Image-processing technique for suppressing ribs in chest radiographs by means of massive training artificial neural network (MTANN)

When lung nodules overlap with ribs or clavicles in chest radiographs, it can be difficult for radiologists as well as computer-aided diagnostic (CAD) schemes to detect these nodules. In this paper, we developed an image-processing technique for suppressing the contrast of ribs and clavicles in chest radiographs by means of a multiresolution massive training artificial neural network (MTANN). An MTANN is a highly nonlinear filter that can be trained by use of input chest radiographs and the corresponding "teaching" images. We employed "bone" images obtained by use of a dual-energy subtraction technique as the teaching images. For effective suppression of ribs having various spatial frequencies, we developed a multiresolution MTANN consisting of multiresolution decomposition/composition techniques and three MTANNs for three different-resolution images. After training with input chest radiographs and the corresponding dual-energy bone images, the multiresolution MTANN was able to provide "bone-image-like" images which were similar to the teaching bone images. By subtracting the bone-image-like images from the corresponding chest radiographs, we were able to produce "soft-tissue-image-like" images where ribs and clavicles were substantially suppressed. We used a validation test database consisting of 118 chest radiographs with pulmonary nodules and an independent test database consisting of 136 digitized screen-film chest radiographs with 136 solitary pulmonary nodules collected from 14 medical institutions in this study. When our technique was applied to nontraining chest radiographs, ribs and clavicles in the chest radiographs were suppressed substantially, while the visibility of nodules and lung vessels was maintained. Thus, our image-processing technique for rib suppression by means of a multiresolution MTANN would be potentially useful for radiologists as well as for CAD schemes in detection of lung nodules on chest radiographs.     

基于小波变换的多神经网络胸片骨骼分离

胸片中,因大量肺结点被锁骨或肋骨遮挡而被放射科医生忽略。为了从胸片图像中分割出骨骼结构,提出了一种基于小波变换的多分辨率人工神经网络,以获取去除骨骼结构的虚拟软组织胸片。该方法可有效保证肺结点与血管的清晰度,且分离出骨骼和软组织可有效地帮助放射医生检测肺结点。     

Toward automated segmentation of the pathological lung in CT

Conventional methods of lung segmentation rely on a large gray value contrast between lung fields and surrounding tissues. These methods fail on scans with lungs that contain dense pathologies, and such scans occur frequently in clinical practice. We propose a segmentation-by-registration scheme in which a scan with normal lungs is elastically registered to a scan containing pathology. When the resulting transformation is applied to a mask of the normal lungs, a segmentation is found for the pathological lungs. As a mask of the normal lungs, a probabilistic segmentation built up out of the segmentations of 15 registered normal scans is used. To refine the segmentation, voxel classification is applied to a certain volume around the borders of the transformed probabilistic mask. Performance of this scheme is compared to that of three other algorithms: a conventional, a user-interactive and a voxel classification method. The algorithms are tested on 10 three-dimensional thin-slice computed tomography volumes containing high-density pathology. The resulting segmentations are evaluated by comparing them to manual segmentations in terms of volumetric overlap and border positioning measures. The conventional and user-interactive methods that start off with thresholding techniques fail to segment the pathologies and are outperformed by both voxel classification and the refined segmentation-by-registration. The refined registration scheme enjoys the additional benefit that it does not require pathological (hand-segmented) training data.     

Automatic detection of abnormalities in chest radiographs using local texture analysis

A fully automatic method is presented to detect abnormalities in frontal chest radiographs which are aggregated into an overall abnormality score. The method is aimed at finding abnormal signs of a diffuse textural nature, such as they are encountered in mass chest screening against tuberculosis (TB). The scheme starts with automatic segmentation of the lung fields, using active shape models. The segmentation is used to subdivide the lung fields into overlapping regions of various sizes. Texture features are extracted from each region, using the moments of responses to a multiscale filter bank. Additional "difference features" are obtained by subtracting feature vectors from corresponding regions in the left and right lung fields. A separate training set is constructed for each region. All regions are classified by voting among the k nearest neighbors, with leave-one-out. Next, the classification results of each region are combined, using a weighted multiplier in which regions with higher classification reliability weigh more heavily. This produces an abnormality score for each image. The method is evaluated on two databases. The first database was collected from a TB mass chest screening program, from which 147 images with textural abnormalities and 241 normal images were selected. Although this database contains many subtle abnormalities, the classification has a sensitivity of 0.86 at a specificity of 0.50 and an area under the receiver operating characteristic (ROC) curve of 0.820. The second database consist of 100 normal images and 100 abnormal images with interstitial disease. For this database, the results were a sensitivity of 0.97 at a specificity of 0.90 and an area under the ROC curve of 0.986.     

Cooperation of color pixel classification schemes and color watershed: a study for microscopic images

We study the ability of the cooperation of two-color pixel classification schemes (Bayesian and K-means classification) with color watershed. Using color pixel classification alone does not sufficiently accurately extract color regions so we suggest to use a strategy based on three steps: simplification, classification, and color watershed. Color watershed is based on a new aggregation function using local and global criteria. The strategy is performed on microscopic images. Quantitative measures are used to evaluate the resulting segmentations according to a learning set of reference images.     

Shape-based averaging.

A new method for averaging multidimensional images is presented, which is based on signed Euclidean distance maps computed for each of the pixel values. We refer to the algorithm as "shape-based averaging" (SBA) because of its similarity to Raya and Udupa's shape-based interpolation method. The new method does not introduce pixel intensities that were not present in the input data, which makes it suitable for averaging nonnumerical data such as label maps (segmentations). Using segmented human brain magnetic resonance images, SBA is compared to label voting for the purpose of averaging image segmentations in a multiclassifier fashion. SBA, on average, performed as well as label voting in terms of recognition rates of the averaged segmentations. SBA produced more regular and contiguous structures with less fragmentation than did label voting. SBA also was more robust for small numbers of atlases and for low atlas resolutions, in particular, when combined with shape-based interpolation. We conclude that SBA improves the contiguity and accuracy of averaged image segmentations.     

基于SVM后验概率的红外弱小目标检测

针对复杂云层背景中背景边缘干扰严重的问题,提出基于支持向量机(SVM)后验概率的红外弱小目标检测算法。该算法将红外弱小目标检测视作目标与背景的二分类问题,根据红外图像特性,以各像素点8个方向的梯度作为目标和背景的分类依据,选取能够表现目标和背景特征的梯度作为SVM训练样本的主要参考量,设定训练集,并通过训练获得SVM分类模型。基于SVM后验概率的检测算法将待测样本各像素点的8方向梯度作用于分类模型,获得的SVM后验概率作为检测输出。实验结果证明了该算法的有效性。     

Anatomic region-based dynamic range compression for chestradiographs using warping transformation of correlated distribution

The purpose of this paper is to investigate the effectiveness of the authors' novel dynamic range compression (DRC) for chest radiographs. The purpose of DRC is to compress the gray scale range of the image when using narrow dynamic range viewing systems such as monitors. First, an automated segmentation method was used to detect the lung region. The combined region of mediastinum, heart, and subdiaphragm was defined based on the lung region. The correlated distributions, between a pixel value and its neighboring averaged pixel value, for the lung region and the combined region were calculated. According to the appearance of overlapping of two distributions, the warping function was decided. After pixel values were warped, the pixel value range of the lung region was compressed while preserving the detail information, because the warping function compressed the range of the averaged pixel values while preserving the pixel value range for the pixels which had had the same averaged pixel value. The performance was evaluated with the authors' criterion function which was the contrast divided by the moment, where the contrast and the moment represent the sum of the differences between the pixel values and the averaged values of eight pixels surrounding that pixel, and the sum of the differences between the pixel values and the averaged value of all pixels in the region-of-interest, respectively. For 71 screening chest images from Johns Hopkins University Hospital (Baltimore, MD), this method improved our criterion function at 11.7% on average. The warping transformation algorithm based on the correlated distribution was effective in compressing the dynamic range while simultaneously preserving the detail information     

基于高斯曲面阈值的CR数字胸片图像分割

以CR数字胸片图像为研究对象,提出基于高斯曲面阈值法的分割算法,它方便后继处理,可以得到比较完整的肋骨信息,为后期的计算机辅助诊断提供更为可靠的实验数据,是一种更为有效的算法.     

Automatic detection of rib borders in chest radiographs

An algorithm for detection of posterior rib borders in chest radiographs is presented. The algorithm first determines the thoracic cage boundary to restrict the area of search for the ribs. It then finds approximate rib borders using a knowledge-based Hough transform. Finally, the algorithm localizes the rib borders using an active contour model. Results of the proposed rib finding algorithm on 10 chest radiographs are presented.     

Efficient segmentation by sparse pixel classification

Segmentation methods based on pixel classification are powerful but often slow. We introduce two general algorithms, based on sparse classification, for optimizing the computation while still obtaining accurate segmentations. The computational costs of the algorithms are derived, and they are demonstrated on real 3-D magnetic resonance imaging and 2-D radiograph data. We show that each algorithm is optimal for specific tasks, and that both algorithms allow a speedup of one or more orders of magnitude on typical segmentation tasks.     

基于点特异度和自适应分类策略的眼底图像分割方法

提出基于点特异度和自适应分类策略的血管分割方法（SSVD, specificity and self-adaptive vessel detection）,首先给出点特异度的定义,通过设置高点特异度阈值,实现主血管的提取,然后由多主体进行自适应像素分类,将每个未确定像素作为一个Agent,在多尺度点特异度阈值范围内,根据邻域Agent状态修订自身状态,逐步完成对像素的分类,最后通过多窗口去噪对噪音进行滤除完成对图像血管结构的分割。将SSVD方法应用到DRIVE数据库眼底图像的血管分割中,实验结果表明该方法要比现有其他方法具有更高的准确度和效率。     

Automatic lung segmentation for accurate quantitation of volumetricX-ray CT images

Segmentation of pulmonary X-ray computed tomography (CT) images is a precursor to most pulmonary image analysis applications. This paper presents a fully automatic method for identifying the lungs in three-dimensional (3-D) pulmonary X-ray CT images. The method has three main steps. First, the lung region is extracted from the CT images by gray-level thresholding. Then, the left and right lungs are separated by identifying the anterior and posterior junctions by dynamic programming. Finally, a sequence of morphological operations is used to smooth the irregular boundary along the mediastinum in order to obtain results consistent with those obtained by manual analysis, in which only the most central pulmonary arteries are excluded from the lung region. The method has been tested by processing 3-D CT data sets from eight normal subjects, each imaged three times at biweekly intervals with lungs at 90% vital capacity. We present results by comparing our automatic method to manually traced borders from two image analysts. Averaged over all volumes, the root mean square difference between the computer and human analysis is 0.8 pixels (0.54 mm). The mean intrasubject change in tissue content over the three scans was 2.75% +/- 2.29% (mean +/- standard deviation).     

Super-resolution target identification from remotely sensed imagesusing a Hopfield neural network

Fuzzy classification techniques have been developed recently to estimate the class composition of image pixels, but their output provides no indication of how these classes are distributed spatially within the instantaneous field of view represented by the pixel. As such, while the accuracy of land cover target identification has been improved using fuzzy classification, it remains for robust techniques that provide better spatial representation of land cover to be developed. Such techniques could provide more accurate land cover metrics for determining social or environmental policy, for example. The use of a Hopfield neural network to map the spatial distribution of classes more reliably using prior information of pixel composition determined from fuzzy classification was investigated. An approach was adopted that used the output from a fuzzy classification to constrain a Hopfield neural network formulated as an energy minimization tool. The network converges to a minimum of an energy function, defined as a goal and several constraints. Extracting the spatial distribution of target class components within each pixel was, therefore, formulated as a constraint satisfaction problem with an optimal solution determined by the minimum of the energy function. This energy minimum represents a “best guess” map of the spatial distribution of class components in each pixel. The technique was applied to both synthetic and simulated Landsat TM imagery, and the resultant maps provided an accurate and improved representation of the land covers studied, with root mean square errors (RMSEs) for Landsat imagery of the order of 0.09 pixels in the new fine resolution image recorded     

Computer Classification of Pneumoconiosis from Radiographs of Coal Workers

The accurate categorization of profusion of opacities in radiographs of coal workers is a significant medical problem. In this study, the feasibility of computer classification of profusion was investigated. Standard pattern recognition techniques were used except for the spatial moments which were computed as measurements of the texture patterns. A normal-abnormal classification was performed on 178 zonal samples and resulted in a training classification rate of 99% and a testing rate of 97%. A four category classification was also performed for the zonal samples with a correct classification rate of 84%. The zonal decisions were used to obtain overall film profusion. The results of this classification compared favorably with readings by radiologists. This study provides positive evidence for a quantitative approach to the classification of profusion. The significance of this study with respect to the understanding and measurement of lung pathology from radiographs is that an alternative or supplement to the presently used visual analysis is demonstrated.     

A Shift-Invariant Neural Network for the Lung Field Segmentation in Chest Radiography

We have developed a computerized method using a neural network for the segmentation of lung fields in chest radiography. The lung is the primary region of interest in routine chest radiography diagnosis. Since computer is expected to perform disease pattern search automatically, it is important to design appropriate algorithms to delineate the region of interest. A reliable segmentation method is essential to facilitate subsequent searches for image patterns associated with lung diseases. In this study, we employed a shift invariant neural network coupled with error back-propagation training method to extract the lung fields. A set of computer algorithms were also developed for smoothing the initially detected edges of lung fields. Our preliminary results indicated that 86% of the segmented lung fields globally matched the original chest radiographs. We also found that the method facilitates the development of computer algorithms in the field of computer-aided diagnosis.     

A sequence-based extension of mean-field annealing using the forward/backward algorithm: application to image segmentation

Optimization based on mean-field theory, including mean-field annealing (MFA), is widely used for discrete optimization (label assignment) problems defined on the pixel sites of an image. One formulation of MFA is via maximum entropy, where one seeks the joint distribution over the (random) assignments subject to an average level of cost. MFA is obtained by assuming the assignments at each pixel are statistically independent, given the observed image. Alternatively, we make the less restrictive assumption of independent row labelings. The independence assumption means that at each step, MFA optimizes over only one pixel, whereas our method jointly optimizes over an entire row, i.e., our method is less greedy. In principle, an MFA extension could be developed that explicitly re-estimates the row labeling distributions, but such an approach is, in practice, infeasible. Even so, we can indirectly implement this re-estimation, by re-estimating quantities that determine the row labeling distributions. These quantities are the a posteriori site probabilities, re-estimable via the well-known forward/backward (F/B) algorithm. Thus, our algorithm, which descends in the ME Lagrangian/free energy, consists of iterative application of F/B to the image rows (columns). At convergence, maximum a posteriori site labeling is performed. Our method was applied to segmentation of both real and synthetic noise-corrupted images. It achieved lower Markov random field model potentials and better segmentations compared with other methods, and, in high noise, standard MFA.     

Performance-based classifier combination in atlas-based image segmentation using expectation-maximization parameter estimation

It is well known in the pattern recognition community that the accuracy of classifications obtained by combining decisions made by independent classifiers can be substantially higher than the accuracy of the individual classifiers. We have previously shown this to be true for atlas-based segmentation of biomedical images. The conventional method for combining individual classifiers weights each classifier equally (vote or sum rule fusion). In this paper, we propose two methods that estimate the performances of the individual classifiers and combine the individual classifiers by weighting them according to their estimated performance. The two methods are multiclass extensions of an expectation-maximization (EM) algorithm for ground truth estimation of binary classification based on decisions of multiple experts (Warfield et al., 2004). The first method performs parameter estimation independently for each class with a subsequent integration step. The second method considers all classes simultaneously. We demonstrate the efficacy of these performance-based fusion methods by applying them to atlas-based segmentations of three-dimensional confocal microscopy images of bee brains. In atlas-based image segmentation, multiple classifiers arise naturally by applying different registration methods to the same atlas, or the same registration method to different atlases, or both. We perform a validation study designed to quantify the success of classifier combination methods in atlas-based segmentation. By applying random deformations, a given ground truth atlas is transformed into multiple segmentations that could result from imperfect registrations of an image to multiple atlas images. In a second evaluation study, multiple actual atlas-based segmentations are combined and their accuracies computed by comparing them to a manual segmentation. We demonstrate in both evaluation studies that segmentations produced by combining multiple individual registration-based segmentations are more accurate for the two classifier fusion methods we propose, which weight the individual classifiers according to their EM-based performance estimates, than for simple sum rule fusion, which weights each classifier equally.