Cross-correlation digital registration techniques for multispectral cell images.

Several variants of the cross-correlation technique of automatic digital image registration are tested and compared on a set of multispectral images of lymphoblastic bone marrow cells. Factors investigated are normalization, application of a gradient and subtraction of the average image density from each image point. Several of the variants are shown to have low error rates. A gradient method requiring a small degree of manual intervention is concluded to be the most sensitive approach.     

Investigation of iterative image reconstruction in three-dimensional optoacoustic tomography

Iterative image reconstruction algorithms for optoacoustic tomography (OAT), also known as photoacoustic tomography, have the ability to improve image quality over analytic algorithms due to their ability to incorporate accurate models of the imaging physics, instrument response and measurement noise. However, to date, there have been few reported attempts to employ advanced iterative image reconstruction algorithms for improving image quality in three-dimensional (3D) OAT. In this work, we implement and investigate two iterative image reconstruction methods for use with a 3D OAT small animal imager: namely a penalized least-squares (PLS) method employing a quadratic smoothness penalty and a PLS method employing a total variation norm penalty. The reconstruction algorithms employ accurate models of the ultrasonic transducer impulse responses. Experimental data sets are employed to compare the performances of the iterative reconstruction algorithms to that of a 3D filtered backprojection (FBP) algorithm. By the use of quantitative measures of image quality, we demonstrate that the iterative reconstruction algorithms can mitigate image artifacts and preserve spatial resolution more effectively than FBP algorithms. These features suggest that the use of advanced image reconstruction algorithms can improve the effectiveness of 3D OAT while reducing the amount of data required for biomedical applications.     

Hyperspectral and multispectral bioluminescence optical tomography for small animal imaging

For bioluminescence imaging studies in small animals, it is important to be able to accurately localize the three-dimensional (3D) distribution of the underlying bioluminescent source. The spectrum of light produced by the source that escapes the subject varies with the depth of the emission source because of the wavelength-dependence of the optical properties of tissue. Consequently, multispectral or hyperspectral data acquisition should help in the 3D localization of deep sources. In this paper, we describe a framework for fully 3D bioluminescence tomographic image acquisition and reconstruction that exploits spectral information. We describe regularized tomographic reconstruction techniques that use semi-infinite slab or FEM-based diffusion approximations of photon transport through turbid media. Singular value decomposition analysis was used for data dimensionality reduction and to illustrate the advantage of using hyperspectral rather than achromatic data. Simulation studies in an atlas-mouse geometry indicated that sub-millimeter resolution may be attainable given accurate knowledge of the optical properties of the animal. A fixed arrangement of mirrors and a single CCD camera were used for simultaneous acquisition of multispectral imaging data over most of the surface of the animal. Phantom studies conducted using this system demonstrated our ability to accurately localize deep point-like sources and show that a resolution of 1.5 to 2.2 mm for depths up to 6 mm can be achieved. We also include an in vivo study of a mouse with a brain tumour expressing firefly luciferase. Co-registration of the reconstructed 3D bioluminescent image with magnetic resonance images indicated good anatomical localization of the tumour.     

Radiologist evaluation of a multispectral image compression algorithm for magnetic resonance images

With the advent of teleradiology and picture archiving and communication systems (PACS), the expense and time required for image transmission and long term image archiving become important. The use of validated image compression algorithms can greatly reduce these costs. A lossy, multispectral image compression scheme at compression ratios (CR) of 25∶1 and 32∶1 was used for a set of 26 different patient MR exams. The original and compressed/decompressed (CD) image sets were evaluated in a blinded fashion by four radiologists in two phases. The main objective was to determine whether radiologic interpretation would vary between the two types of CD image sets and the corresponding originals. In general, the compression algorithm caused a slight decrease in image quality; however, the interpretation of pathology did not change between the original and CD image sets. In only one case at the maximum CR=32 did one of four radiologists change the interpretation of pathology after CD. In this study, lossy multispectral image compression of MR images at CR=25 maintained diagnostic integrity. This could play a significant role in image storage and communications.     

Fast iterative image reconstruction methods for fully 3D multispectral bioluminescence tomography

We investigate fast iterative image reconstruction methods for fully 3D multispectral bioluminescence tomography for applications in small animal imaging. Our forward model uses a diffusion approximation for optically inhomogeneous tissue, which we solve using a finite element method (FEM). We examine two approaches to incorporating the forward model into the solution of the inverse problem. In a conventional direct calculation approach one computes the full forward model by repeated solution of the FEM problem, once for each potential source location. We describe an alternative on-the-fly approach where one does not explicitly solve for the full forward model. Instead, the solution to the forward problem is included implicitly in the formulation of the inverse problem, and the FEM problem is solved at each iteration for the current image estimate. We evaluate the convergence speeds of several representative iterative algorithms. We compare the computation cost of those two approaches, concluding that the on-the-fly approach can lead to substantial reductions in total cost when combined with a rapidly converging iterative algorithm.     

Rational variety mapping for contrast-enhanced nonlinear unsupervised segmentation of multispectral images of unstained specimen

A methodology is proposed for nonlinear contrast-enhanced unsupervised segmentation of multispectral (color) microscopy images of principally unstained specimens. The methodology exploits spectral diversity and spatial sparseness to find anatomical differences between materials (cells, nuclei, and background) present in the image. It consists of rth-order rational variety mapping (RVM) followed by matrix/tensor factorization. Sparseness constraint implies duality between nonlinear unsupervised segmentation and multiclass pattern assignment problems. Classes not linearly separable in the original input space become separable with high probability in the higher-dimensional mapped space. Hence, RVM mapping has two advantages: it takes implicitly into account nonlinearities present in the image (ie, they are not required to be known) and it increases spectral diversity (ie, contrast) between materials, due to increased dimensionality of the mapped space. This is expected to improve performance of systems for automated classification and analysis of microscopic histopathological images. The methodology was validated using RVM of the second and third orders of the experimental multispectral microscopy images of unstained sciatic nerve fibers (nervus ischiadicus) and of unstained white pulp in the spleen tissue, compared with a manually defined ground truth labeled by two trained pathophysiologists. The methodology can also be useful for additional contrast enhancement of images of stained specimens.     

High-resolution simulation of B0 field conditions in the human heart from segmented computed tomography images

B₀ inhomogeneity leads to imaging artifacts in cardiac magnetic resonance imaging (MRI), in particular dark band artifacts with steady-state free precession pulse sequences. The limited spatial resolution of MR-derived in vivo B₀ maps and the lack of population data prevent systematic analysis of the problem at hand and the development of optimized B₀ shim strategies. We used readily available clinical computed tomography (CT) images to simulate the B₀ conditions in the human heart at high spatial resolution. Calculated B₀ fields showed consistency with MRI-based B₀ measurements. The B₀ maps for both the simulations and in vivo measurements showed local field inhomogeneities in the vicinity of lung tips with dominant Z3 spherical harmonic terms in the field distribution. The presented simulation approach allows for the derivation of B₀ field conditions at high spatial resolution from CT images and enables the development of subject- and population-specific B₀ shim strategies for the human heart.     

Morphologic tomography of nonspherical particles using multispectral diffusing light measurements

A series of phantom experiments are conducted to demonstrate the ability of a T-matrix-based inverse algorithm for tomographic recovery of morphologic characteristics of nonspherical particles embedded in heterogeneous turbid media. Diffusely scattered light at several wavelengths along the boundary of the phantom are collected and analyzed to allow for simultaneous extraction of the size, concentration, and aspect ratio of the spheroidal particles.     

Automated segmentation of multiple sclerosis lesions in multispectral MR imaging using fuzzy clustering

A method is presented for fully automated detection of Multiple Sclerosis (MS) lesions in multispectral magnetic resonance (MR) imaging. Based on the Fuzzy C-Means (FCM) algorithm, the method starts with a segmentation of an MR image to extract an external CSF/lesions mask, preceded by a local image contrast enhancement procedure. This binary mask is then superimposed on the corresponding data set yielding an image containing only CSF structures and lesions. The FCM is then reapplied to this masked image to obtain a mask of lesions and some undesired substructures which are removed using anatomical knowledge. Any lesion size found to be less than an input bound is eliminated from consideration. Results are presented for test runs of the method on 10 patients. Finally, the potential of the method as well as its limitations are discussed.     

Interpretation of principal components of the reflectance spectra obtained from multispectral images of exposed pig brain

The spatial variation in reflectance such as the blood-vessel pattern can be observed in the image of cerebral cortex. This spatial variation is mainly caused by the difference in concentrations of oxy- and deoxyhemoglobin in the tissue. We analyze the reflectance spectra obtained from multispectral images of pig cortex by principal component analysis to extract information that relates to physiological parameters such as the concentrations of oxy- and deoxyhemoglobin and physical parameters such as mean optical path length. The light propagation in a model of exposed pig cortex is predicted by Monte Carlo simulation to estimate the interpretation of physiological and physical meanings of the principal components. The spatial variance of reflectance spectra of the pig cortex can be approximately described by the first principal component. The first principal component reflects the spectrum of hemoglobin in the cortical tissue multiplied by the mean optical path length. These results imply that the wavelength dependence of mean optical path length can be experimentally estimated from the first principal component of the reflectance spectra obtained from multispectral image of cortical tissue.     

Motion estimation for cardiac emission tomography by optical flow methods

This paper describes a new method for estimating the 3D, non-rigid object motion in a time sequence of images. The method is a generalization of a standard optical flow algorithm that is incorporated into a successive quadratic approximation framework. The method was evaluated for gated cardiac emission tomography using images obtained from a mathematical, 4D phantom and a physical, dynamic phantom. The results showed that the proposed method offers improved motion estimation accuracy relative to the standard optical flow method. Convergence of the proposed algorithm was evidenced with a monotonically decreasing objective function value with iteration. Practical application of the motion estimation method in cardiac emission tomography includes quantitative myocardial motion estimation and 4D, motion-compensated image reconstruction.     

基于深度学习的舌象自监督聚类

目的：解决人工智能舌诊领域数据标注成本较高且带有较强主观性的问题。方法：基于深度学习中的对比学习技术,对舌象进行自监督聚类。该方法首先利用卷积神经网络将不同数据增强模式下的舌象映射到潜在空间,并在学习同类实例之间共同特征的同时尽可能区分非同类实例;随后利用高斯混合模型对网络提取的特征向量进行聚类。结果：在无需引入先验知识的情况下,利用300张舌象仪采集的无标签图像取得了52.54%的聚类纯度。结论：该方法一定程度上将医疗工作者从费事费力的数据标注工作中解放出来。除应用于自动化舌象分类外,该方法还可进一步针对不同病症的特殊舌象症候群进行聚类分析,其提取的舌象特征也可为舌体分割、舌色分类、苔质分区等下游任务提供预训练的参考。     

Physiological clustering of visual channels in the mouse retina

Anatomy predicts that mammalian retinas should have in excess of 12 physiological channels, each encoding a specific aspect of the visual scene. Although several channels have been correlated with morphological cell types, the number of morphological types generally exceeds the known physiological types. Here, we attempted to sort the ganglion cells of the mouse retina purely on a physiological basis. The null hypothesis was that the outputs of the ganglion cells form a continuum or should be divided into only a few types. We recorded the spiking output of 471 retinal ganglion cells on a multielectrode array while presenting 4 classes of visual stimuli. Five parameters were chosen to describe each cell's response characteristics, including relative amplitude of the ON and OFF responses, response latency, response transience, direction selectivity, and the receptive field surround. We compared the results of four clustering routines and judged the results using the relevant validation indices. The optimal partition was the 12-cluster solution of the Fuzzy Gustafson-Kessel algorithm. This classification contained three visual channels that carried predominately OFF responses, six that carried ON responses, and three that carried both ON and OFF information. They differed in other parameters as well. Other evidence suggests that the true number of cell types in the mouse retina may be somewhat larger than 12, and a definitive typology will probably require broader stimulus sets and characterization of more response parameters. Nonetheless, the present results do allow us to reject the null hypothesis: it appears that in addition to well-known cell types (such as the ON-OFF direction selectivity cells) numerous other cell classes can be identified in the mouse retina based solely on their responses to a standard set of simple visual stimuli.     

Fuzzy modeling of electrical impedance tomography images of the lungs

OBJECTIVES: Aiming to improve the anatomical resolution of electrical impedance tomography images, we developed a fuzzy model based on electrical impedance tomography's high temporal resolution and on the functional pulmonary signals of perfusion and ventilation. INTRODUCTION: Electrical impedance tomography images carry information about both ventilation and perfusion. However, these images are difficult to interpret because of insufficient anatomical resolution, such that it becomes almost impossible to distinguish the heart from the lungs. METHODS: Electrical impedance tomography data from an experimental animal model were collected during normal ventilation and apnea while an injection of hypertonic saline was administered. The fuzzy model was elaborated in three parts: a modeling of the heart, the pulmonary ventilation map and the pulmonary perfusion map. Image segmentation was performed using a threshold method, and a ventilation/perfusion map was generated. RESULTS: Electrical impedance tomography images treated by the fuzzy model were compared with the hypertonic saline injection method and computed tomography scan images, presenting good results. The average accuracy index was 0.80 when comparing the fuzzy modeled lung maps and the computed tomography scan lung mask. The average ROC curve area comparing a saline injection image and a fuzzy modeled pulmonary perfusion image was 0.77. DISCUSSION: The innovative aspects of our work are the use of temporal information for the delineation of the heart structure and the use of two pulmonary functions for lung structure delineation. However, robustness of the method should be tested for the imaging of abnormal lung conditions. CONCLUSIONS: These results showed the adequacy of the fuzzy approach in treating the anatomical resolution uncertainties in electrical impedance tomography images.