmentation approach used to compare both sequences by automatic 3D lumen segmentation, allowing for evaluation of differences of aortic lumen diameters as well as differences in length comparing both acquisition techniques at every possible location. METHODS AND MATERIALS Sixteen healthy volunteers underwent 1.5-T-MR Angiography (MRA). For each volunteer, two different MR sequences were performed, CE-MRA: gradient echo Turbo FLASH sequence and N-MRA: respiratory-and-cardiac-gated, T2-weighted 3D SSFP. Datasets were segmented using a 3D model-based ellipse-fitting approach with a single seed point placed manually above the celiac trunk. The segmented volumes were manually cropped from left subclavian artery to celiac trunk to avoid error due to side branches. Diameters, volumes and centerline length were computed for intraindividual comparison. For statistical analysis the Wilcoxon-Signed-Ranked-Test was used. RESULTS Average centerline length obtained based on N-MRA was 239.0±23.4 mm compared to 238.6±23.5 mm for CE-MRA without significant difference (P=0.877). Average maximum diameter obtained based on N-MRA was 25.7±3.3 mm compared to 24.1±3.2 mm for CE-MRA (P<0.001). In agreement with the difference in diameters, volumes obtained based on N-MRA (100.1±35.4 cm(3)) were consistently and significantly larger compared to CE-MRA (89.2±30.0 cm(3)) (P<0.001). CONCLUSIONS 3D morphometry shows highly similar centerline lengths for N-MRA and CE-MRA, but systematically higher diameters and volumes for N-MRA.

Volumetric analysis of lung nodules in computed tomography (CT): comparison of two different segmentation algorithm softwares and two different reconstruction filters on automated volume calculation.

BACKGROUND A precise detection of volume change allows for better estimating the biological behavior of the lung nodules. Postprocessing tools with automated detection, segmentation, and volumetric analysis of lung nodules may expedite radiological processes and give additional confidence to the radiologists. PURPOSE To compare two different postprocessing software algorithms (LMS Lung, Median Technologies; LungCARE®, Siemens) in CT volumetric measurement and to analyze the effect of soft (B30) and hard reconstruction filter (B70) on automated volume measurement. MATERIAL AND METHODS Between January 2010 and April 2010, 45 patients with a total of 113 pulmonary nodules were included. The CT exam was performed on a 64-row multidetector CT scanner (Somatom Sensation, Siemens, Erlangen, Germany) with the following parameters: collimation, 24x1.2 mm; pitch, 1.15; voltage, 120 kVp; reference tube current-time, 100 mAs. Automated volumetric measurement of each lung nodule was performed with the two different postprocessing algorithms based on two reconstruction filters (B30 and B70). The average relative volume measurement difference (VME%) and the limits of agreement between two methods were used for comparison. RESULTS At soft reconstruction filters the LMS system produced mean nodule volumes that were 34.1% (P < 0.0001) larger than those by LungCARE® system. The VME% was 42.2% with a limit of agreement between -53.9% and 138.4%.The volume measurement with soft filters (B30) was significantly larger than with hard filters (B70); 11.2% for LMS and 1.6% for LungCARE®, respectively (both with P < 0.05). LMS measured greater volumes with both filters, 13.6% for soft and 3.8% for hard filters, respectively (P < 0.01 and P > 0.05). CONCLUSION There is a substantial inter-software (LMS/LungCARE®) as well as intra-software variability (B30/B70) in lung nodule volume measurement; therefore, it is mandatory to use the same equipment with the same reconstruction filter for the follow-up of lung nodule volume.

Shallow Dialogue Processing Using Machine Learning Algorithms (or not)

This paper presents a shallow dialogue analysis model, aimed at human-human dialogues in the context of staff or business meetings. Four components of the model are defined, and several machine learning techniques are used to extract features from dialogue transcripts: maximum entropy classifiers for dialogue acts, latent semantic analysis for topic segmentation, or decision tree classifiers for discourse markers. A rule-based approach is proposed for solving cross-modal references to meeting documents. The methods are trained and evaluated thanks to a common data set and annotation format. The integration of the components into an automated shallow dialogue parser opens the way to multimodal meeting processing and retrieval applications.