A semiautomated approach for artefact removal in serial tissue cryosections.

Thinly sliced serial tissue sections of an organ can be imaged using optical microscopy at a resolution detailing individual cells. When the tissue sections are first subjected to in situ hybridization or immunohistochemistry, these data sets can be analysed for changes in gene expression and gene products. Such spatial information is important for understanding the functional effects of experimental or environmental challenges to the organism. However, a critical step in analysing these data sets is mitigating artefacts that result from the preparation of the tissue sections. In this paper, we describe an automated method with manual validation tools that together enable detecting and addressing artefacts including dust particles and air bubbles.

Computational tools for quantitative breast morphometry based on 3D scans.

Quantitative analysis of breast morphometry is critical to breast plastic surgery. Recently, three-dimensional (3D) photography has emerged as a strong new alternative for breast morphometry analysis in comparison to other existing techniques. 3D photography enables the capture of the entire breast surface topology virtually in a single snapshot and without any direct contact with the patient, thus causing minimal discomfort. In this paper, we present a set of computational tools for the quantitative analysis of two key morphological properties of the breast that are of interest to breast plastic surgery based on 3D scans, namely breast shape and volume. The breast shape is modeled using a compact geometric model capable of capturing the global shape of the breast with very few parameters. Specifically, the shape model is deduced by applying a set of five global deformations to a geometric primitive. These deformations, defined using very intuitive parameters, closely model the key shape variables that surgeons inherently use to describe the overall shape of the breast. Patient-specific parameters of the breast shape model are automatically recovered by fitting a generic breast shape model to the 3D scan of the patient's breast using a physics-based deformable model framework. The mean error of fit between the automatically fitted shape model and the actual breast surface for 12 subjects varied between 0.9 and 2.6 mm. These results are very encouraging considering the fact that only 17 parameters are used to determine the shape of the breast. The breast volume is estimated automatically by first localizing the breast on a 3D scan of the patient's torso and then computing the volume enclosed between an interpolated breast-less torso surface and the actual breast. The volume estimated by the proposed method was found to be within the intra-operator variability among five segmentation trials performed manually by an expert on 3D torso scans of three subjects.

Automatic delineation of the inner thoracic region in non-contrast CT data.

The inner thoracic region consists of several important anatomical structures and an accurate delineation of this region is an essential step for various biomedical image analysis applications. In this paper, we present a fully automatic graph-based method for the delineation of the inner thoracic region in non-contrast cardiac CT data. In particular, we reformulate the problem of delineating the inner thoracic region as an optimal surface segmentation problem, the solution to which is obtained by computing the minimum-cost closed set in a node-weighted directed graph. Comparing the results obtained using our method with manual segmentations performed by an expert on non-contrast cardiac CT scans of 20 randomly selected patients indicated an overlap of 99.1 +/- 0.2%.

Towards cardiovascular risk stratification using imaging data.

In spite of the advancement and proliferation of cardiovascular imaging data, the rate of deaths due to unpredicted heart attack remains high. Thus, it becomes imperative to develop novel computational tools to mine quantitative parameters from imaging data for early detection and diagnosis of asymptomatic cardiovascular disease. In this paper, we present our progress towards developing a computational framework to mine cardiac imaging data and provide quantitative measures for developing a new risk assessment method. Specifically, we present computational methods developed for the detection of coronary calcification and segmentation of thoracic aorta in non-contrast cardiac computed tomography, and detection of neovessels in plaques in intravascular ultrasound imaging data.

Toward unsupervised classification of calcified arterial lesions.

There is growing evidence that calcified arterial deposits play a crucial role in the pathogenesis of cardiovascular disease. This paper investigates the challenging problem of unsupervised calcified lesion classification. We propose an algorithm, US-CALC (UnSupervised Calcified Arterial Lesion Classification), that discriminates arterial lesions from non-arterial lesions. The proposed method first mines the characteristics of calcified lesions using a novel optimization criterion and then identifies a subset of lesion features which is optimal for classification. Second, a two stage clustering is deployed to discriminate between arterial and non-arterial lesions. A histogram intersection distance measure is incorporated to determine cluster proximity. The clustering hierarchies are carefully validated and the final clusters are determined by a new intracluster compactness measure. Experimental results indicate an average accuracy of approximately 80% on a database of electron beam CT heart scans.

Wavelet-based Bayesian image estimation: from marginal and bivariate prior models to multivariate prior models.

Prior models play an important role in the wavelet-based Bayesian image estimation problem. Although it is well known that a residual dependency structure always remains among natural image wavelet coefficients, only few multivariate prior models with a closed parametric form are available in the literature. In this paper, we develop new multivariate prior models that not only match well with the observed statistics of the wavelet coefficients of natural images, but also have a simple parametric form. These prior models are very effective for Bayesian image estimation and lead to an improved estimation performance over related earlier techniques.

Automatic centerline extraction of irregular tubular structures using probability volumes from multiphoton imaging.

In this paper, we present a general framework for extracting 3D centerlines from volumetric datasets. Unlike the majority of previous approaches, we do not require a prior segmentation of the volume nor we do assume any particular tubular shape. Centerline extraction is performed using a morphology-guided level set model. Our approach consists of: i) learning the structural patterns of a tubular-like object, and ii) estimating the centerline of a tubular object as the path with minimal cost with respect to outward flux in gray level images. Such shortest path is found by solving the Eikonal equation. We compare the performance of our method with existing approaches in synthetic, CT, and multiphoton 3D images, obtaining substantial improvements, especially in the case of irregular tubular objects.

Automated characterization of gene expression patterns with an atlas of the mouse brain.

A spatio-temporal map of gene activity in the brain would be an important contribution to the understanding of brain development, disease, and function. Such a resource is now possible using high-throughput in situ hybridization, a method for transcriptome-wide acquisition of cellular resolution gene expression patterns in serial tissue sections. However, querying an enormous quantity of image data requires computational methods for describing and organizing gene expression patterns in a consistent manner. In addressing this, we have developed procedures for automated annotation of gene expression patterns in the postnatal mouse brain.