TMaCS: A hybrid template matching and classification system for partially-automated particle selection

Selection of particle images from electron micrographs presents a bottleneck in determining the structures of macromolecular assemblies by single particle electron cryomicroscopy (cryo-EM). The problem is particularly important when an experimentalist wants to improve the resolution of a 3D map by increasing by tens or hundreds of thousands of images the size of the dataset used for calculating the map. Although several existing methods for automatic particle image selection work well for large protein complexes that produce high-contrast images, it is well known in the cryo-EM community that small complexes that give low-contrast images are often refractory to existing automated particle image selection schemes. Here we develop a method for partially-automated particle image selection when an initial 3D map of the protein under investigation is already available. Candidate particle images are selected from micrographs by template matching with template images derived from projections of the existing 3D map. The candidate particle images are then used to train a support vector machine, which classifies the candidates as particle images or non-particle images. In a final step in the analysis, the selected particle images are subjected to projection matching against the initial 3D map, with the correlation coefficient between the particle image and the best matching map projection used to assess the reliability of the particle image. We show that this approach is able to rapidly select particle images from micrographs of a rotary ATPase, a type of membrane protein complex involved in many aspects of biology.     

Identification of Cryptosporidium parvum oocysts by an artificial neural network approach

Microscopic detection of Cryptosporidium parvum oocysts is time-consuming, requires trained analysts, and is frequently subject to significant human errors. Artificial neural networks (ANN) were developed to help identify immunofluorescently labeled C. parvum oocysts. A total of 525 digitized images of immunofluorescently labeled oocysts, fluorescent microspheres, and other miscellaneous nonoocyst images were employed in the training of the ANN. The images were cropped to a 36- by 36-pixel image, and the cropped images were placed into two categories, oocyst and nonoocyst images. The images were converted to grayscale and processed into a histogram of gray color pixel intensity. Commercially available software was used to develop and train the ANN. The networks were optimized by varying the number of training images, number of hidden neurons, and a combination of these two parameters. The network performance was then evaluated using a set of 362 unique testing images which the network had never "seen" before. Under optimized conditions, the correct identification of authentic oocyst images ranged from 81 to 97%, and the correct identification of nonoocyst images ranged from 78 to 82%, depending on the type of fluorescent antibody that was employed. The results indicate that the ANN developed were able to generalize the training images and subsequently discern previously unseen oocyst images efficiently and reproducibly. Thus, ANN can be used to reduce human errors associated with the microscopic detection of Cryptosporidium oocysts.     

Gradient direction dependencies in natural images

We have used information-theoretic measures to compute the amount of dependency which exists between two and three gradient directions at separate locations in an ensemble of natural images. Control experiments were performed on other image classes: phase randomized natural images, whitened natural images and Gaussian noise images. The results show that, for an ensemble of natural images, the amount of 2-point and 3-point gradient direction dependency is equivalent to its ensemble of phase randomized natural images. Therefore, we conclude that the amount of gradient direction dependency in an ensemble of natural images is determined by the ensemble's mean power spectrum rather than the phase spectra of the images. Moreover, this relationship does not extend to individual natural images, the amount of dependency between gradient magnitudes, or gradient directions at high gradient magnitude locations.     

Adaptation Aftereffects in the Perception of Radiological Images

Radiologists must classify and interpret medical images on the basis of visual inspection. We examined how the perception of radiological scans might be affected by common processes of adaptation in the visual system. Adaptation selectively adjusts sensitivity to the properties of the stimulus in current view, inducing an aftereffect in the appearance of stimuli viewed subsequently. These perceptual changes have been found to affect many visual attributes, but whether they are relevant to medical image perception is not well understood. To examine this we tested whether aftereffects could be generated by the characteristic spatial structure of radiological scans, and whether this could bias their appearance along dimensions that are routinely used to classify them. Measurements were focused on the effects of adaptation to images of normal mammograms, and were tested in observers who were not radiologists. Tissue density in mammograms is evaluated visually and ranges from "dense" to "fatty." Arrays of images varying in intermediate levels between these categories were created by blending dense and fatty images with different weights. Observers first adapted by viewing image samples of dense or fatty tissue, and then judged the appearance of the intermediate images by using a texture matching task. This revealed pronounced perceptual aftereffects – prior exposure to dense images caused an intermediate image to appear more fatty and vice versa. Moreover, the appearance of the adapting images themselves changed with prolonged viewing, so that they became less distinctive as textures. These aftereffects could not be accounted for by the contrast differences or power spectra of the images, and instead tended to follow from the phase spectrum. Our results suggest that observers can selectively adapt to the properties of radiological images, and that this selectivity could strongly impact the perceived textural characteristics of the images.     

Identification of Cryptosporidium parvum Oocysts by an Artificial Neural Network Approach

Microscopic detection of Cryptosporidium parvum oocysts is time-consuming, requires trained analysts, and is frequently subject to significant human errors. Artificial neural networks (ANN) were developed to help identify immunofluorescently labeled C. parvum oocysts. A total of 525 digitized images of immunofluorescently labeled oocysts, fluorescent microspheres, and other miscellaneous nonoocyst images were employed in the training of the ANN. The images were cropped to a 36- by 36-pixel image, and the cropped images were placed into two categories, oocyst and nonoocyst images. The images were converted to grayscale and processed into a histogram of gray color pixel intensity. Commercially available software was used to develop and train the ANN. The networks were optimized by varying the number of training images, number of hidden neurons, and a combination of these two parameters. The network performance was then evaluated using a set of 362 unique testing images which the network had never “seen” before. Under optimized conditions, the correct identification of authentic oocyst images ranged from 81 to 97%, and the correct identification of nonoocyst images ranged from 78 to 82%, depending on the type of fluorescent antibody that was employed. The results indicate that the ANN developed were able to generalize the training images and subsequently discern previously unseen oocyst images efficiently and reproducibly. Thus, ANN can be used to reduce human errors associated with the microscopic detection of Cryptosporidium oocysts.     

Techniques of vision: photography, disco and renderings of present perceptions in highland Papua

This article draws on ethnographic and historical materials to interpret several related photographs and ritual images created by Fuyuge-speakers of highland Papua. The images are analysed as techniques of vision. That is, the images are shown to bring forth present perceptions in a timely manner, as evaluated by Fuyuge men and women. It is shown how the representations are informed by distinctive Fuyuge retentions of the past in the present, and extensions of the present in the future. The article suggests that the tendency to 'read' photographic images for their semiotic connotations misses the local perceptions at work in their creation and use. Rather, photographic and ritual images are analogous to the relations of persons that produce them, with their mutually connected temporality.     

On the extraction and alignment of image edges

In this paper the types of image filtering, segmentation and correlational processes which are consistent with the ability of human observers to align edges in arbitrary two-dimensional images are considered. Edge-only versions of a number of images were computed by pre-filtering, segmenting and using laplacians to extract luminance boundaries. The matching of such edge-only images with the originals seems well predicted by a (derivative) matched filtering (cross-correlation) process showing some bias toward edges contained within the middle frequency range of the images used.     

Real-time MRI and articulatory coordination in speech

This paper describes the real-time MRI technique and its use for the study of speech production. The two major problems, (i) the simultaneous recording of the MR images and the speech signal and (ii) the synchronisation of the images and of the speech signal, are addressed. Measurement accuracy on real-time images is evaluated by comparison with similar measurements on static MR images.     

In vivo diffusion-perfusion magnetic resonance imaging of acute cerebral ischemia.

We compared the anatomic extent and severity of ischemic brain injury shown on diffusion-weighted magnetic resonance (MR) images, with cerebral tissue perfusion deficits demonstrated by a nonionic intravascular T2*-shortening magnetic susceptibility contrast agent used in conjunction with standard T2-weighted spin-echo and gradient-echo echo-planar images. Diffusion-weighted images displayed increased signal intensity in the vascular territory of the middle cerebral artery 25-40 min after permanent occlusion, whereas T2-weighted images without contrast were negative or equivocal for at least 2-3 h after stroke was induced. Contrast-enhanced T2-weighted and echo-planar images revealed perfusion deficits that were spatially closely related to the anatomic regions of ischemic tissue injury. These data indicate that diffusion-weighted MR images are very sensitive to early onset pathophysiologic changes induced by acute cerebral ischemia. Combined sequential diffusion-perfusion imaging enables noninvasive in vivo examination of the relationship between hypoperfusion and evolving ischemic brain injury.     

High-speed digital microscopy

High-speed imaging is an ideal technique to accurately resolve the temporal and spatial characteristics of rapid events at either the molecular or cellular level. In this article, the digital imaging techniques used to simultaneously acquire transillumination phase-contrast images, at 240 images s(-1) (high-speed), to characterize ciliary beat frequency, and fluorescence images, at 30 images s(-1) (fast), to measure intracellular calcium concentration ([Ca2+]i), are described. With this technique, a precise correlation between the changes in ciliary beat frequency with changes in [Ca2+]i can be made. Simultaneous imaging is achieved by using different wavelengths of light to form the phase-contrast and fluorescent images and selectively directing these light wavelengths to different cameras with dichroic mirrors and bandpass filters. High-speed images compatible with standard video recording equipment are obtained by prematurely resetting the raster scan of a CCD camera with additional vertical synchronization pulses. The fast [Ca2+]i images are determined using the ratiometric dye fura-2 and a recording technique that monitors rapid changes in fluorescence at a single wavelength and uses intermittent reference images for calibration.     

Accuracy of target delineation by positron emission tomography-based auto-segmentation methods after deformable image registration: A phantom study

PurposeDiagnostic positron emission tomography and computed tomography (PET/CT) images can be fused to the planning CT images by a deformable image registration (DIR). The aim of this study was to evaluate the standardized uptake value (SUV) and target delineation on deformed PET images.MethodsWe used a cylindrical phantom and removable inserts of four spheres (16–38 mm in diameter) and three ellipsoids with a volume equal to the 38-mm-diameter sphere (S38) in each. S38 was filled with 18F-fluorodeoxyglucose activity, and then PET/CT images were acquired. The contours of S38 were generated using original PET images by PET auto-segmentation (PET-AS) methods of (1) SUV2.5, (2) 40% of maximum SUV (SUV40%max), and (3) gradient-based (GB), and were deformed to the other inserts by DIR. We compared the volumes and the SUV_max with the generated contours using the deformed PET images.ResultsThe SUV_max was slightly decreased by DIR; the mean absolute difference was −0.10 ± 0.04. For SUV2.5 and SUV40%max, the differences in S38 volumes between the original and deformed PET images were less than 5%, regardless of deformation type. For the GB, the contoured volumes obtained from deformed PET images were larger than those of the original PET images for the deformation type of ellipsoids. When the S38 was deformed to the 16-mm-diameter sphere, the maximum volume difference was −22.8%.ConclusionsAlthough SUV fluctuations by DIR were negligible, the target delineation on deformed PET images by the GB should be carefully considered owing to the distortion of intensity profiles.     

Stereo super-resolution images detection based on multi-scale feature extraction and hierarchical feature fusion

Recently, with most mobile phones coming with dual cameras, stereo image super-resolution is becoming increasingly popular in phones and other modern acquisition devices, leading stereo super-resolution images spread widely on the Internet. However, current image forensics methods are carried out in monocular images, and high false positive rate appears when detecting stereo super-resolution images by these methods. Therefore, it is important to develop stereo super-resolution image detection method. In this paper, a convolutional neural network with multi-scale feature extraction and hierarchical feature fusion is proposed to detect the stereo super-resolution images. Multi-atrous convolutions are employed to extract multi-scale features and be adapt for varying stereo super-resolution images, and hierarchical feature fusion further improve the performance and robustness of the model. Experimental results demonstrate that the proposed network can detect stereo super-resolution images effectively and achieve strong generalization and robustness. To the best of our knowledge, it is the first attempt to investigate the performance of current forensics methods when tested under stereo super-resolution images, and represent the first study of stereo super-resolution images detection. We believe that it can raise the awareness about the security of stereo super-resolution images.     

基于VTK的医学图像三维可视化系统

医学图像的三维可视化可以通过可视化工具包（VTK）提供的API实现。VTK是医学图像可视化的开法工具包,它把可视化的算法封装起来,利用简单的代码生成所需图形。基于VTK的医学图像三维可视化系统阐述了如何借助VTKAPI读入二维医学图像序列、操作二维图像、重建三维图像以及进行三维图像可视化的全套方案,为临床医生的诊断、治疗提供了有益的途径。     

An automatic method for removing empty camera trap images using ensemble learning

Camera traps often produce massive images, and empty images that do not contain animals are usually overwhelming. Deep learning is a machine‐learning algorithm and widely used to identify empty camera trap images automatically. Existing methods with high accuracy are based on millions of training samples (images) and require a lot of time and personnel costs to label the training samples manually. Reducing the number of training samples can save the cost of manually labeling images. However, the deep learning models based on a small dataset produce a large omission error of animal images that many animal images tend to be identified as empty images, which may lead to loss of the opportunities of discovering and observing species. Therefore, it is still a challenge to build the DCNN model with small errors on a small dataset. Using deep convolutional neural networks and a small‐size dataset, we proposed an ensemble learning approach based on conservative strategies to identify and remove empty images automatically. Furthermore, we proposed three automatic identifying schemes of empty images for users who accept different omission errors of animal images. Our experimental results showed that these three schemes automatically identified and removed 50.78%, 58.48%, and 77.51% of the empty images in the dataset when the omission errors were 0.70%, 1.13%, and 2.54%, respectively. The analysis showed that using our scheme to automatically identify empty images did not omit species information. It only slightly changed the frequency of species occurrence. When only a small dataset was available, our approach provided an alternative to users to automatically identify and remove empty images, which can significantly reduce the time and personnel costs required to manually remove empty images. The cost savings were comparable to the percentage of empty images removed by models.     

Canvases of Political Competition: Image Production as Politics in Tamil Nadu,India

In this article, I explore the production of political images in the Indian state of Tamil Nadu. The state is known for the ubiquitous presence of banners, murals and posters in its public spaces, featuring prominent politicians and actors. It is commonly argued that these images help to convey the heroic or exclusive status of political leaders. However, such images are actually produced by party workers and therefore do not simply transpose status and image. Instead, political supporters praise their leader via these images and act as ‘kingmakers’ in constructing reputation and power. Simultaneously, by putting political images on display, supporters also authorise their own power. While praise is important in showing a person's dedication to a political party, the images, in the motivations of their producers, are suffused with ambivalence and competition as well. Hence, I argue, political image practices are not representative of politics, they are politics.     

基于启发式A^＊算法的超声图像颈动脉内膜提取

从超声图像准确提取颈动脉内膜,为基于颈动脉超声图像判断动脉粥样硬化服务。方法提出一种基于启发式A*算法从超声图像中提取颈动脉内膜边缘的方法。先使用图像分割法区分血管腔和血管壁,再采用结合图像灰度值特点的A*算法准确地提取颈动脉内膜边缘。结果通过对临床采集的32幅颈动脉超声图像的分析研究,表明本方法自动提取的结果与医生手工描绘的结果基本吻合。结论本方法有望应用于超声图像颈动脉内膜的自动提取。     

Using statistical image models for objective evaluation of spot detection in two-dimensional gels

Protein spot detection is central to the analysis of two-dimensional electrophoresis gel images. There are many commercially available packages, each implementing a protein spot detection algorithm. Despite this, there have been relatively few studies comparing the performance characteristics of the different packages. This is in part due to the fact that different packages employ different sets of user-adjustable parameters. It is also partly due to the fact that the images are complex. To carry out an evaluation, "ground truth" data specifying spot position, shape and intensities needs to be defined subjectively on selected test images. We address this problem by proposing a method of evaluation using synthetic images with unambiguous interpretation. The characteristics of the spots in the synthetic images are determined from statistical models of the shape, intensity, size, spread and location of real spot data. The distribution of parameters is described using a Gaussian mixture model obtained from training images. The synthetic images allow us to investigate the effects of individual image properties, such as signal-to-noise ratios and degree of spot overlap, by measuring quantifiable outcomes, e.g. accuracy of spot position, false positive and false negative detection. We illustrate the approach by carrying out quantitative evaluations of spot detection on a number of widely used analysis packages.     

Fusing multi-season UAS images with convolutional neural networks to map tree species in Amazonian forests

Remote sensing images obtained by unoccupied aircraft systems (UAS) across different seasons enabled capturing of species-specific phenological patterns of tropical trees. The application of UAS multi-season images to classify tropical tree species is still poorly understood. In this study, we used RGB images from different seasons obtained by a low-cost UAS and convolutional neural networks (CNNs) to map tree species in an Amazonian forest. Individual tree crowns (ITC) were outlined in the UAS images and identified to the species level using forest inventory data. The CNN model was trained with images obtained in February, May, August, and November. The classification accuracy in the rainy season (November and February) was higher than in the dry season (May and August). Fusing images from multiple seasons improved the average accuracy of tree species classification by up to 21.1 percentage points, reaching 90.5%. The CNN model can learn species-specific phenological characteristics that impact the classification accuracy, such as leaf fall in the dry season, which highlights its potential to discriminate species in various conditions. We produced high-quality individual tree crown maps of the species using a post-processing procedure. The combination of multi-season UAS images and CNNs has the potential to map tree species in the Amazon, providing valuable insights for forest management and conservation initiatives.