Structured output SVM for remote sensing image classification

In the recent years, kernel methods have revealed very powerful tools in many application domains in general and in remote sensing image classification in particular. The special characteristics of remote sensing images (high dimension, few labeled samples and different noise sources) are efficiently dealt with kernel machines. In this paper, we propose the use of structured output learning to improve remote sensing image classification based on kernels. Structured output learning is concerned with the design of machine learning algorithms that not only implement input-output mapping, but also take into account the relations between output labels, thus generalizing unstructured kernel methods. We analyze the framework and introduce it to the remote sensing community. Output similarity is here encoded into SVM classifiers by modifying the model loss function and the kernel function either independently or jointly. Experiments on a very high resolution (VHR) image classification problem shows promising results and opens a wide field of research with structured output kernel methods.

MP2RAGE, a self bias-field corrected sequence for improved segmentation and T-1-mapping at high field

The large spatial inhomogeneity in transmit B, field (B-1(+)) observable in human MR images at hi h static magnetic fields (B-0) severely impairs image quality. To overcome this effect in brain T-1-weighted images the, MPRAGE sequence was modified to generate two different images at different inversion times MP2RAGE By combining the two images in a novel fashion, it was possible to create T-1-weigthed images where the result image was free of proton density contrast, T-2* contrast, reception bias field, and, to first order transmit field inhomogeneity. MP2RAGE sequence parameters were optimized using Bloch equations to maximize contrast-to-noise ratio per unit of time between brain tissues and minimize the effect of B-1(+) variations through space. Images of high anatomical quality and excellent brain tissue differentiation suitable for applications such as segmentation and voxel-based morphometry were obtained at 3 and 7 T. From such T-1-weighted images, acquired within 12 min, high-resolution 3D T-1 maps were routinely calculated at 7 T with sub-millimeter voxel resolution (0.65-0.85 mm isotropic). T-1 maps were validated in phantom experiments. In humans, the T, values obtained at 7 T were 1.15 +/- 0.06 s for white matter (WM) and 1.92 +/- 0.16 s for grey matter (GM), in good agreement with literature values obtained at lower spatial resolution. At 3 T, where whole-brain acquisitions with 1 mm isotropic voxels were acquired in 8 min the T-1 values obtained (0.81 +/- 0.03 S for WM and 1.35 +/- 0.05 for GM) were once again found to be in very good agreement with values in the literature. (C) 2009 Elsevier Inc. All rights reserved.

Targetting of the ventro-intermediate nucleus using ultra-high field (7T) MRI for gamma knife surgery purposes: A pilot in vivo study on healthy subjects.

INTRODUCTION: Gamma Knife surgery (GKS) is a non-invasive neurosurgical stereotactic procedure, increasingly used as an alternative to open functional procedures. This includes targeting of the ventro-intermediate nucleus of the thalamus (e.g. Vim) for tremor. We currently perform an indirect targeting, as the Vim is not visible on current 3Tesla MRI acquisitions. Our objective was to enhance anatomic imaging (aiming at refining the precision of anatomic target selection by direct visualisation) in patients treated for tremor with Vim GKS, by using high field 7T MRI. MATERIALS AND METHODSH: Five young healthy subjects were scanned on 3 (T1-w and diffusion tensor imaging) and 7T (high-resolution susceptibility weighted images (SWI)) MRI in Lausanne. All images were further integrated for the first time into the Gamma Plan Software(®) (Elekta Instruments, AB, Sweden) and co-registered (with T1 was a reference). A simulation of targeting of the Vim was done using various methods on the 3T images. Furthermore, a correlation with the position of the found target with the 7T SWI was performed. The atlas of Morel et al. (Zurich, CH) was used to confirm the findings on a detailed analysis inside/outside the Gamma Plan. RESULTS: The use of SWI provided us with a superior resolution and an improved image contrast within the basal ganglia. This allowed visualization and direct delineation of some subgroups of thalamic nuclei in vivo, including the Vim. The position of the target, as assessed on 3T, perfectly matched with the supposed one of the Vim on the SWI. Furthermore, a 3-dimensional model of the Vim-target area was created on the basis of the obtained images. CONCLUSION: This is the first report of the integration of SWI high field MRI into the LGP, aiming at the improvement of targeting validation of the Vim in tremor. The anatomical correlation between the direct visualization on 7T and the current targeting methods on 3T (e.g. quadrilatere of Guyot, histological atlases) seems to show a very good anatomical matching. Further studies are needed to validate this technique, both by improving the accuracy of the targeting of the Vim (potentially also other thalamic nuclei) and to perform clinical assessment.

The feasibility of 350 μm spatial resolution coronary magnetic resonance angiography at 3 T in humans.

PURPOSE: The purposes of this study were to (1) develop a high-resolution 3-T magnetic resonance angiography (MRA) technique with an in-plane resolution approximate to that of multidetector coronary computed tomography (MDCT) and a voxel size of 0.35 × 0.35 × 1.5 mm³ and to (2) investigate the image quality of this technique in healthy participants and preliminarily in patients with known coronary artery disease (CAD). MATERIALS AND METHODS: A 3-T coronary MRA technique optimized for an image acquisition voxel as small as 0.35 × 0.35 × 1.5 mm³ (high-resolution coronary MRA [HRC]) was implemented and the coronary arteries of 22 participants were imaged. These included 11 healthy participants (average age, 28.5 years; 5 men) and 11 participants with CAD (average age, 52.9 years; 5 women) as identified on MDCT. In addition, the 11 healthy participants were imaged using a method with a more common spatial resolution of 0.7 × 1 × 3 mm³ (regular-resolution coronary MRA [RRC]). Qualitative and quantitative comparisons were made between the 2 MRA techniques. RESULTS: Normal vessels and CAD lesions were successfully depicted at 350 × 350 μm² in-plane resolution with adequate signal-to-noise ratio (SNR) and contrast-to-noise ratio. The CAD findings were consistent among MDCT and HRC. The HRC showed a 47% improvement in sharpness despite a reduction in SNR (by 72%) and in contrast-to-noise ratio (by 86%) compared with the regular-resolution coronary MRA. CONCLUSION: This study, as a first step toward substantial improvement in the resolution of coronary MRA, demonstrates the feasibility of obtaining at 3 T a spatial resolution that approximates that of MDCT. The acquisition in-plane pixel dimensions are as small as 350 × 350 μm² with a 1.5-mm slice thickness. Although SNR is lower, the images have improved sharpness, resulting in image quality that allows qualitative identification of disease sites on MRA consistent with MDCT.

MP2RAGE, a self bias-field corrected sequence for improved segmentation and T1-mapping at high field.

The large spatial inhomogeneity in transmit B(1) field (B(1)(+)) observable in human MR images at high static magnetic fields (B(0)) severely impairs image quality. To overcome this effect in brain T(1)-weighted images, the MPRAGE sequence was modified to generate two different images at different inversion times, MP2RAGE. By combining the two images in a novel fashion, it was possible to create T(1)-weighted images where the result image was free of proton density contrast, T(2) contrast, reception bias field, and, to first order, transmit field inhomogeneity. MP2RAGE sequence parameters were optimized using Bloch equations to maximize contrast-to-noise ratio per unit of time between brain tissues and minimize the effect of B(1)(+) variations through space. Images of high anatomical quality and excellent brain tissue differentiation suitable for applications such as segmentation and voxel-based morphometry were obtained at 3 and 7 T. From such T(1)-weighted images, acquired within 12 min, high-resolution 3D T(1) maps were routinely calculated at 7 T with sub-millimeter voxel resolution (0.65-0.85 mm isotropic). T(1) maps were validated in phantom experiments. In humans, the T(1) values obtained at 7 T were 1.15+/-0.06 s for white matter (WM) and 1.92+/-0.16 s for grey matter (GM), in good agreement with literature values obtained at lower spatial resolution. At 3 T, where whole-brain acquisitions with 1 mm isotropic voxels were acquired in 8 min, the T(1) values obtained (0.81+/-0.03 s for WM and 1.35+/-0.05 for GM) were once again found to be in very good agreement with values in the literature.

Active learning methods for remote sensing image classification

In this paper, we propose two active learning algorithms for semiautomatic definition of training samples in remote sensing image classification. Based on predefined heuristics, the classifier ranks the unlabeled pixels and automatically chooses those that are considered the most valuable for its improvement. Once the pixels have been selected, the analyst labels them manually and the process is iterated. Starting with a small and nonoptimal training set, the model itself builds the optimal set of samples which minimizes the classification error. We have applied the proposed algorithms to a variety of remote sensing data, including very high resolution and hyperspectral images, using support vector machines. Experimental results confirm the consistency of the methods. The required number of training samples can be reduced to 10% using the methods proposed, reaching the same level of accuracy as larger data sets. A comparison with a state-of-the-art active learning method, margin sampling, is provided, highlighting advantages of the methods proposed. The effect of spatial resolution and separability of the classes on the quality of the selection of pixels is also discussed.

Silver-assisted laser desorption ionization for high spatial resolution imaging mass spectrometry of olefins from thin tissue sections.

Silver has been demonstrated to be a powerful cationization agent in mass spectrometry (MS) for various olefinic species such as cholesterol and fatty acids. This work explores the utility of metallic silver sputtering on tissue sections for high resolution imaging mass spectrometry (IMS) of olefins by laser desorption ionization (LDI). For this purpose, sputtered silver coating thickness was optimized on an assorted selection of mouse and rat tissues including brain, kidney, liver, and testis. For mouse brain tissue section, the thickness was adjusted to 23 ± 2 nm of silver to prevent ion suppression effects associated with a higher cholesterol and lipid content. On all other tissues, a thickness of at 16 ± 2 nm provided the best desorption/ionization efficiency. Characterization of the species by MS/MS showed a wide variety of olefinic compounds allowing the IMS of different lipid classes including cholesterol, arachidonic acid, docosahexaenoic acid, and triacylglyceride 52:3. A range of spatial resolutions for IMS were investigated from 150 μm down to the high resolution cellular range at 5 μm. The applicability of direct on-tissue silver sputtering to LDI-IMS of cholesterol and other olefinic compounds presents a novel approach to improve the amount of information that can be obtained from tissue sections. This IMS strategy is thus of interest for providing new biological insights on the role of cholesterol and other olefins in physiological pathways or disease.

Side scan sonar image and geologic interpretation of the Ria de Pontevedra seafloor (Galicia, NW Spain).

High-resolution side scan sonar has been used for mapping the seafloor of the Ría de Pontevedra. Four backscatter patterns have been mapped within the Ría: (1) Pattern with isolated reflections, correlated with granite and metamorphic outcrops and located close to the coastal prominence and Ons and Onza Islands. (2) Pattern of strong reflectivity usually located around the basement outcrops and near the coastline and produced by coarse-grained sediment. (3) Pattern of weak backscatter is correlated with fine sand to mud and comprising large areas in the central and deep part of the Ría, where the bottom currents are weak. It is generally featureless, except where pockmarks and anthropogenic features are present. (4) Patches of strong and weak backscatter are located in the boundary between coarse and fine-grained sediments and they are due to the effect of strong bottom currents. The presence of megaripples associated to both patterns of strong reflectivity and sedimentary patches indicate bedload transport of sediment during high energy conditions (storms). Side scan sonar records and supplementary bathymetry, bottom samples and hydrodynamic data reveal that the distribution of seafloor sediment is strongly related to oceanographic processes and the particular morphology and topography of the Ría.

Ultra-High Field (7 Tesla) MRI for Gamma Knife Surgery Targeting of the Ventro-Intermediate Nucleus: A Pilot in Vivo Study

Aim: Gamma Knife surgery (GKS) is a non-invasive neurosurgical stereotactic procedure, increasingly used as an alternative to open functional procedures. This includes the targeting of the ventro-intermediate (Vim) nucleus of the thalamus for tremor. We currently perform an indirect targeting, using the "quadrilatere of Guyot," as the Vim nucleus is not visible on current 3 Tesla (T) MRI acquisitions. The primary objective of the current study was to enhance anatomic imaging for Vim GKS using high-field (7 T) MRI, with the aim of refining the visualization and precision of anatomical targeting. Method: Five young healthy subjects (mean age 23 years) were scanned both on 3 and 7 T MRI in Lausanne University Hospital (CHUV) and Center for Biomedical Imaging (CIBM). Classical T1-weighted MPRAGE, T2 CISS sequences (replacing former ventriculography) and diffusion tensor imaging were acquired at 3T. We obtained high-resolution susceptibility weighted images (SWI) at 7T for the visualization of thalamic subparts. SWI was further integrated for the first time into Leksell Gamma Plan® (LGP) software and co-registered with the 3T images. A simulation of targeting of the Vim was done using the "quadrilatere of Guyot" methodology on the 3T images. Furthermore, a correlation with the position of the found target on SWI was performed. The atlas of Morel et al. was used to confirm the findings on a detailed computer analysis outside LGP. Also, 3T and 7T MRI of one patient undergoing GKS Vim thalamotomy, were obtained before and 2 years after the procedure, and studied similarly. Results: The use of SWI provided a superior resolution and improved image contrast within the central gray matter. This allowed visualization and direct delineation of groups of thalamic nuclei in vivo, including the Vim. The position of the target, as assessed with the "quadrilatere of Guyot" method on 3 T, perfectly matched with the supposed one of the Vim on the SWI. Furthermore, a 3-dimensional model of the Vim target area was created on the basis of 3T and 7T images. Conclusion: This is the first report of the integration of SWI high-field MRI into the LGP in healthy subjects and in one patient treated GKS Vim thalamotomy. This approach aims at the improvement of targeting validation and further direct targeting of the Vim in tremor. The anatomical correlation between the direct visualization on 7T and the current targeting methods on 3T seems to show a very good anatomical matching.

A post-processing pipeline to reconstruct the developing fetal brain using low-resolution MRI

Introduction. Development of the fetal brain surfacewith concomitant gyrification is one of the majormaturational processes of the human brain. Firstdelineated by postmortem studies or by ultrasound, MRIhas recently become a powerful tool for studying in vivothe structural correlates of brain maturation. However,the quantitative measurement of fetal brain developmentis a major challenge because of the movement of the fetusinside the amniotic cavity, the poor spatial resolution,the partial volume effect and the changing appearance ofthe developing brain. Today extensive efforts are made todeal with the âeurooepost-acquisitionâeuro reconstruction ofhigh-resolution 3D fetal volumes based on severalacquisitions with lower resolution (Rousseau, F., 2006;Jiang, S., 2007). We here propose a framework devoted tothe segmentation of the basal ganglia, the gray-whitetissue segmentation, and in turn the 3D corticalreconstruction of the fetal brain. Method. Prenatal MRimaging was performed with a 1-T system (GE MedicalSystems, Milwaukee) using single shot fast spin echo(ssFSE) sequences in fetuses aged from 29 to 32gestational weeks (slice thickness 5.4mm, in planespatial resolution 1.09mm). For each fetus, 6 axialvolumes shifted by 1 mm were acquired (about 1 min pervolume). First, each volume is manually segmented toextract fetal brain from surrounding fetal and maternaltissues. Inhomogeneity intensity correction and linearintensity normalization are then performed. A highspatial resolution image of isotropic voxel size of 1.09mm is created for each fetus as previously published byothers (Rousseau, F., 2006). B-splines are used for thescattered data interpolation (Lee, 1997). Then, basalganglia segmentation is performed on this superreconstructed volume using active contour framework witha Level Set implementation (Bach Cuadra, M., 2010). Oncebasal ganglia are removed from the image, brain tissuesegmentation is performed (Bach Cuadra, M., 2009). Theresulting white matter image is then binarized andfurther given as an input in the Freesurfer software(http://surfer.nmr.mgh.harvard.edu/) to provide accuratethree-dimensional reconstructions of the fetal brain.Results. High-resolution images of the cerebral fetalbrain, as obtained from the low-resolution acquired MRI,are presented for 4 subjects of age ranging from 29 to 32GA. An example is depicted in Figure 1. Accuracy in theautomated basal ganglia segmentation is compared withmanual segmentation using measurement of Dice similarity(DSI), with values above 0.7 considering to be a verygood agreement. In our sample we observed DSI valuesbetween 0.785 and 0.856. We further show the results ofgray-white matter segmentation overlaid on thehigh-resolution gray-scale images. The results arevisually checked for accuracy using the same principlesas commonly accepted in adult neuroimaging. Preliminary3D cortical reconstructions of the fetal brain are shownin Figure 2. Conclusion. We hereby present a completepipeline for the automated extraction of accuratethree-dimensional cortical surface of the fetal brain.These results are preliminary but promising, with theultimate goal to provide âeurooemovieâeuro of the normal gyraldevelopment. In turn, a precise knowledge of the normalfetal brain development will allow the quantification ofsubtle and early but clinically relevant deviations.Moreover, a precise understanding of the gyraldevelopment process may help to build hypotheses tounderstand the pathogenesis of several neurodevelopmentalconditions in which gyrification have been shown to bealtered (e.g. schizophrenia, autismâeuro¦). References.Rousseau, F. (2006), 'Registration-Based Approach forReconstruction of High-Resolution In Utero Fetal MR Brainimages', IEEE Transactions on Medical Imaging, vol. 13,no. 9, pp. 1072-1081. Jiang, S. (2007), 'MRI of MovingSubjects Using Multislice Snapshot Images With VolumeReconstruction (SVR): Application to Fetal, Neonatal, andAdult Brain Studies', IEEE Transactions on MedicalImaging, vol. 26, no. 7, pp. 967-980. Lee, S. (1997),'Scattered data interpolation with multilevel B-splines',IEEE Transactions on Visualization and Computer Graphics,vol. 3, no. 3, pp. 228-244. Bach Cuadra, M. (2010),'Central and Cortical Gray Mater Segmentation of MagneticResonance Images of the Fetal Brain', ISMRM Conference.Bach Cuadra, M. (2009), 'Brain tissue segmentation offetal MR images', MICCAI.

New Developments and Applications of the MP2RAGE Sequence - Focusing the Contrast and High Spatial Resolution R1 Mapping.

MR structural T1-weighted imaging using high field systems (>3T) is severely hampered by the existing large transmit field inhomogeneities. New sequences have been developed to better cope with such nuisances. In this work we show the potential of a recently proposed sequence, the MP2RAGE, to obtain improved grey white matter contrast with respect to conventional T1-w protocols, allowing for a better visualization of thalamic nuclei and different white matter bundles in the brain stem. Furthermore, the possibility to obtain high spatial resolution (0.65 mm isotropic) R1 maps fully independent of the transmit field inhomogeneities in clinical acceptable time is demonstrated. In this high resolution R1 maps it was possible to clearly observe varying properties of cortical grey matter throughout the cortex and observe different hippocampus fields with variations of intensity that correlate with known myelin concentration variations.

Sublimation of new matrix candidates for high spatial resolution imaging mass spectrometry of lipids: enhanced information in both positive and negative polarities after 1,5-diaminonapthalene deposition.

Matrix sublimation has demonstrated to be a powerful approach for high-resolution matrix-assisted laser desorption ionization (MALDI) imaging of lipids, providing very homogeneous solvent-free deposition. This work presents a comprehensive study aiming to evaluate current and novel matrix candidates for high spatial resolution MALDI imaging mass spectrometry of lipids from tissue section after deposition by sublimation. For this purpose, 12 matrices including 2,5-dihydroxybenzoic acid (DHB), sinapinic acid (SA), α-cyano-4-hydroxycinnamic acid (CHCA), 2,6-dihydroxyacetphenone (DHA), 2',4',6'-trihydroxyacetophenone (THAP), 3-hydroxypicolinic acid (3-HPA), 1,8-bis(dimethylamino)naphthalene (DMAN), 1,8,9-anthracentriol (DIT), 1,5-diaminonapthalene (DAN), p-nitroaniline (NIT), 9-aminoacridine (9-AA), and 2-mercaptobenzothiazole (MBT) were investigated for lipid detection efficiency in both positive and negative ionization modes, matrix interferences, and stability under vacuum. For the most relevant matrices, ion maps of the different lipid species were obtained from tissue sections at high spatial resolution and the detected peaks were characterized by matrix-assisted laser desorption ionization time-of-flight/time-of-flight (MALDI-TOF/TOF) mass spectrometry. First proposed for imaging mass spectrometry (IMS) after sublimation, DAN has demonstrated to be of high efficiency providing rich lipid signatures in both positive and negative polarities with high vacuum stability and sub-20 μm resolution capacity. Ion images from adult mouse brain were generated with a 10 μm scanning resolution. Furthermore, ion images from adult mouse brain and whole-body fish tissue sections were also acquired in both polarity modes from the same tissue section at 100 μm spatial resolution. Sublimation of DAN represents an interesting approach to improve information with respect to currently employed matrices providing a deeper analysis of the lipidome by IMS.

Classification of very high spatial resolution imagery using mathematical morphology and support vector machines

We investigate the relevance of morphological operators for the classification of land use in urban scenes using submetric panchromatic imagery. A support vector machine is used for the classification. Six types of filters have been employed: opening and closing, opening and closing by reconstruction, and opening and closing top hat. The type and scale of the filters are discussed, and a feature selection algorithm called recursive feature elimination is applied to decrease the dimensionality of the input data. The analysis performed on two QuickBird panchromatic images showed that simple opening and closing operators are the most relevant for classification at such a high spatial resolution. Moreover, mixed sets combining simple and reconstruction filters provided the best performance. Tests performed on both images, having areas characterized by different architectural styles, yielded similar results for both feature selection and classification accuracy, suggesting the generalization of the feature sets highlighted.

Sparse image reconstruction on the sphere: implications of a new sampling theorem.

We study the impact of sampling theorems on the fidelity of sparse image reconstruction on the sphere. We discuss how a reduction in the number of samples required to represent all information content of a band-limited signal acts to improve the fidelity of sparse image reconstruction, through both the dimensionality and sparsity of signals. To demonstrate this result, we consider a simple inpainting problem on the sphere and consider images sparse in the magnitude of their gradient. We develop a framework for total variation inpainting on the sphere, including fast methods to render the inpainting problem computationally feasible at high resolution. Recently a new sampling theorem on the sphere was developed, reducing the required number of samples by a factor of two for equiangular sampling schemes. Through numerical simulations, we verify the enhanced fidelity of sparse image reconstruction due to the more efficient sampling of the sphere provided by the new sampling theorem.

An efficient total variation algorithm for super-resolution in fetal brain MRI with adaptive regularization.

Although fetal anatomy can be adequately viewed in new multi-slice MR images, many critical limitations remain for quantitative data analysis. To this end, several research groups have recently developed advanced image processing methods, often denoted by super-resolution (SR) techniques, to reconstruct from a set of clinical low-resolution (LR) images, a high-resolution (HR) motion-free volume. It is usually modeled as an inverse problem where the regularization term plays a central role in the reconstruction quality. Literature has been quite attracted by Total Variation energies because of their ability in edge preserving but only standard explicit steepest gradient techniques have been applied for optimization. In a preliminary work, it has been shown that novel fast convex optimization techniques could be successfully applied to design an efficient Total Variation optimization algorithm for the super-resolution problem. In this work, two major contributions are presented. Firstly, we will briefly review the Bayesian and Variational dual formulations of current state-of-the-art methods dedicated to fetal MRI reconstruction. Secondly, we present an extensive quantitative evaluation of our SR algorithm previously introduced on both simulated fetal and real clinical data (with both normal and pathological subjects). Specifically, we study the robustness of regularization terms in front of residual registration errors and we also present a novel strategy for automatically select the weight of the regularization as regards the data fidelity term. Our results show that our TV implementation is highly robust in front of motion artifacts and that it offers the best trade-off between speed and accuracy for fetal MRI recovery as in comparison with state-of-the art methods.