How to measure image quality in tissue-based diagnosis (diagnostic surgical pathology)

Background

Automated image analysis, measurements of virtual slides, and open access electronic measurement user systems require standardized image quality assessment in tissue-based diagnosis.

Aims

To describe the theoretical background and the practical experiences in automated image quality estimation of colour images acquired from histological slides.

Theory, material and measurements

Digital images acquired from histological slides should present with textures and objects that permit automated image information analysis. The quality of digitized images can be estimated by spatial independent and local filter operations that investigate in homogenous brightness, low peak to noise ratio (full range of available grey values), maximum gradients, equalized grey value distribution, and existence of grey value thresholds. Transformation of the red-green-blue (RGB) space into the hue-saturation-intensity (HSI) space permits the detection of colour and intensity maxima/minima. The feature distance of the original image to its standardized counterpart is an appropriate measure to quantify the actual image quality. These measures have been applied to a series of H&;E stained, fluorescent (DAPI, Texas Red, FITC), and immunohistochemically stained (PAP, DAB) slides. More than 5,000 slides have been measured and partly analyzed in a time series.

Results

Analysis of H&;E stained slides revealed low shading corrections (10%) and moderate grey value standardization (10 – 20%) in the majority of cases. Immunohistochemically stained slides displayed greater shading and grey value correction. Fluorescent stained slides are often revealed to high brightness. Images requiring only low standardization corrections possess at least 5 different statistically significant thresholds, which are useful for object segmentation. Fluorescent images of good quality only posses one singular intensity maximum in contrast to good images obtained from H&;E stained slides that present with 2 – 3 intensity maxima.

Conclusion

Evaluation of image quality and creation of formally standardized images should be performed prior to automatic analysis of digital images acquired from histological slides. Spatial dependent and local filter operations as well as analysis of the RGB and HSI spaces are appropriate methods to reproduce evaluated formal image quality.

     

A Reference-Free Method for Brightness Compensation and Contrast Enhancement of Micrographs of Serial Sections

Three-dimensional (3D) reconstruction of an organ or tissue from a stack of histologic serial sections provides valuable morphological information. The procedure includes section preparation of the organ or tissue, micrographs acquisition, image registration, 3D reconstruction, and visualization. However, the brightness and contrast through the image stack may not be consistent due to imperfections in the staining procedure, which may cause difficulties in micro-structure identification using virtual sections, region segmentation, automatic target tracing, etc. In the present study, a reference-free method, Sequential Histogram Fitting Algorithm (SHFA), is therefore developed for adjusting the severe and irregular variance of brightness and contrast within the image stack. To apply the SHFA, the gray value histograms of individual images are first calculated over the entire image stack and a set of landmark gray values are chosen. Then the histograms are transformed so that there are no abrupt changes in progressing through the stack. Finally, the pixel gray values of the original images are transformed into the desired ones based on the relationship between the original and the transformed histograms. The SHFA is tested on an image stacks from mouse kidney sections stained with toluidine blue, and captured by a slide scanner. As results, the images through the entire stack reveal homogenous brightness and consistent contrast. In addition, subtle color differences in the tissue are well preserved so that the morphological details can be recognized, even in virtual sections. In conclusion, compared with the existing histogram-based methods, the present study provides a practical method suitable for compensating brightness, and improving contrast of images derived from a large number of serial sections of biological organ.     

Fully automated measurements by light microscopy of tissue sections using a cellular array computer

Software was developed for the acquisition, segmentation and analysis of microscopic OD-images on a VICOM digital image processor, extended with a VISIOMORPH morphoprocessor board. The delineation algorithms for peroxisomes, lysosomes, and nuclei in liver, kidney, and adrenal gland sections start by thresholding the difference between the original image and a low pass filtered version. The resulting binary mask is then processed by morphological operations in order to produce an object overlay. The efficiency of the programs is evaluated by comparing delineated objects at different OD-levels, created by varying the stain or by multiplying the original pixel values with constant factors. Manual delineation on some images is also used as a reference. More complex algorithms are used for the delineation of muscle fibres in ATP-ase-stained sections and immunocytochemically labelled cells in monolayer preparations. Muscle images from parallel sections with different stainings are matched with a coordinate transform, enabling the transfer of the object mask from a single delineated image to the unprocessed images and thus obtain all necessary information for fibre classification. After segmentation, the OD-images and their object overlays are fed into a data extraction program, measuring for each delineated object user-selected features. Data are sent to a VAX for statistical interpretation.     

Simple method for quantification of fast plasma membrane movements.

We present a method for the quantification of the fast plasma membrane movements that are involved in ruffling, blebbing, fast shape change, and fast translocation. The method is based on the Kontron Vidas image analysis computer program. Video images from cells viewed through an inverted microscope were transmitted to the computer. The procedure was as follows: 4 consecutive video images were averaged (image 1); 28 s later a second set of 4 video images was averaged (image 2); image 2 was subtracted from image 1 and the grey level of each pixel of the resulting image was increased with 128 grey level units, resulting in the subtraction image, showing a uniform grey background speckled with brighter and darker spots corresponding to areas of movement. These spots were discriminated and turned into white objects against a black background. Interactive editing was used to delete artefacts that resulted from floating debris. The total area of the discriminated objects was measured, and the parameter motile area in micron2 per cell was calculated. We have applied our method to the study of motility induced in epithelial cell lines by the tumor promoter 12-O-tetradecanoyl-phorbol-13-acetate and by epidermal growth factor.     

基于颜色名和OpponentSIFT特征的鳞翅目昆虫图像识别

【目的】本研究旨在探索使用先进的计算机视觉技术实现对昆虫图像的自动分类方法。【方法】通过预处理对采集的昆虫标本图像去除背景,获得昆虫图像的前景蒙板,并由蒙板确定的轮廓计算出前景图像的最小包围盒,剪切出由最小包围盒确定的前景有效区域,然后对剪切得到的图像进行特征提取。首先提取颜色名特征,把原来的RGB（Red-Green-Blue）图像的像素值映射到11种颜色名空间,其值表示RGB值属于该颜色名的概率,每个颜色名平面划分成3×3像素大小的网格,用每格的概率均值作为网格中心点的描述子,最后用空阈金字塔直方图统计的方式形成颜色名视觉词袋特征;其次提取OpponentSIFT(Opponent Scale Invariant Feature Transform)特征,首先把RGB图像变换到对立色空间,对该空间每通道提取SIFT特征,最后用空域池化和直方图统计方法形成OpponentSIFT视觉词袋。将两种词袋特征串接后得到该昆虫图像的特征向量。使用昆虫图像样本训练集提取到的特征向量训练SVM（Support Vector Machine）分类器,使用这些训练得到的分类器即可实现对鳞翅目昆虫的分类识别。【结果】该方法在包含10种576个样本的昆虫图像数据库中进行了测试,取得了100%的识别正确率。【结论】试验结果证明基于颜色名和OpponentSIFT特征可以有效实现对鳞翅目昆虫图像的识别。     

Computer-assisted image analysis of the distributions of peptidergic terminals in the central nucleus of the amygdala: A preliminary study

When viewed under dark-field illumination, peptidergic terminals in sections stained by the Sternberger PAP immunocytochemical method are seen as individual points of light. Under high magnification, the degree of brightness of various areas of immunoreactive terminals is seen to be a function of the density of terminals in these areas. By analyzying the relative brightness of the immunostained central nucleus of the amygdala (CNA) with an EyeCom II PDP-$\text{11}\text{34}$ image analysis system, we have obtained a relative evaluation of the density distribution of neurotensin (NT)-, substance P (SP), VIP-, angiotensin II (AII), m-enkephalin (m-ENK) and somatostatin (SS)-immunoreactive terminals in terms of normal morphology and following a brain lesion. The EyeCom II system divides the presented image into 307200 picture elements (pixels) and assigns one of 256 grey values to the average brightness with each pixel. We have aggregated the grey level frequencies into 5 levels where level 1 corresponds to the highest terminal density and level 5 to the lowest density. At level 1, only NT- and VIP-immunoreactive terminals occupy a significant percentage of the cross-sectional area of the CNA (20%). About 15% of the area of the CNA has VIP terminals with level 5 density. The distributions of the top 20% of the terminal density range of NT, SP, AII and VIP support a classical medial/lateral division of the nucleus. The distribution of the same range of SS- and ENK terminals suggests a dorsoventral division of the CNA. A preliminary study indicates that comparison of grey level frequency histograms generated by image analysis from homologous lesioned and unlesioned sections of the CNA can yield useful information regarding post-lesion changes in the distribution of immunoreactive terminals.     

基于对象特征的山东省丘陵地区多时相遥感土地覆被自动分类

对于基于像元的土地覆被分类来说,植被的分类是难点。使用多时相面向对象分类方法可以较好的解决这个问题。以山东省烟台市丘陵地区为研究区,采用Landsat TM(Landsat Thematic Mapper remotely sensed imagery)、DEM(Digital Elevation Model)、坡度、坡位、坡向等多种数据,利用基于对象特征的多时相分类方法对研究区进行土地覆盖自动分类。首先对影像进行多尺度分割并检验分割结果选取合适的分割尺度,然后分析对象的光谱、纹理、形状特征。根据各类地物的光谱特征、地理相关性、形状、空间分布等特征,明确类别之间的差异。建立决策树使用隶属度函数进行模糊分类,借助支持向量机提高分类精度。研究结果表明,通过使用多时相影像采用面向对象分类方法,相对于传统的基于像素的分类可以明显提高分类精度,尤其是解决了乔灌草的区分问题。     

A Modified DWT-SVD Algorithm for T1-w Brain MR Images Contrast Enhancement

BackgroundImage contrast enhancement is considered as the most useful technique permitting a better appearance of the low contrast images. This paper presents a modified Discret Wavelet Transform - Singular Value Decomposition (DWT-SVD) approach for the enhancement of low contrast Brain MR Images used for brain tissues exploration.MethodsThe proposed technique is processed as follows: first of all, we consider low contrast T1-weighted MRI slices as input images (A1) on which we apply General Histogram Equalization (GHE) algorithm to have equalized images referred as (A2) having zero as mean and one as variance. Secondly, by using the Discrete Wavelet Transform (DWT) algorithm, both (A1) and (A2) are divided into low and high frequency sub-bands. On the low frequency (LL1) and (LL2) sub-bands, SVD is processed in order to generate three matrix factorization (U, V and Σ) where the maximums of (U) and (V) matrix are used for the estimation of a correction coefficient (ξ). Our contribution in this paper is to estimate the new singular value matrix (New Σ) using a weighted sum of both original and equalized singular value matrix thanks to an adjustable parameter (μ) for the targeted low contrast images. This parameter, ranged between 0.05 and 0.95, is determined empirically according to the input low contrast image. Finally, the enhanced resulting image is easily reconstructed using the Inverse SVD (ISVD) and the Inverse DWT (IDWT) processes.ResultsThe database considered in our research consisted of 120 MR brain images where T1-weighted MR brain modality are selected for the contrast enhancement process. Considering the qualitative results, our proposed contrast enhancement method have shown better distinction between brain tissues and have preserved all White Matter (WM), Gray Matter (GM) and Cerebro-Spinal Fluid (CSF) pixel edges. In fact, histogram plots of images enhanced by proposed method covered all the gray level intensities. For the quantitative results, proposed method gives the highest PSNR, QRCM, SSIM, FSIM and EME values and the lowest AMBE values for (μ) equal to 0.65 as comparing to the rest of methods. These results signifies that proposed contrast enhancement method have provided greater image quality with preservation of image structure, feature and brightness.ConclusionProposed method improved performance of contrast enhancement image without creating unwanted artifact and without destroying image edge information or affecting the specificities of brain tissues. This is due to the use of an empirically (μ) parameter adjustable according to the input MR images. Hence, the proposed approach is appropriate for enhancing contrast of huge type of low contrast images.     

Combinatorial image analysis of DNA microarray features

MOTIVATION: DNA and protein microarrays have become an established leading-edge technology for large-scale analysis of gene and protein content and activity. Contact-printed microarrays has emerged as a relatively simple and cost effective method of choice but its reliability is especially susceptible to quality of pixel information obtained from digital scans of spotted features in the microarray image. RESULTS: We address the statistical computation requirements for optimizing data acquisition and processing of digital scans. We consider the use of median filters to reduce noise levels in images and top-hat filters to correct for trends in background values. We also consider, as alternative estimators of spot intensity, discs of fixed radius, proportions of histograms and k-means clustering, either with or without a square-root intensity transformation and background subtraction. We identify, using combinatoric procedures, optimal filter and estimator parameters, in achieving consistency among the replicates of a gene on each microarray. Our results, using test data from microarrays of HCMV, indicate that a highly effective approach for improving reliability and quality of microarray data is to apply a 21 by 21 top-hat filter, then estimate spot intensity as the mean of the largest 20% of pixel values in the target region, after a square-root transformation, and corrected for background, by subtracting the mean of the smallest 70% of pixel values. AVAILABILITY: Fortran90 subroutines implementing these methods are available from the authors, or at http://www.bioss.ac.uk/~chris.     

Imaging technique implemented in CellTracks system

BACKGROUND: We developed the CellTracks cell analysis system that, similar to flow cytometry, yields multiparameter information by which the cells can be differentiated. We describe the implementation of a laser scanning imaging method in the system. Image analysis of the cells improves the specificity of cell classification, especially in cases where the particular cells are found relatively infrequently and one has to discriminate between artifacts and real events. METHODS: Fluorescent images of immunomagnetically labeled and aligned cells are obtained by passing the cells through a laser focus. The laser focus is smaller than the objects and subsequent frames captured by a regular surveillance CCD camera with a frame grabber board represent different parts of the cells. Complete images of the cells are constructed by shifting each image with respect to each other and adding individual pixel values. RESULTS: The power of combining a fluorescent image with multiparametric data is demonstrated by imaging fluorescent and magnetically labeled beads and cells. The image gives additional information about the dye distribution across the objects. Changes in dye distribution as a function of time were observed in leukocytes labeled with the red fluorescent label, Oxazine750, which are imaged at different time intervals. CONCLUSIONS: An imaging technique implemented in the CellTracks system provides high-resolution fluorescent images of events previously identified by the system. The images of the fluorescent cells enhance the ability to classify rare events.     

Removing high contrast artifacts via digital inpainting in cryo-electron tomography: an application of compressed sensing

To cope with poor quality in cryo-electron tomography images, electron-dense markers, such as colloidal goldbeads, are often used to assist image registration and analysis algorithms. However, these markers can create artifacts that occlude a specimen due to their high contrast, which can also cause failure of some image processing algorithms. One way of reducing these artifacts is to replace high contrast objects with pixel densities that blend into the surroundings in the projection domain before volume reconstruction. In this paper, we propose digital inpainting via compressed sensing (CS) as a new method to achieve this goal. We show that cryo-ET projections are sparse in the discrete cosine transform (DCT) domain, and, by finding the sparsest DCT domain decompositions given uncorrupted pixels, we can fill in the missing pixel values that are occluded by high contrast objects without discontinuities. Our method reduces visual artifacts both in projections and in tomograms better than conventional algorithms, such as polynomial interpolation and random noise inpainting.     

Pixel-based analysis of multiple images for the identification of changes: a novel approach applied to unravel proteome patterns [corrected] of 2-D electrophoresis gel images

A novel approach for revealing patterns of proteome variation among series of 2-DE gel images is presented. The approach utilises image alignment to ensure that each pixel represents the same information across all gels. Gel images are normalised, and background corrected, followed by unfolding of the images to 1-D pixel vectors and analysing pixel vectors by multivariate data modelling. Information resulting from the data analysis is refolded back to the image domain for visualisation and interpretation. The method is rapid and suitable for automatic routines applied after the gel alignment. The approach is compared with spot volume analysis to illustrate how this approach can solve persistent problems like mismatch of protein spots, erroneous missing values and failure to detect variation in overlapping proteins. The method may also detect variation in the border area of saturated proteins. The approach is given the name pixel-based analysis of multiple images for the identification of changes (PMC). The method can be used for multiple images in general. Effects of pretreatment of the images are discussed.     

Multispectral/hyperspectral image enhancement for biological cell analysis.

BACKGROUND: Microscopes form projected images from illuminated objects, such as cellular tissue, which are recorded at a distance through the optical system's field of view. A telescope on a satellite or airplane also forms images with a similar optical projection of objects on the ground. Typical visible illuminations form a displayed set of three-color channels (Red Green Blue [RGB]) that are combined from three image sensor arrays (e.g., focal plane arrays) into a single pixel coding for each color present in the image. Analysis of these RGB color images develops a qualitative image representation of the objects. METHODS: Independent component analysis (ICA) is used for analysis and enhancement of multispectral images, and compared with the similar and widely used principal component analysis. RESULTS: The data examples indicate that the ICA enhancement, and the resulting RGB image combination display, can be useful in processing datacubes of cellular data where isolation of unknown subtle image elements representing objects is desired. CONCLUSIONS: ICA image enhancement can aid processing of datacubes of cellular data by clarifying subtle image elements. These parallelizable algorithms can be implemented for real-time, online analysis.     

A Slantlet transform based intelligent system for magnetic resonance brain image classification

The present paper proposes the development of a new approach for automated diagnosis, based on classification of magnetic resonance (MR) human brain images. Wavelet transform based methods are a well-known tool for extracting frequency space information from non-stationary signals. In this paper, the proposed method employs an improved version of orthogonal discrete wavelet transform (DWT) for feature extraction, called Slantlet transform, which can especially be useful to provide improved time localization with simultaneous achievement of shorter supports for the filters. For each two-dimensional MR image, we have computed its intensity histogram and Slantlet transform has been applied on this histogram signal. Then a feature vector, for each image, is created by considering the magnitudes of Slantlet transform outputs corresponding to six spatial positions, chosen according to a specific logic. The features hence derived are used to train a neural network based binary classifier, which can automatically infer whether the image is that of a normal brain or a pathological brain, suffering from Alzheimer's disease. An excellent classification ratio of 100% could be achieved for a set of benchmark MR brain images, which was significantly better than the results reported in a very recent research work employing wavelet transform, neural networks and support vector machines.     

The importance of optical optimization in whole slide imaging (WSI) and digital pathology imaging

In the last 10 years, whole slide imaging (WSI) has seen impressive progress not only in image quality and scanning speed but also in the variety of systems available to pathologists. However, we have noticed that most systems have relatively simple optics axes and rely on software to optimize image quality and colour balance. While much can be done in software, this study examines the importance of optics, in particular optical filters, in WSI.Optical resolution is a function of the wavelength of light used and the numerical aperture of the lens system (Resolution = (f) wavelength/2 NA). When illumining light is not conditioned correctly with filters, there is a tendency for the wavelength to shift to longer values (more red) because of the characteristics of the lamps in common use. Most microscopes (but remarkably few WSI devices) correct for this with ND filter for brightness and Blue filter (depends on the light source) for colour correction.Using H&E slides research microscopes (Axiophot, Carl Zeiss MicroImaging, Inc. NY. Eclipse 50i., Nikon Inc. NY) at 20x, an attached digital camera (SPOT RT741 Slider Color, Diagnosis Instruments., MI USA), and a filter set, we examined the effect of filters and software enhancement on digital image quality. The focus value (as evaluated by focus evaluation software developed in house and SPOT imaging Software v4.6) was used as a proxy for image quality. Resolution of tissue features was best with the use of both the Blue and ND filters (in addition to software enhancement). Images without filters but with software enhancement while superficially good, lacked some details of specimen morphology and were unclear compared with the images with filters.The results indicate that the appropriate use of optical filters could measurably improve the appearance and resolution of WSI images.     

Stereo disparity computation using Gabor filters

A solution to the correspondence problem for stereopsis is proposed using the differences in the complex phase of local spatial frequency components. One-dimensional spatial Gabor filters (Gabor 1946; Marcelja 1980), at different positions and spatial frequencies are convolved with each member of a stereo pair. The difference between the complex phase at corresponding points in the two images is used to find the stereo disparity. Disparity values are combined across spatial frequencies for each image location. Three-dimensional depth maps have been computed from real images under standard lighting conditions, as well as from random-dot stereograms (Julesz 1971). The algorithm can discriminate disparities significantly smaller than the width of a pixel. It is possible that a similar mechanism might be used in the human visual system.     

A secure lossless recovery for medical images based on image encoding and data self-embedding

To protect the security of data outsourced to the cloud, the tampers detection and recovery for outsourced image have aroused the concern of people. A secure tampering detection and lossless recovery for medical images (MI) using permutation ordered binary (POB) number system is proposed. In the proposed scheme, the region of interest (ROI) of MI is first extracted, and then, ROI is divided into some no-overlapping blocks, and image encoding is conducted on these blocks based on the better compression performance of JPEG-LS for medical image. After that, the generated compression data by all the blocks are divided into high 4-bit and low 4-bit planes, and shuffling and combination are used to generate two plane images. Owing to the substantial redundancies space in the compressed data, the data of each plane are spread to the size of the original image. Lastly, authentication data of two bits is obtained for every pixel and inserted into the pixel itself within the each plane, and the corresponding 10-bit data is transformed into the POB value of 8-bit. Furthermore, encryption is implemented on the above image to produce two shares which can be outsourced to the cloud server. The users can detect tampered part and recover original image when they down load the shares from the cloud. Extensive experiments on some ordinary medical image and COVID-19 image datasets show that the proposed approach can locate the tampered parts within the MI, and the original MI can be recovered without any loss even if one of the shares are totally destroyed, or two shares are tampered at the ration not more than 50%. Some comparisons and analysis are given to show the better performance of the scheme.

     

Imaging obstructed ventilation with NMR using inert fluorinated gases.

We partially obstructed the left bronchi of rats and imaged an inert insoluble gas, SF(6), in the lungs with NMR using a technique that clearly differentiates obstructed and normal ventilation. When the inhaled fraction of O(2) is high, SF(6) concentrates dramatically in regions of the lung with low ventilation-to-perfusion ratios (VA/Q); therefore, these regions are brighter in an image than where VA/Q values are normal or high. A second image, made when the inhaled fraction of O(2) is low, serves as a reference because the SF(6) fraction is nearly uniform, regardless of VA/Q. The quotient of the first and second images displays the low-VA/Q regions and is corrected for other causes of brightness variation. The technique may provide sufficient quantification of VA/Q to be a useful research tool. The noise in the quotient image is described by the probability density function for the quotient of two normal random variables. When the signal-to-noise ratio of the denominator image is >10, the signal-to-noise ratio of the quotient image is similar to that of the parent images and decreases with pixel value.