Comparative evaluation of autofocus algorithms for a real-time system for automatic detection of Mycobacterium tuberculosis

Microscopy images must be acquired at the optimal focal plane for the objects of interest in a scene. Although manual focusing is a standard task for a trained observer, automatic systems often fail to properly find the focal plane under different microscope imaging modalities such as bright field microscopy or phase contrast microscopy. This article assesses several autofocus algorithms applied in the study of fluorescence-labeled tuberculosis bacteria. The goal of this work was to find the optimal algorithm in order to build an automatic real-time system for diagnosing sputum smear samples, where both accuracy and computational time are important. We analyzed 13 focusing methods, ranging from well-known algorithms to the most recently proposed functions. We took into consideration criteria that are inherent to the autofocus function, such as accuracy, computational cost, and robustness to noise and to illumination changes. We also analyzed the additional benefit provided by preprocessing techniques based on morphological operators and image projection profiling.     

Robust Spectral Clustering Using Statistical Sub-Graph Affinity Model

Spectral clustering methods have been shown to be effective for image segmentation. Unfortunately, the presence of image noise as well as textural characteristics can have a significant negative effect on the segmentation performance. To accommodate for image noise and textural characteristics, this study introduces the concept of sub-graph affinity, where each node in the primary graph is modeled as a sub-graph characterizing the neighborhood surrounding the node. The statistical sub-graph affinity matrix is then constructed based on the statistical relationships between sub-graphs of connected nodes in the primary graph, thus counteracting the uncertainty associated with the image noise and textural characteristics by utilizing more information than traditional spectral clustering methods. Experiments using both synthetic and natural images under various levels of noise contamination demonstrate that the proposed approach can achieve improved segmentation performance when compared to existing spectral clustering methods.     

Imaging system for morphometric assessment of absorption or fluorescence in stained cells

An image acquisition and processing system has been developed for quantitative microscopy of absorption or fluorescence in stained cells. Three different light transducers are used in the system to exploit the best characteristics of these sensors for different biological measurements. A digital scanner, in the form of a linear array charge-coupled device (CCD), acquires data with high spatial and photometric resolution. A color (RGB) camera is employed when spectral information is required for the segmentation of cellular subcomponents. An image-intensified charged-injection device (CID) camera provides for very low light intensity measurements, primarily for fluorescence-labeled cells. Properties of these transducers, such as contrast transfer function, linearity, and photo-response nonuniformity, have been measured. Two dedicated image processing units were incorporated into the system. The front-end processor, based on a digital signal processor, provides functions such as object detection, raw image calibration, compression, artifact removal, and filtering. The second image processor is associated with the frame memory and includes a histogram processor, a dedicated arithmetic logic unit for image processing functions, and a graphics module for one-bit overlay functions. An interactive program was developed to acquire cell images and to experiment with a range of segmentation algorithms, feature extractions, and other image processing functions. The results of any image operation are displayed on the video monitor. Once a desired processing sequence is determined, the sequence may be stored to become part of a command library and can be executed thereafter as a single instruction.     

基于频谱图像灰度共生矩阵的无损测温方法

本文提出了一种基于哈达玛变换的频谱图像灰度共生矩阵(Hadamard-GLCM)的高强度聚焦超声治疗无损测温方法。利用高强度聚焦超声辐照新鲜离体猪肉组织,获取辐照前后的B超图像的减影图像,采用Hadamard变换对其进行处理,获取频谱图像,将频谱图像的灰度共生矩阵惯性矩作为反应温度变化的信息参数。实验表明:不仅单组数据的Hadamard-GLCM惯性矩(HGMI)和温度能很好的线性拟合,而且多组数据的Hadamard-GLCM惯性矩与温度也成近似的线性关系,而且斜率非常接近,拟合度更接近1,误差小,对温度的分辨能力高,容错能力强,与传统的测温方法相比有着明显的优势,能为HIFU治疗过程中的无损测温提供有效的实时依据。     

An effective approach to selecting the appropriate pan-sharpening method in digital change detection of natural ecosystems

In an image fusion process, the spatial resolution of a multispectral image is improved by a panchromatic band. However, due to the spatial and spectral resolution differences between these two data sets, the enhanced image may have two distortions, spatial and spectral. Therefore, to evaluate the efficiency of the pan-sharpening method, the status of these two types of distortions is examined. Unfortunately, there is still no developed acceptance index that can thoroughly investigate the quality of the pansharpened image; moreover, most of the proposed methods for reviewing the quality of output images have been developed with an emphasis on the residential area. Accordingly, to assess the quality of the pansharpened image in this study, we evaluated highly effective conventional methods, such as visual examinations, quantitative evaluation and impact analysis regarding the change detection process of mangrove forests. Finally, we suggested a simple yet efficient approach for such research in natural ecosystems. In the proposed method, based on the nature of the ecosystem, a spectral vegetation index is applied to the pansharpened images; the spectral quality of the images is further evaluated based on two parameters, 1) the areas under the curves of the histogram of the spectral vegetation index in the natural ecosystem region and 2) its centroid. The spatial quality of the pansharpened images is evaluated through implementing of two transects perpendicular to each other in the images of the spectral vegetation index, and creating a spatial deviation on them. With expert reviews and visual evaluation of the pansharpened images, the proposed method, especially in natural ecosystems, has more advantages as regards assessing the quality of the fused images. Based on the evaluations, among 11 methods of pansharpening, including Ehlers Fusion, FuzeGO, Gram-Schmidt, HPF, HCS, PCA, Modified IHS, Brovey Transform, Projective Resolution Merge, Wavelet IHS, and Wavelet PCA; the HPF method the Brovey Transform and Modified IHS methods respectively showed the best performance in the digital change detection of Mangrove forests.     

Adaptive color basis transformation. An aid in image segmentation

Multispectral images of stained cells enable the use of color differences to segment and/or to discriminate between image components, such as cell types and cellular subcomponents. When the spectral characteristics of the image components do not change over the area of a slide or from slide to slide, one can create a constant weighted linear combination of spectral images to generate one-dimensional or two-dimensional images that have the desired contrast between the image components that must be discriminated. However, when the spectral characteristics are not constant, i.e., when they vary from image to image, a constant weighted linear combination cannot be employed; instead, an appropriate solution must be found for each selected image. This is usually a time-consuming, manual procedure that cannot be employed in a fully automated process of discriminating and segmenting stained cells. This paper describes an algorithm that uses principal components decomposition basis vectors to generate a nonstatic weighted linear combination of color images that can be used by an automated system. This algorithm relies on a semiconstant relationship between the areas (sizes) of the image components that are to be discriminated and/or segmented. The technique has been successfully applied as an aid in the segmentation of images of stained cervical smears; the images were acquired with a three-chip CCD camera that generates three broad-band color images.     

Electronic light microscopy: present capabilities and future prospects

Electronic light microscopy involves the combination of microscopic techniques with electronic imaging and digital image processing, resulting in dramatic improvements in image quality and ease of quantitative analysis. In this review, after a brief definition of digital images and a discussion of the sampling requirements for the accurate digital recording of optical images, I discuss the three most important imaging modalities in electronic light microscopy-video-enhanced contrast microscopy, digital fluorescence microscopy and confocal scanning microscopy-considering their capabilities, their applications, and recent developments that will increase their potential. Video-enhanced contrast microscopy permits the clear visualisation and real-time dynamic recording of minute objects such as microtubules, vesicles and colloidal gold particles, an order of magnitude smaller than the resolution limit of the light microscope. It has revolutionised the study of cellular motility, and permits the quantitative tracking of organelles and gold-labelled membrane bound proteins. In combination with the technique of optical trapping (optical tweezers), it permits exquisitely sensitive force and distance measurements to be made on motor proteins. Digital fluorescence microscopy enables low-light-level imaging of fluorescently labelled specimens. Recent progress has involved improvements in cameras, fluorescent probes and fluorescent filter sets, particularly multiple bandpass dichroic mirrors, and developments in multiparameter imaging, which is becoming particularly important for in situ hybridisation studies and automated image cytometry, fluorescence ratio imaging, and time-resolved fluorescence. As software improves and small computers become more powerful, computational techniques for out-of-focus blur deconvolution and image restoration are becoming increasingly important. Confocal microscopy permits convenient, high-resolution, non-invasive, blur-free optical sectioning and 3D image acquisition, but suffers from a number of limitations. I discuss advances in confocal techniques that address the problems of temporal resolution, spherical and chromatic aberration, wavelength flexibility and cross-talk between fluorescent channels, and describe new optics to enhance axial resolution and the use of two-photon excitation to reduce photobleaching. Finally, I consider the desirability of establishing a digital image database, the BioImage database, which would permit the archival storage of, and public Internet access to, multidimensional image data from all forms of biological microscopy. Submission of images to the BioImage database would be made in coordination with the scientific publication of research results based upon these data. In the context of electronic light microscopy, this would be particularly useful for three-dimensional images of cellular structure and video sequences of dynamic cellular processes, which are otherwise hard to communicate. However, it has the wider significance of allowing correlative studies on data obtained from many different microscopies and from sequence and crystallographic investigations. It also opens the door to interactive hypermedia access to the multidimensional image data, and multimedia publishing ventures based upon this.Presented at the XXXVII Symposium of the Society for Histochemistry, 23 September 1995, Rigi Kaltbad, Switzerland     

Noninvasive diagnosis of melanoma with tensor decomposition-based feature extraction from clinical color image

We propose a method for feature extraction from clinical color images, with application in classification of skin lesions. Proposed feature extraction method is based on tensor decomposition of the clinical color image of skin lesion. Since color image is naturally represented as a three-way tensor, it is reasonable to use multi-way techniques to capture the underlying information contained in the image. Extracted features are elements of the core tensor in the corresponding multi-way decomposition, and represent spatial-spectral profile of the lesion. In contrast to common methods that exploit either texture or spectral diversity of the tumor only, the proposed approach simultaneously captures spatial and spectral characteristics. The procedure is tested on a problem of noninvasive diagnosis of melanoma from the clinical color images of skin lesions, with overall sensitivity 82.1% and specificity 86.9%. Our method compares favorably with the state of the art results reported in the literature and provides an interesting alternative to the existing approaches.     

Passive autofocusing for digital endoscopic imaging systems.

An autofocus is a desirable feature of an endoscope, because it relieves the user from performing a task which can be automated and thereby prevents unnecessary interruptions in the work to be performed. Autofocusing is in general best achieved by an active system, i.e., on the basis of a distance measurement. Yet, in handheld medical endoscopes such a method is unsuited because of the added weight associated with the necessary electromechanical components. Autofocusing should rather be performed passively, furthermore, in applications which are particularly critical with respect to safety, e.g., in the eye, a stable and reliable operation in real time and without interruption is necessary. Passive autofocus strategies applied to date and known to the authors lead however to algorithms which are either too slow for a real time implementation and/or are influenced by the structure of the object which is to be brought into focus. Accordingly, a new autofocus procedure has been developed which exhibits a stable and reliable operation in real time under all circumstances of interest. It is based on the squared differences of the intensity of adjacent points in both dimensions of a plane image (Square Plane Sum Modulus Difference, SPSMD) and as such particularly suitable for digital camera systems and real-time needs (typically, 30 evaluations per second on an image of 1024 x 1024 pixels). The SPSMD criterion is more sensitive, has a larger SNR than other focus criteria known to the authors and exhibits in particular no secondary extrema which could adversely affect proper focusing. As it includes intensity differences in both (perpendicular) directions in the image plane, it is essentially independent of image structures.     

NetMHCIIpan-2.0 - Improved pan-specific HLA-DR predictions using a novel concurrent alignment and weight optimization training procedure

Background

Macrophages represent the front lines of our immune system; they recognize and engulf pathogens or foreign particles thus initiating the immune response. Imaging macrophages presents unique challenges, as most optical techniques require labeling or staining of the cellular compartments in order to resolve organelles, and such stains or labels have the potential to perturb the cell, particularly in cases where incomplete information exists regarding the precise cellular reaction under observation. Label-free imaging techniques such as Raman microscopy are thus valuable tools for studying the transformations that occur in immune cells upon activation, both on the molecular and organelle levels. Due to extremely low signal levels, however, Raman microscopy requires sophisticated image processing techniques for noise reduction and signal extraction. To date, efficient, automated algorithms for resolving sub-cellular features in noisy, multi-dimensional image sets have not been explored extensively.

Results

We show that hybrid z-score normalization and standard regression (Z-LSR) can highlight the spectral differences within the cell and provide image contrast dependent on spectral content. In contrast to typical Raman imaging processing methods using multivariate analysis, such as single value decomposition (SVD), our implementation of the Z-LSR method can operate nearly in real-time. In spite of its computational simplicity, Z-LSR can automatically remove background and bias in the signal, improve the resolution of spatially distributed spectral differences and enable sub-cellular features to be resolved in Raman microscopy images of mouse macrophage cells. Significantly, the Z-LSR processed images automatically exhibited subcellular architectures whereas SVD, in general, requires human assistance in selecting the components of interest.

Conclusions

The computational efficiency of Z-LSR enables automated resolution of sub-cellular features in large Raman microscopy data sets without compromise in image quality or information loss in associated spectra. These results motivate further use of label free microscopy techniques in real-time imaging of live immune cells.     

Multiple excitation confocal analysis of targets in nuclei of cytogenetic preparations

OBJECTIVE: To demonstrate that cellular preparations requiring color analysis of different domains stained by molecular cytogenetic methods (fluorescence in situ hybridization) can be processed by spectral analysis of fluorescent emissions by either factor analysis of medical image sequences (FAMIS) or a META confocal configuration to isolate fluorescent probes. STUDY DESIGN: Three-dimensional sequences of images obtained by spectral analysis in a META confocal microscope (Carl Zeiss SAS, Jena, Germany) were analyzed by META processing and the FAMIS algorithm, which provides factor curves. META and factor images were then the result of image-processing methods that cover emission spectra. RESULTS: Factor curves and factor or META images can help to analyze targets inside nuclei. CONCLUSION: It is possible to process preparations containing numerous spots on different colors to differentiate stained targets and to improve visualization and detection.     

Cellular logic implementation on a generalized image processor applied to biomedical image processing

The implementation on a generalized image processor of a cellular logic package that performs non-recursive cellular-logic operations (CLOs) in real time is described. This system takes advantage of up to 20 512 X 512 X 8-bit memory planes within the image processor and can manipulate cells with up to 256 symbolic states. The flexibility of the image processor allows the use of an expanded cellular transition set, beyond bit-on or bit-off, as well as application-specific neighborhood configurations. The use of concurrent data-dependent global calculations, including CLO iteration termination control, is described. The array processor implementation specifics are discussed. This general cellular logic package is applied to biomedical images in the Image Processing Laboratory, Department of Radiological Sciences, University of California at Los Angeles. Geometric information is acquired from the images using real-time operators on the image array processor. This information includes image segmentation, area calculation, object counting, centroid determination and shape analysis. Initial clinical results are presented, and possible future medical applications are discussed.     

Feature extraction from three-dimensional images in quantitative microscopy

Different methods are investigated in selecting and generating the appropriate microscope images for analysis of three-dimensional objects in quantitative microscopy. Traditionally, the ‘best’ focused image from a set is used for quantitative analysis. Such an objectively determined image is optimal for the extraction of some features, but may not be the best image for the extraction of all features. Various methods using multiple images are here developed to obtain a tighter distribution for all features.Three different approaches for analysis of images of stained cervical cells were analyzed. In the first approach, features are extracted from each image in the set. The feature values are then averaged to give the final result. In the second approach, a set of varying focused images are reconstructed to obtain a set of in-focus images. Features are then extracted from this set and averaged. In the third approach, a set of images in the three-dimensional scene is compressed into a single two-dimensional image. Four different compression methods are used. Features are then extracted from the resulting two-dimensional image. The third approach is employed on both the raw and transformed images.Each approach has its advantages and disadvantages. The first approach is fast and produces reasonable results. The second approach is more computationally expensive but produces the best results. The last approach overcomes the memory storage problem of the first two approaches since the set of images is compressed into one. The method of compression using the highest gradient pixel produces better results overall than other data reduction techniques and produces results comparable to the first approach.     

Predicting protein structural classes with pseudo amino acid composition: an approach using geometric moments of cellular automaton image

A novel approach was developed for predicting the structural classes of proteins based on their sequences. It was assumed that proteins belonging to the same structural class must bear some sort of similar texture on the images generated by the cellular automaton evolving rule [Wolfram, S., 1984. Cellular automation as models of complexity. Nature 311, 419-424]. Based on this, two geometric invariant moment factors derived from the image functions were used as the pseudo amino acid components [Chou, K.C., 2001. Prediction of protein cellular attributes using pseudo amino acid composition. Proteins: Struct., Funct., Genet. (Erratum: ibid., 2001, vol. 44, 60) 43, 246-255] to formulate the protein samples for statistical prediction. The success rates thus obtained on a previously constructed benchmark dataset are quite promising, implying that the cellular automaton image can help to reveal some inherent and subtle features deeply hidden in a pile of long and complicated amino acid sequences.     

Improved confocal characterization of targets in nuclei of cytogenetic preparations

OBJECTIVE: To show that cellular preparations requiring depth analysis of different domains stained by molecular cytogenetic methods (fluorescence in situ hybridization and primed in situ) can be improved by regularized factor analysis of medical image sequences (FAMIS) to isolate fluorescent probes by means of intensity depth profiles of fluorochromes, to track relevant DNA sequences (cosmids and centromeres) in cell nuclei during interphase and to improve the use of cytogenetic techniques resulting in flat preparations of whole cells that are assumed to preserve probe access to their targets. STUDY DESIGN: 3D sequences of images obtained by depth displacement in a confocal microscope were first analyzed by the FAMIS algorithm, which provides factor curves. Factor images then resulted from regularization methods that improve signal/noise ratio while preserving target contours. RESULTS: Factor curves and regularized factor images helped analyze targets inside nuclei. CONCLUSION: It is possible to process preparations containing numerous spots (even when they are on different planes) to differentiate stained targets, to investigate depth differences and to improve visualization and detection.     

UAV-based classification of maritime Antarctic vegetation types using GEOBIA and random forest

Development of vegetation communities in areas of Antarctica without permanent ice cover emphasizes the need for effective remote sensing techniques for proper monitoring of local environmental changes. Detection and mapping of vegetation by image classification remains limited in the Antarctic environment due to the complexity of its surface cover, and the spatial heterogeneity and spectral homogeneity of cryptogamic vegetation. As ultra-high resolution aerial images allow a comprehensive analysis of vegetation, this study aims to identify different types of vegetation cover (i.e., algae, mosses, and lichens) in an ice-free area of  Hope Bay, on the northern tip of the Antarctic Peninsula. Using the geographic object-based image analysis (GEOBIA) approach, remote sensing data sets are tested in the random forest classifier in order to distinguish vegetation classes within vegetated areas. Because species of algae, mosses, and lichens may have similar spectral characteristics, subclasses are established. The results show that when only the mean values of green, red, and NIR bands are considered, the subclasses have low separability. Variations in accuracy and visual changes are identified according to the set of features used in the classification. Accuracy improves when multilayer information is used. A combination of spectral and morphometric products and by-products provides the best result for the detection and delineation of different types of vegetation, with an overall accuracy of 0.966 and a Kappa coefficient of 0.946. The method allowed for the identification of units primarily composed of algae, mosses, and lichens as well as differences in communities. This study demonstrates that ultra-high spatial resolution data can provide the necessary properties for the classification of vegetation in Maritime Antarctica, even in images obtained by sensors with low spectral resolution.     

A Parallel Distributed-Memory Particle Method Enables Acquisition-Rate Segmentation of Large Fluorescence Microscopy Images

Modern fluorescence microscopy modalities, such as light-sheet microscopy, are capable of acquiring large three-dimensional images at high data rate. This creates a bottleneck in computational processing and analysis of the acquired images, as the rate of acquisition outpaces the speed of processing. Moreover, images can be so large that they do not fit the main memory of a single computer. We address both issues by developing a distributed parallel algorithm for segmentation of large fluorescence microscopy images. The method is based on the versatile Discrete Region Competition algorithm, which has previously proven useful in microscopy image segmentation. The present distributed implementation decomposes the input image into smaller sub-images that are distributed across multiple computers. Using network communication, the computers orchestrate the collectively solving of the global segmentation problem. This not only enables segmentation of large images (we test images of up to 10¹⁰ pixels), but also accelerates segmentation to match the time scale of image acquisition. Such acquisition-rate image segmentation is a prerequisite for the smart microscopes of the future and enables online data compression and interactive experiments.     

自动对焦算法评估线虫脂滴图像的研究

自动对焦是实现线虫自动化筛选的一个重要步骤．在光学显微镜系统中,通过采集同一个视野下不同焦面的图像,再通过清晰度评价函数对这些图像进行运算,得到的最大值被认为是最佳对焦位置．在本研究中,对16种常用的自动对焦算法以及最近提出的一些算法进行了评估,通过评估找出最适合线虫脂滴图像的自动对焦算法,从而搭建一套线虫脂滴自动化筛选系统．同时就对焦精度、运算时间、抗噪声能力、对焦曲线等特征进行了分析评价,结果表明,大多数算法对线虫脂滴图像都有较好的表现,特别是绝对Tenengrad算法在对焦精度上有最好的表现,我们将优选该算法应用到线虫脂滴自动化筛选系统中．     

Multiview Locally Linear Embedding for Effective Medical Image Retrieval

Content-based medical image retrieval continues to gain attention for its potential to assist radiological image interpretation and decision making. Many approaches have been proposed to improve the performance of medical image retrieval system, among which visual features such as SIFT, LBP, and intensity histogram play a critical role. Typically, these features are concatenated into a long vector to represent medical images, and thus traditional dimension reduction techniques such as locally linear embedding (LLE), principal component analysis (PCA), or laplacian eigenmaps (LE) can be employed to reduce the “curse of dimensionality”. Though these approaches show promising performance for medical image retrieval, the feature-concatenating method ignores the fact that different features have distinct physical meanings. In this paper, we propose a new method called multiview locally linear embedding (MLLE) for medical image retrieval. Following the patch alignment framework, MLLE preserves the geometric structure of the local patch in each feature space according to the LLE criterion. To explore complementary properties among a range of features, MLLE assigns different weights to local patches from different feature spaces. Finally, MLLE employs global coordinate alignment and alternating optimization techniques to learn a smooth low-dimensional embedding from different features. To justify the effectiveness of MLLE for medical image retrieval, we compare it with conventional spectral embedding methods. We conduct experiments on a subset of the IRMA medical image data set. Evaluation results show that MLLE outperforms state-of-the-art dimension reduction methods.