We propose a deep mixture of multimodal hierarchical variational auto-encoders called MMHVAE that synthesizes missing images from observed images in different modalities. MMHVAE's design focuses on tackling four challenges: (i) creating a complex latent representation of multimodal data to generate high-resolution images; (ii) encouraging the variational distributions to estimate the missing information needed for cross-modal image synthesis; (iii) learning to fuse multimodal information in the context of missing data; (iv) leveraging dataset-level information to handle incomplete data sets at training time. Extensive experiments are performed on the challenging problem of pre-operative brain multi-parametric magnetic resonance and intra-operative ultrasound imaging.

We present a novel approach to calibrating linguistic expressions of certainty, e.g., "Maybe" and "Likely". Unlike prior work that assigns a single score to each certainty phrase, we model uncertainty as distributions over the simplex to capture their semantics more accurately. To accommodate this new representation of certainty, we generalize existing measures of miscalibration and introduce a novel post-hoc calibration method. Leveraging these tools, we analyze the calibration of both humans (e.g., radiologists) and computational models (e.g., language models) and provide interpretable suggestions to improve their calibration.

We introduce MHVAE, a deep hierarchical variational autoencoder (VAE) that synthesizes missing images from various modalities. Extending multi-modal VAEs with a hierarchical latent structure, we introduce a probabilistic formulation for fusing multi-modal images in a common latent representation while having the flexibility to handle incomplete image sets as input. Moreover, adversarial learning is employed to generate sharper images. Extensive experiments are performed on the challenging problem of joint intra-operative ultrasound (iUS) and Magnetic Resonance (MR) synthesis. Our model outperformed multi-modal VAEs, conditional GANs, and the current state-of-the-art unified method (ResViT) for synthesizing missing images, demonstrating the advantage of using a hierarchical latent representation and a principled probabilistic fusion operation. Our code is publicly available.

Lung ultrasound (LUS) is an important imaging modality used by emergency physicians to assess pulmonary congestion at the patient bedside. B-line artifacts in LUS videos are key findings associated with pulmonary congestion. Not only can the interpretation of LUS be challenging for novice operators, but visual quantification of B-lines remains subject to observer variability. In this work, we investigate the strengths and weaknesses of multiple deep learning approaches for automated B-line detection and localization in LUS videos. We curate and publish,BEDLUS, a new ultrasound dataset comprising 1,419 videos from 113 patients with a total of 15,755 expert-annotated B-lines. Based on this dataset, we present a benchmark of established deep learning methods applied to the task of B-line detection. To pave the way for interpretable quantification of B-lines, we propose a novel “single-point” approach to B-line localization using only the point of origin. Our results show that (a) the area under the receiver operating characteristic curve ranges from 0.864 to 0.955 for the benchmarked detection methods, (b) within this range, the best performance is achieved by models that leverage multiple successive frames as input, and (c) the proposed single-point approach for B-line localization reaches an F$_{1}$-score of 0.65, performing on par with the inter-observer agreement. The dataset and developed methods can facilitate further biomedical research on automated interpretation of lung ultrasound with the potential to expand the clinical utility.

MOTIVATION: Mass spectrometry imaging (MSI) provides rich biochemical information in a label-free manner and therefore holds promise to substantially impact current practice in disease diagnosis. However, the complex nature of MSI data poses computational challenges in its analysis. The complexity of the data arises from its large size, high-dimensionality and spectral nonlinearity. Preprocessing, including peak picking, has been used to reduce raw data complexity; however, peak picking is sensitive to parameter selection that, perhaps prematurely, shapes the downstream analysis for tissue classification and ensuing biological interpretation. RESULTS: We propose a deep learning model, massNet, that provides the desired qualities of scalability, nonlinearity and speed in MSI data analysis. This deep learning model was used, without prior preprocessing and peak picking, to classify MSI data from a mouse brain harboring a patient-derived tumor. The massNet architecture established automatically learning of predictive features, and automated methods were incorporated to identify peaks with potential for tumor delineation. The model's performance was assessed using cross-validation, and the results demonstrate higher accuracy and a substantial gain in speed compared to the established classical machine learning method, support vector machine. AVAILABILITY AND IMPLEMENTATION: https://github.com/wabdelmoula/massNet. The data underlying this article are available in the NIH Common Fund's National Metabolomics Data Repository (NMDR) Metabolomics Workbench under project id (PR001292) with http://dx.doi.org/10.21228/M8Q70T. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

We propose a novel pairwise distance measure between image keypoint sets, for the purpose of large-scale medical image indexing. Our measure generalizes the Jaccard index to account for soft set equivalence (SSE) between keypoint elements, via an adaptive kernel framework modeling uncertainty in keypoint appearance and geometry. A new kernel is proposed to quantify the variability of keypoint geometry in location and scale. Our distance measure may be estimated between${O}\,{(}{N}^{{\,{2}}}{)}$image pairs in${O}\,{(}{N}\,\text {log}{N}\,{)}$operations via keypoint indexing. Experiments report the first results for the task of predicting family relationships from medical images, using 1010 T1-weighted MRI brain volumes of 434 families including monozygotic and dizygotic twins, siblings and half-siblings sharing 100%-25% of their polymorphic genes. Soft set equivalence and the keypoint geometry kernel improve upon standard hard set equivalence (HSE) and appearance kernels alone in predicting family relationships. Monozygotic twin identification is near 100%, and three subjects with uncertain genotyping are automatically paired with their self-reported families, the first reported practical application of image-based family identification. Our distance measure can also be used to predict group categories, sex is predicted with an AUC = 0.97. Software is provided for efficient fine-grained curation of large, generic image datasets.

In this paper, we present a deep learning method, DDMReg, for accurate registration between diffusion MRI (dMRI) datasets. In dMRI registration, the goal is to spatially align brain anatomical structures while ensuring that local fiber orientations remain consistent with the underlying white matter fiber tract anatomy. DDMReg is a novel method that uses joint whole-brain and tract-specific information for dMRI registration. Based on the successful VoxelMorph framework for image registration, we propose a novel registration architecture that leverages not only whole brain information but also tract-specific fiber orientation information. DDMReg is an unsupervised method for deformable registration between pairs of dMRI datasets: it does not require nonlinearly pre-registered training data or the corresponding deformation fields as ground truth. We perform comparisons with four state-of-the-art registration methods on multiple independently acquired datasets from different populations (including teenagers, young and elderly adults) and different imaging protocols and scanners. We evaluate the registration performance by assessing the ability to align anatomically corresponding brain structures and ensure fiber spatial agreement between different subjects after registration. Experimental results show that DDMReg obtains significantly improved registration performance compared to the state-of-the-art methods. Importantly, we demonstrate successful generalization of DDMReg to dMRI data from different populations with varying ages and acquired using different acquisition protocols and different scanners.

We propose and demonstrate a representation learning approach by maximizing the mutual information between local features of images and text. The goal of this approach is to learn useful image representations by taking advantage of the rich information contained in the free text that describes the findings in the image. Our method trains image and text encoders by encouraging the resulting representations to exhibit high local mutual information. We make use of recent advances in mutual information estimation with neural network discriminators. We argue that the sum of local mutual information is typically a lower bound on the global mutual information. Our experimental results in the downstream image classification tasks demonstrate the advantages of using local features for image-text representation learning.

Ensembling is now recognized as an effective approach for increasing the predictive performance and calibration of deep networks. We introduce a new approach, Parameter Ensembling by Perturbation (PEP), that constructs an ensemble of parameter values as random perturbations of the optimal parameter set from training by a Gaussian with a single variance parameter. The variance is chosen to maximize the log-likelihood of the ensemble average (𝕃) on the validation data set. Empirically, and perhaps surprisingly, 𝕃 has a well-defined maximum as the variance grows from zero (which corresponds to the baseline model). Conveniently, calibration level of predictions also tends to grow favorably until the peak of 𝕃 is reached. In most experiments, PEP provides a small improvement in performance, and, in some cases, a substantial improvement in empirical calibration. We show that this "PEP effect'' (the gain in log-likelihood) is related to the mean curvature of the likelihood function and the empirical Fisher information. Experiments on ImageNet pre-trained networks including ResNet, DenseNet, and Inception showed improved calibration and likelihood. We further observed a mild improvement in classification accuracy on these networks. Experiments on classification benchmarks such as MNIST and CIFAR-10 showed improved calibration and likelihood, as well as the relationship between the PEP effect and overfitting; this demonstrates that PEP can be used to probe the level of overfitting that occurred during training. In general, no special training procedure or network architecture is needed, and in the case of pre-trained networks, no additional training is needed.

Fully convolutional neural networks (FCNs), and in particular U-Nets, have achieved state-of-the-art results in semantic segmentation for numerous medical imaging applications. Moreover, batch normalization and Dice loss have been used successfully to stabilize and accelerate training. However, these networks are poorly calibrated i.e. they tend to produce overconfident predictions for both correct and erroneous classifications, making them unreliable and hard to interpret. In this paper, we study predictive uncertainty estimation in FCNs for medical image segmentation. We make the following contributions: 1) We systematically compare cross-entropy loss with Dice loss in terms of segmentation quality and uncertainty estimation of FCNs; 2) We propose model ensembling for confidence calibration of the FCNs trained with batch normalization and Dice loss; 3) We assess the ability of calibrated FCNs to predict segmentation quality of structures and detect out-of-distribution test examples. We conduct extensive experiments across three medical image segmentation applications of the brain, the heart, and the prostate to evaluate our contributions. The results of this study offer considerable insight into the predictive uncertainty estimation and out-of-distribution detection in medical image segmentation and provide practical recipes for confidence calibration. Moreover, we consistently demonstrate that model ensembling improves confidence calibration.

We introduce an approach for image segmentation based on sparse correspondences between keypoints in testing and training images. Keypoints represent automatically identified distinctive image locations, where each keypoint correspondence suggests a transformation between images. We use these correspondences to transfer the label maps of entire organs from the training images to the test image. The keypoint transfer algorithm includes three steps: 1) keypoint matching; 2) voting-based keypoint labeling; and 3) keypoint-based probabilistic transfer of organ segmentations. We report segmentation results for abdominal organs in whole-body CT and MRI, as well as in contrast-enhanced CT and MRI. Our method offers a speed-up of about three orders of magnitude in comparison with common multi-atlas segmentation while achieving an accuracy that compares favorably. Moreover, keypoint transfer does not require the registration to an atlas or a training phase. Finally, the method allows for the segmentation of scans with a highly variable field-of-view.

Image guidance improves tissue sampling during biopsy by allowing the physician to visualize the tip and trajectory of the biopsy needle relative to the target in MRI, CT, ultrasound, or other relevant imagery. This paper reports a system for fast automatic needle tip and trajectory localization and visualization in MRI that has been developed and tested in the context of an active clinical research program in prostate biopsy. To the best of our knowledge, this is the first reported system for this clinical application and also the first reported system that leverages deep neural networks for segmentation and localization of needles in MRI across biomedical applications. Needle tip and trajectory were annotated on 583 T2-weighted intra-procedural MRI scans acquired after needle insertion for 71 patients who underwent transperineal MRI-targeted biopsy procedure at our institution. The images were divided into two independent training-validation and test sets at the patient level. A deep 3-D fully convolutional neural network model was developed, trained, and deployed on these samples. The accuracy of the proposed method, as tested on previously unseen data, was 2.80-mm average in needle tip detection and 0.98° in needle trajectory angle. An observer study was designed in which independent annotations by a second observer, blinded to the original observer, were compared with the output of the proposed method. The resultant error was comparable to the measured inter-observer concordance, reinforcing the clinical acceptability of the proposed method. The proposed system has the potential for deployment in clinical routine.

This paper presents a method for automatically calibrating and assessing the calibration quality of an externally tracked 2-D ultrasound (US) probe by scanning arbitrary, natural tissues, as opposed a specialized calibration phantom as is the typical practice. A generative topic model quantifies the posterior probability of calibration parameters conditioned on local 2-D image features arising from a generic underlying substrate. Auto-calibration is achieved by identifying the maximum a-posteriori image-to-probe transform, and calibration quality is assessed online in terms of the posterior probability of the current image-to-probe transform. Both are closely linked to the 3-D point reconstruction error (PRE) in aligning feature observations arising from the same underlying physical structure in different US images. The method is of practical importance in that it operates simply by scanning arbitrary textured echogenic structures, e.g., in-vivo tissues in the context of the US-guided procedures, without requiring specialized calibration procedures or equipment. Observed data take the form of local scale-invariant features that can be extracted and fit to the model in near real-time. Experiments demonstrate the method on a public data set of in vivo human brain scans of 14 unique subjects acquired in the context of neurosurgery. Online calibration assessment can be performed at approximately 3 Hz for the US images of pixels. Auto-calibration achieves an internal mean PRE of 1.2 mm and a discrepancy of [2 mm, 6 mm] in comparison to the calibration via a standard phantom-based method.

Segmentation is a fundamental task for extracting semantically meaningful regions from an image. The goal of segmentation algorithms is to accurately assign object labels to each image location. However, image-noise, shortcomings of algorithms, and image ambiguities cause uncertainty in label assignment. Estimating the uncertainty in label assignment is important in multiple application domains, such as segmenting tumors from medical images for radiation treatment planning. One way to estimate these uncertainties is through the computation of posteriors of Bayesian models, which is computationally prohibitive for many practical applications. On the other hand, most computationally efficient methods fail to estimate label uncertainty. We therefore propose in this paper the Active Mean Fields (AMF) approach, a technique based on Bayesian modeling that uses a mean-field approximation to efficiently compute a segmentation and its corresponding uncertainty. Based on a variational formulation, the resulting convex model combines any label-likelihood measure with a prior on the length of the segmentation boundary. A specific implementation of that model is the Chan-Vese segmentation model (CV), in which the binary segmentation task is defined by a Gaussian likelihood and a prior regularizing the length of the segmentation boundary. Furthermore, the Euler-Lagrange equations derived from the AMF model are equivalent to those of the popular Rudin-Osher-Fatemi (ROF) model for image denoising. Solutions to the AMF model can thus be implemented by directly utilizing highly-efficient ROF solvers on log-likelihood ratio fields. We qualitatively assess the approach on synthetic data as well as on real natural and medical images. For a quantitative evaluation, we apply our approach to the icgbench dataset.

Using image-based descriptors to investigate clinical hypotheses and therapeutic implications is challenging due to the notorious "curse of dimensionality" coupled with a small sample size. In this paper, we present a low-dimensional analysis of anatomical shape variability in the space of diffeomorphisms and demonstrate its benefits for clinical studies. To combat the high dimensionality of the deformation descriptors, we develop a probabilistic model of principal geodesic analysis in a bandlimited low-dimensional space that still captures the underlying variability of image data. We demonstrate the performance of our model on a set of 3D brain MRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Our model yields a more compact representation of group variation at substantially lower computational cost than models based on the high-dimensional state-of-the-art approaches such as tangent space PCA (TPCA) and probabilistic principal geodesic analysis (PPGA).

There is increasing evidence that iron deposition occurs in specific regions of the brain in normal aging and neurodegenerative disorders such as Parkinson's, Huntington's, and Alzheimer's disease. Iron deposition changes the magnetic susceptibility of tissue, which alters the MR signal phase, and allows estimation of susceptibility differences using quantitative susceptibility mapping (QSM). We present a method for quantifying susceptibility by inversion of a perturbation model, or "QSIP." The perturbation model relates phase to susceptibility using a kernel calculated in the spatial domain, in contrast to previous Fourier-based techniques. A tissue/air susceptibility atlas is used to estimate B0 inhomogeneity. QSIP estimates in young and elderly subjects are compared to postmortem iron estimates, maps of the Field-Dependent Relaxation Rate Increase, and the L1-QSM method. Results for both groups showed excellent agreement with published postmortem data and in vivo FDRI: statistically significant Spearman correlations ranging from Rho=0.905 to Rho=1.00 were obtained. QSIP also showed improvement over FDRI and L1-QSM: reduced variance in susceptibility estimates and statistically significant group differences were detected in striatal and brainstem nuclei, consistent with age-dependent iron accumulation in these regions.

The main contribution of this work is a framework to register anatomical structures characterized as a point set where each point has an associated symmetric matrix. These matrices can represent problem-dependent characteristics of the registered structure. For example, in airways, matrices can represent the orientation and thickness of the structure. Our framework relies on a dense tensor field representation which we implement sparsely as a kernel mixture of tensor fields. We equip the space of tensor fields with a norm that serves as a similarity measure. To calculate the optimal transformation between two structures we minimize this measure using an analytical gradient for the similarity measure and the deformation field, which we restrict to be a diffeomorphism. We illustrate the value of our tensor field model by comparing our results with scalar and vector field based models. Finally, we evaluate our registration algorithm on synthetic data sets and validate our approach on manually annotated airway trees.

Deformable image registration is used increasingly in image-guided interventions and other applications. However, validation and characterization of registration performance remain areas that require further study. We propose an analysis methodology for deriving tolerance limits on the initial conditions for deformable registration that reliably lead to a successful registration. This approach results in a concise summary of the probability of registration failure, while accounting for the variability in the test data. The (β, γ) tolerance limit can be interpreted as a value of the input parameter that leads to successful registration outcome in at least 100β% of cases with the 100γ% confidence. The utility of the methodology is illustrated by summarizing the performance of a deformable registration algorithm evaluated in three different experimental setups of increasing complexity. Our examples are based on clinical data collected during MRI-guided prostate biopsy registered using publicly available deformable registration tool. The results indicate that the proposed methodology can be used to generate concise graphical summaries of the experiments, as well as a probabilistic estimate of the registration outcome for a future sample. Its use may facilitate improved objective assessment, comparison and retrospective stress-testing of deformable.

Graph-based methods have become popular in recent years and have successfully addressed tasks like segmentation and deformable registration. Their main strength is optimality of the obtained solution while their main limitation is the lack of precision due to the grid-like representations and the discrete nature of the quantized search space. In this paper we introduce a novel approach for combined segmentation/registration of brain tumors that adapts graph and sampling resolution according to the image content. To this end we estimate the segmentation and registration marginals towards adaptive graph resolution and intelligent definition of the search space. This information is considered in a hierarchical framework where uncertainties are propagated in a natural manner. State of the art results in the joint segmentation/registration of brain images with low-grade gliomas demonstrate the potential of our approach.

Identifying patterns from the neuroimaging recordings of brain activity related to the unobservable psychological or mental state of an individual can be treated as a unsupervised pattern recognition problem. The main challenges, however, for such an analysis of fMRI data are: a) defining a physiologically meaningful feature-space for representing the spatial patterns across time; b) dealing with the high-dimensionality of the data; and c) robustness to the various artifacts and confounds in the fMRI time-series. In this paper, we present a network-aware feature-space to represent the states of a general network, that enables comparing and clustering such states in a manner that is a) meaningful in terms of the network connectivity structure; b)computationally efficient; c) low-dimensional; and d) relatively robust to structured and random noise artifacts. This feature-space is obtained from a spherical relaxation of the transportation distance metric which measures the cost of transporting ``mass'' over the network to transform one function into another. Through theoretical and empirical assessments, we demonstrate the accuracy and efficiency of the approximation, especially for large problems. While the application presented here is for identifying distinct brain activity patterns from fMRI, this feature-space can be applied to the problem of identifying recurring patterns and detecting outliers in measurements on many different types of networks, including sensor, control and social networks.

This paper investigates ordinal image description for invariant feature correspondence. Ordinal description is a meta-technique which considers image measurements in terms of their ranks in a sorted array, instead of the measurement values themselves. Rank-ordering normalizes descriptors in a manner invariant under monotonic deformations of the underlying image measurements, and therefore serves as a simple, non-parametric substitute for ad hoc scaling and thresholding techniques currently used. Ordinal description is particularly well-suited for invariant features, as the high dimensionality of state-of-the-art descriptors permits a large number of unique rank-orderings, and the computationally complex step of sorting is only required once after geometrical normalization. Correspondence trials based on a benchmark data set show that in general, rank-ordered SIFT (SIFT-rank) descriptors outperform other state-of-the-art descriptors in terms of precision-recall, including standard SIFT and GLOH.

We model images and the anatomy that they are derived from as stationary and jointly ergodic random processes. Using an empirically-observed property of anatomy, and data processing inequality arguments, we arrive at optimality criteria for mutual information in the ensemble domain. Using ergodicity, we transfer the criteria to single pairs of images in the spatial domain, where it applies to the popular mutual information-based registration approach that is used in practice.

This paper presents feature-based morphometry (FBM), a new, fully data-driven technique for identifying group-related differences in volumetric imagery. In contrast to most morphometry methods which assume one-to-one correspondence between all subjects, FBM models images as a collage of distinct, localized image features which may not be present in all subjects. FBM thus explicitly accounts for the case where the same anatomical tissue cannot be reliably identified in all subjects due to disease or anatomical variability. A probabilistic model describes features in terms of their appearance, geometry, and relationship to subgroups of a population, and is automatically learned from a set of subject images and group labels. Features identified indicate group-related anatomical structure that can potentially be used as disease biomarkers or as a basis for computer-aided diagnosis. Scale-invariant image features are used, which reflect generic, salient patterns in the image. Experiments validate FBM clinically in the analysis of normal (NC) and Alzheimer's (AD) brain images using the freely available OASIS database. FBM automatically identifies known structural differences between NC and AD subjects in a fully data-driven fashion, and obtains an equal error classification rate of 0.78 on new subjects.

We introduce an algorithm for segmenting brain magnetic resonance (MR) images into anatomical compartments such as the major tissue classes and neuro-anatomical structures of the gray matter. The algorithm is guided by prior information represented within a tree structure. The tree mirrors the hierarchy of anatomical structures and the subtrees correspond to limited segmentation problems. The solution to each problem is estimated via a conventional classifier. Our algorithm can be adapted to a wide range of segmentation problems by modifying the tree structure or replacing the classifier. We evaluate the performance of our new segmentation approach by revisiting a previously published statistical group comparison between first-episode schizophrenia patients, first-episode affective psychosis patients, and comparison subjects. The original study is based on 50 MR volumes in which an expert identified the brain tissue classes as well as the superior temporal gyrus, amygdala, and hippocampus. We generate analogous segmentations using our new method and repeat the statistical group comparison. The results of our analysis are similar to the original findings, except for one structure (the left superior temporal gyrus) in which a trend-level statistical significance (p = 0.07) was observed instead of statistical significance.

Characterizing the performance of image segmentation approaches has been a persistent challenge. Performance analysis is important since segmentation algorithms often have limited accuracy and precision. Interactive drawing of the desired segmentation by human raters has often been the only acceptable approach, and yet suffers from intra-rater and inter-rater variability. Automated algorithms have been sought in order to remove the variability introduced by raters, but such algorithms must be assessed to ensure they are suitable for the task. The performance of raters (human or algorithmic) generating segmentations of medical images has been difficult to quantify because of the difficulty of obtaining or estimating a known true segmentation for clinical data. Although physical and digital phantoms can be constructed for which ground truth is known or readily estimated, such phantoms do not fully reflect clinical images due to the difficulty of constructing phantoms which reproduce the full range of imaging characteristics and normal and pathological anatomical variability observed in clinical data. Comparison to a collection of segmentations by raters is an attractive alternative since it can be carried out directly on the relevant clinical imaging data. However, the most appropriate measure or set of measures with which to compare such segmentations has not been clarified and several measures are used in practice. We present here an expectation-maximization algorithm for simultaneous truth and performance level estimation (STAPLE). The algorithm considers a collection of segmentations and computes a probabilistic estimate of the true segmentation and a measure of the performance level represented by each segmentation. The source of each segmentation in the collection may be an appropriately trained human rater or raters, or may be an automated segmentation algorithm. The probabilistic estimate of the true segmentation is formed by estimating an optimal combination of the segmentations, weighting each segmentation depending upon the estimated performance level, and incorporating a prior model for the spatial distribution of structures being segmented as well as spatial homogeneity constraints. STAPLE is straightforward to apply to clinical imaging data, it readily enables assessment of the performance of an automated image segmentation algorithm, and enables direct comparison of human rater and algorithm performance.

We present a new multimodal image registration method based on the a priori knowledge of the class label mappings between two segmented input images. A joint class histogram between the image pairs is estimated by assigning each bin value equal to the total number of occurrences of the corresponding class label pairs. The discrepancy between the observed and expected joint class histograms should be minimized when the transformation is optimal. Kullback-Leibler distance (KLD) is used to measure the difference between these two histograms. Based on the probing experimental results on a synthetic dataset as well as a pair of precisely registered 3D clinical volumes, we show that, with the knowledge of the expected joint class histogram, our method obtained longer capture range and fewer local optimal points as compared with the conventional mutual information (MI) based registration method. We also applied the proposed method to a 2D-3D rigid registration problems between DSA and MRA volumes. Based on manually selected markers, we found that the accuracies of our method and the MI-based method are comparable. Moreover, our method is more computationally efficient than the MI-based method.

We describe an animated electro-holographic visualization of brain lesions due to the progression of multiple sclerosis. A research case study is used which documents the expression of visible brain lesions in a series of magnetic resonance imaging (MRI) volumes collected over the interval of one year. Some of the salient information resident within this data is described, and the motivation for using a dynamic spatial display to explore its spatial and temporal characteristics is stated. We provide a brief overview of spatial displays in medical imaging applications, and then describe our experimental visualization pipeline, from the processing of MRI datasets, through model construction, computer graphic rendering, and hologram encoding. The utility, strengths and shortcomings of the electro-holographic visualization are described and future improvements are suggested.

We propose a shape-based approach to curve evolution for the segmentation of medical images containing known object types. In particular, motivated by the work of Leventon, Grimson, and Faugeras, we derive a parametric model for an implicit representation of the segmenting curve by applying principal component analysis to a collection of signed distance representations of the training data. The parameters of this representation are then manipulated to minimize an objective function for segmentation. The resulting algorithm is able to handle multidimensional data, can deal with topological changes of the curve, is robust to noise and initial contour placements, and is computationally efficient. At the same time, it avoids the need for point correspondences during the training phase of the algorithm. We demonstrate this technique by applying it to two medical applications; two-dimensional segmentation of cardiac magnetic resonance imaging (MRI) and three-dimensional segmentation of prostate MRI.

We propose a model-based curve evolution technique for segmentation of images containing known object types. In particular, motivated by the work of Leventon et al. (2000), we derive a parametric model for an implicit representation of the segmenting curve by applying principal component analysis to a collection of signed distance representations of the training data, The parameters of this representation are then calculated to minimize an objective function for segmentation. We found the resulting algorithm to be computationally efficient, able to handle multidimensional data, robust to noise and initial contour placements, while at the same time, avoiding the need for point correspondences during the training phase of the algorithm. We demonstrate this technique by applying it to two medical applications.

The registration of pre-operative volumetric datasets to intra-operative two-dimensional images provides an improved way of verifying patient position and medical instrument location. In applications from orthopedics to neurosurgery, it has great value in maintaining up-to-date information about changes due to intervention. We propose a mutual information-based registration algorithm which establishes the proper alignment via a stochastic gradient ascent strategy. Our main contribution lies in estimating probability density measures of image intensities with a sparse histogramming method which could lead to potential speedup over existing registration procedures and deriving the gradient estimates required by the maximization procedure. Experimental results are presented on fluoroscopy and CT datasets of a real skull, and on a CT-derived dataset of a real skull, a plastic skull and a plastic lumbar spine segment.

A simple method is presented for detecting, localizing and recognizing classes of objects, that accommodates wide variation in an object's pose. The method utilizes a small two-dimensional template that is warped into an image, and converts localization to a one-dimensional sub-problem, with the search for a match between image and template executed by dynamic programming. The method recovers three of the six degrees of freedom of motion (2 translation, 1 rotation), and accommodates two more DOF in the search process (1 rotation, 1 translation), and is extensible to the final DOF. Experiments demonstrate that the method provides an efficient search strategy that outperforms normalized correlation. This is demonstrated in the example domain of face detection and localization, and is extended to more general detection tasks. An additional technique recovers a rough object pose from the match results, and is used in a two stage recognition experiment using maximization of mutual information.

While the role and utility of Magnetic Resonance Images as a diagnostic tool are well established in current clinical practice, there are a number of emerging medical arenas in which MRI can play an equally important role. In this article, we consider the problem of image-guided surgery, and provide an overview of a series of techniques that we have recently developed in order to automatically utilize MRI-based anatomical reconstructions for surgical guidance and navigation.

We describe a fully-automatic three-dimensional (3-D)-segmentation technique for brain magnetic resonance (MR) images. By means of Markov random fields (MRF's) the segmentation algorithm captures three features that are of special importance for MR images, i.e., nonparametric distributions of tissue intensities, neighborhood correlations, and signal inhomogeneities. Detailed simulations and real MR images demonstrate the performance of the segmentation algorithm. In particular, the impact of noise, inhomogeneity, smoothing, and structure thickness are analyzed quantitatively. Even single-echo MR images are well classified into gray matter, white matter, cerebrospinal fluid, scalp-bone, and background. A simulated annealing and an iterated conditional modes implementation are presented.

There is a need for frameless guidance systems to help surgeons plan the exact location for incisions, to define the margins of tumors, and to precisely identify locations of neighboring critical structures. The authors have developed an automatic technique for registering clinical data, such as segmented magnetic resonance imaging (MRI) or computed tomography (CT) reconstructions, with any view of the patient on the operating table. The authors demonstrate on the specific example of neurosurgery. The method enables a visual mix of live video of the patient and the segmented three-dimensional (3-D) MRI or CT model. This supports enhanced reality techniques for planning and guiding neurosurgical procedures and allows us to interactively view extracranial or intracranial structures nonintrusively. Extensions of the method include image guided biopsies, focused therapeutic procedures, and clinical studies involving change detection over time sequences of images.

Intensity-based classification of MR images has proven problematic, even when advanced techniques are used. Intrascan and interscan intensity inhomogeneities are a common source of difficulty. While reported methods have had some success in correcting intrascan inhomogeneities, such methods require supervision for the individual scan. This paper describes a new method called adaptive segmentation that uses knowledge of tissue intensity properties and intensity inhomogeneities to correct and segment MR images. Use of the expectation-maximization (EM) algorithm leads to a method that allows for more accurate segmentation of tissue types as well as better visualization of magnetic resonance imaging (MRI) data, that has proven to be effective in a study that includes more than 1000 brain scans. Implementation and results are described for segmenting the brain in the following types of images: axial (dual-echo spin-echo), coronal [three dimensional Fourier transform (3-DFT) gradient-echo T1-weighted] all using a conventional head coil, and a sagittal section acquired using a surface coil. The accuracy of adaptive segmentation was found to be comparable with manual segmentation, and closer to manual segmentation than supervised multivariant classification while segmenting gray and white matter.

A new information-theoretic approach is presented for finding the pose of an object in an image. The technique does not require information about the surface properties of the object, besides its shape, and is robust with respect to variations of illumination. In our derivation, few assumptions are made about the nature of the imaging process. As a result, the algorithms are quite general and can foreseeably be used in a wide variety of imaging situations. Experiments are presented that demonstrate the approach in registering magnetic resonance images, aligning a complex 3D object model to real scenes including clutter and occlusion, tracking a human head in a video sequence and aligning a view-based 2D object model to real images. The method is based on a formulation of the mutual information between the model and the image. As applied in this paper, the technique is intensity-based, rather than feature-based. It works well in domains where edge or gradient-magnitude based methods have difficulty, yet it is more robust then traditional correlation. Additionally, it has an efficient implementation that is based on stochastic approximation.>

There is a need for frameless guidance systems to help neurosurgeons to plan the exact location of a craniotomy, to define the margins of tumors and to precisely identify locations of neighboring critical structures. We have developed an automatic technique for registering clinical data, such as segmented MRI or CT reconstructions, with the patient's head on the operating table. A second method calibrates the position of a video camera relative to the patient. The combination allows a visual mix of live video of the patient with the segmented 3D MRI or CT model, enabling enhanced reality techniques for planning and guiding neurosurgical procedures, and to interactively view extracranial or intracranial structures non-intrusively. Extensions of the method include image guided biopsies, focused therapeutic procedures and clinical studies involving change detection over time sequences of images.>

A simple MAP model-matching criterion that captures important aspects of recognition in controlled situations is described. A detailed metrical object model is assumed. A probabilistic model of image features is combined with a simple prior on both the pose and the feature interpretations to yield a mixed objective function. The parameters that appear in the probabilistic models can be derived from images in the application domain. By extremizing the objective function, an optimal matching between model and image feature results. Within this framework, good models of feature uncertainty allow for robustness despite inaccuracy in feature detection. In addition, the relative likelihood of features arising from either the object or the background can be evaluated in a rational way. The objective function provides a simple and uniform means of evaluating match and pose hypotheses. Several linear projection and feature models are discussed. An experimental implementation of MAP model matching, among features derived from low-resolution edge images, is described.>

An image restoration model that performs piecewise-constant restorations on images corrupted with very high noise levels is presented. The model uses a sigmoid nonlinearity at each pixel site to produce a restoration with sharp boundaries without using an explicit line process. The restoration is produced efficiently using a stochastic search procedure at constant temperature that typically requires less than 50 iterations. The model is able to restore images with up to 70% of the pixels corrupted with noise. The model is massively parallel with local neighbor interactions (four nearest neighbors), and it can be implemented on a large parallel computer or a VLSI chip

The role of vision for the task of robot mobility is discussed. An efficient technique is presented for detecting, tracking, and locating three-dimensional line segments. The utility of this technique has been demonstrated by the SRI mobile robot, which uses it to locate features in an office environment in real time (3-Hz frame rate). A formulation of structure-from-motion using line segments is described. The formulation uses longitudinal as well as transverse information about the endpoints of image line segments. Although two images suffice to form an estimate of a world line segment, more images are used to obtain a better estimate. The system operates in a sequential fashion, using prediction-based feature detection to eliminate the need for global image processing.>

Gaussian filtering is an important tool in image processing and computer vision. In this paper we discuss the background of Gaussian filtering and look at some methods for implementing it. Consideration of the central limit theorem suggests using a cascade of ``simple'' filters as a means of computing Gaussian filters. Among ``simple'' filters, uniform-coefficient finite-impulse-response digital filters are especially economical to implement. The idea of cascaded uniform filters has been around for a while [13], [16]. We show that this method is economical to implement, has good filtering characteristics, and is appropriate for hardware implementation. We point out an equivalence to one of Burt's methods [1], [3] under certain circumstances. As an extension, we describe an approach to implementing a Gaussian Pyramid which requires approximately two addition operations per pixel, per level, per dimension. We examine tradeoffs in choosing an algorithm for Gaussian filtering, and finally discuss an implementation.