Continual improvement of nasopharyngeal carcinoma segmentation with less labeling effort

PurposeConvolutional neural networks (CNNs) offer a promising approach to automated segmentation. However, labeling contours on a large scale is laborious. Here we propose a method to improve segmentation continually with less labeling effort.MethodsThe cohort included 600 patients with nasopharyngeal carcinoma. The proposed method was comprised of four steps. First, an initial CNN model was trained from scratch to perform segmentation of the clinical target volume. Second, a binary classifier was trained using a secondary CNN to identify samples for which the initial model gave a dice similarity coefficient (DSC) < 0.85. Third, the classifier was used to select such samples from the new coming data. Forth, the final model was fine-tuned from the initial model, using only selected samples.ResultsThe classifier can detect poor segmentation of the model with an accuracy of 92%. The proposed segmentation method improved the DSC from 0.82 to 0.86 while reducing the labeling effort by 45%.ConclusionsThe proposed method reduces the amount of labeled training data and improves segmentation by continually acquiring, fine-tuning, and transferring knowledge over long time spans.     

Hybrid Nature Inspired SMO-GBM Classifier for Exudate Classification on Fundus Retinal Images

Background: The diabetic retinopathy can result in loss of vision if not detected in the earlier stages. Exudates are the lesions which play a crucial role in early diagnosis of diabetic retinopathy. The localization of exudates lesions with high values of performance metrics is complicated due to presence of blood vessels and other noisy artifacts. Method: We present computer aided system for classification of retinal fundus images using a novel nature inspired spider monkey optimization for parameter tuning of gradient boosting machines classifier. The image enhancement has been performed with histogram equalization and contourlet transform. The pixels belonging to optic disc region are detected and eliminated using circular Hough transform and Otsu's segmentation method. We have employed Kirsch's matrices for blood vessel detection. The GLCM based feature vector extraction has been employed for textural features. The classification has been performed with hybrid SMO-GBM classifier. Result: We have utilized the STARE database for validation of proposed technique. The proposed system can effectively classify entire image set from test data. The SMO-GBM classifier can further sub-segregate into sub classes with an average accuracy of 97.5%. Conclusion: The proposed approach provides detection and grading of diabetic retinopathy. The abnormality is further categories as soft, moderate and severe. The hybrid SMO-GBM classifier yields a better statistical metrics than the existing exudates classification approaches.     

A Comprehensive Texture Segmentation Framework for Segmentation of Capillary Non-Perfusion Regions in Fundus Fluorescein Angiograms

Capillary non-perfusion (CNP) in the retina is a characteristic feature used in the management of a wide range of retinal diseases. There is no well-established computation tool for assessing the extent of CNP. We propose a novel texture segmentation framework to address this problem. This framework comprises three major steps: pre-processing, unsupervised total variation texture segmentation, and supervised segmentation. It employs a state-of-the-art multiphase total variation texture segmentation model which is enhanced by new kernel based region terms. The model can be applied to texture and intensity-based multiphase problems. A supervised segmentation step allows the framework to take expert knowledge into account, an AdaBoost classifier with weighted cost coefficient is chosen to tackle imbalanced data classification problems. To demonstrate its effectiveness, we applied this framework to 48 images from malarial retinopathy and 10 images from ischemic diabetic maculopathy. The performance of segmentation is satisfactory when compared to a reference standard of manual delineations: accuracy, sensitivity and specificity are 89.0%, 73.0%, and 90.8% respectively for the malarial retinopathy dataset and 80.8%, 70.6%, and 82.1% respectively for the diabetic maculopathy dataset. In terms of region-wise analysis, this method achieved an accuracy of 76.3% (45 out of 59 regions) for the malarial retinopathy dataset and 73.9% (17 out of 26 regions) for the diabetic maculopathy dataset. This comprehensive segmentation framework can quantify capillary non-perfusion in retinopathy from two distinct etiologies, and has the potential to be adopted for wider applications.     

Hemodynamic Segmentation of Brain Perfusion Images with Delay and Dispersion Effects Using an Expectation-Maximization Algorithm

Automatic identification of various perfusion compartments from dynamic susceptibility contrast magnetic resonance brain images can assist in clinical diagnosis and treatment of cerebrovascular diseases. The principle of segmentation methods was based on the clustering of bolus transit-time profiles to discern areas of different tissues. However, the cerebrovascular diseases may result in a delayed and dispersed local perfusion and therefore alter the hemodynamic signal profiles. Assessing the accuracy of the segmentation technique under delayed/dispersed circumstance is critical to accurately evaluate the severity of the vascular disease. In this study, we improved the segmentation method of expectation-maximization algorithm by using the results of hierarchical clustering on whitened perfusion data as initial parameters for a mixture of multivariate Gaussians model. In addition, Monte Carlo simulations were conducted to evaluate the performance of proposed method under different levels of delay, dispersion, and noise of signal profiles in tissue segmentation. The proposed method was used to classify brain tissue types using perfusion data from five normal participants, a patient with unilateral stenosis of the internal carotid artery, and a patient with moyamoya disease. Our results showed that the normal, delayed or dispersed hemodynamics can be well differentiated for patients, and therefore the local arterial input function for impaired tissues can be recognized to minimize the error when estimating the cerebral blood flow. Furthermore, the tissue in the risk of infarct and the tissue with or without the complementary blood supply from the communicating arteries can be identified.     

A Voting-Based Sequential Pattern Recognition Method

We propose a novel method for recognizing sequential patterns such as motion trajectory of biological objects (i.e., cells, organelle, protein molecules, etc.), human behavior motion, and meteorological data. In the proposed method, a local classifier is prepared for every point (or timing or frame) and then the whole pattern is recognized by majority voting of the recognition results of the local classifiers. The voting strategy has a strong benefit that even if an input pattern has a very large deviation from a prototype locally at several points, they do not severely influence the recognition result; they are treated just as several incorrect votes and thus will be neglected successfully through the majority voting. For regularizing the recognition result, we introduce partial-dependency to local classifiers. An important point is that this dependency is introduced to not only local classifiers at neighboring point pairs but also to those at distant point pairs. Although, the dependency makes the problem non-Markovian (i.e., higher-order Markovian), it can still be solved efficiently by using a graph cut algorithm with polynomial-order computations. The experimental results revealed that the proposed method can achieve better recognition accuracy while utilizing the above characteristics of the proposed method.     

Motion segmentation method for hybrid characteristic on human motion

Motion segmentation and analysis are used to improve the process of classification of motion and information gathered on repetitive or periodic characteristic. The classification result is useful for ergonomic and postural safety analysis, since repetitive motion is known to be related to certain musculoskeletal disorders. Past studies mainly focused on motion segmentation on particular motion characteristic with certain prior knowledge on static or periodic property of motion, which narrowed method's applicability. This paper attempts to introduce a method to tackle human joint motion without having prior knowledge. The motion is segmented by a two-pass algorithm. Recursive least square (RLS) is firstly used to estimate possible segments on the input human-motion set. Further, period identification and extra segmentation process are applied to produce meaningful segments. Each of the result segments is modeled by a damped harmonic model, with frequency, amplitude and duration produced as parameters for ergonomic evaluation and other human factor studies such as task safety evaluation and sport analysis. Experiments show that the method can handle periodic, random and mixed characteristics on human motion, which can also be extended to the usage in repetitive motion in workflow and irregular periodic motion like sport movement.     

A watershed approach for improving medical image segmentation

In this paper, a novel watershed approach based on seed region growing and image entropy is presented which could improve the medical image segmentation. The proposed algorithm enables the prior information of seed region growing and image entropy in its calculation. The algorithm starts by partitioning the image into several levels of intensity using watershed multi-degree immersion process. The levels of intensity are the input to a computationally efficient seed region segmentation process which produces the initial partitioning of the image regions. These regions are fed to entropy procedure to carry out a suitable merging which produces the final segmentation. The latter process uses a region-based similarity representation of the image regions to decide whether regions can be merged. The region is isolated from the level and the residual pixels are uploaded to the next level and so on, we recall this process as multi-level process and the watershed is called multi-level watershed. The proposed algorithm is applied to challenging applications: grey matter–white matter segmentation in magnetic resonance images (MRIs). The established methods and the proposed approach are experimented by these applications to a variety of simulating immersion, multi-degree, multi-level seed region growing and multi-level seed region growing with entropy. It is shown that the proposed method achieves more accurate results for medical image oversegmentation.     

An automatic segmentation and classification framework for anti-nuclear antibody images

Autoimmune disease is a disorder of immune system due to the over-reaction of lymphocytes against one's own body tissues. Anti-Nuclear Antibody (ANA) is an autoantibody produced by the immune system directed against the self body tissues or cells, which plays an important role in the diagnosis of autoimmune diseases. Indirect ImmunoFluorescence (IIF) method with HEp-2 cells provides the major screening method to detect ANA for the diagnosis of autoimmune diseases. Fluorescence patterns at present are usually examined laboriously by experienced physicians through manually inspecting the slides with the help of a microscope, which usually suffers from inter-observer variability that limits its reproducibility. Previous researches only provided simple segmentation methods and criterions for cell segmentation and recognition, but a fully automatic framework for the segmentation and recognition of HEp-2 cells had never been reported before. This study proposes a method based on the watershed algorithm to automatically detect the HEp-2 cells with different patterns. The experimental results show that the segmentation performance of the proposed method is satisfactory when evaluated with percent volume overlap (PVO: 89%). The classification performance using a SVM classifier designed based on the features calculated from the segmented cells achieves an average accuracy of 96.90%, which outperforms other methods presented in previous studies. The proposed method can be used to develop a computer-aided system to assist the physicians in the diagnosis of auto-immune diseases.     

基于线性判别分析的基因表达数据分类方法研究

由于基因表达数据高属性维、低样本维的特点,Fisher分类器对该种数据分类性能不是很高。本文提出了Fisher的改进算法Fisher-List。该算法独特之处在于为每个类别确定一个决策阀值,每个阀值既包含总体样本信息,又含有某些对分类至关重要的个体样本信息。本文用实验证明新算法在基因表达数据分类方面比Fisher、LogitBoost、AdaBoost、k-近邻法、决策树和支持向量机具有更高的性能。     

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.     

Stochastic segmentation models for array-based comparative genomic hybridization data analysis

Array-based comparative genomic hybridization (array-CGH) is a high throughput, high resolution technique for studying the genetics of cancer. Analysis of array-CGH data typically involves estimation of the underlying chromosome copy numbers from the log fluorescence ratios and segmenting the chromosome into regions with the same copy number at each location. We propose for the analysis of array-CGH data, a new stochastic segmentation model and an associated estimation procedure that has attractive statistical and computational properties. An important benefit of this Bayesian segmentation model is that it yields explicit formulas for posterior means, which can be used to estimate the signal directly without performing segmentation. Other quantities relating to the posterior distribution that are useful for providing confidence assessments of any given segmentation can also be estimated by using our method. We propose an approximation method whose computation time is linear in sequence length which makes our method practically applicable to the new higher density arrays. Simulation studies and applications to real array-CGH data illustrate the advantages of the proposed approach.     

Computational neural networks: Enhancing supervised learning algorithms via self-organization

A neural network processing scheme is proposed which utilizes a self-organizing Kohonen feature map as the front end to a feedforward classifier network. The results of a series of benchmarking studies based upon artificial statistical pattern recognition tasks indicate that the proposed architecture performs significantly better than conventional feedforward classifier networks when the decision regions are disjoint. This is attributed to the fact that the self-organization process allows internal units in the succeeding classifier network to be sensitive to a specific set of features in the input space at the outset of training.     

Improving ECG classification accuracy using an ensemble of neural network modules

This paper illustrates the use of a combined neural network model based on Stacked Generalization method for classification of electrocardiogram (ECG) beats. In conventional Stacked Generalization method, the combiner learns to map the base classifiers' outputs to the target data. We claim adding the input pattern to the base classifiers' outputs helps the combiner to obtain knowledge about the input space and as the result, performs better on the same task. Experimental results support our claim that the additional knowledge according to the input space, improves the performance of the proposed method which is called Modified Stacked Generalization. In particular, for classification of 14966 ECG beats that were not previously seen during training phase, the Modified Stacked Generalization method reduced the error rate for 12.41% in comparison with the best of ten popular classifier fusion methods including Max, Min, Average, Product, Majority Voting, Borda Count, Decision Templates, Weighted Averaging based on Particle Swarm Optimization and Stacked Generalization.     

FISH Finder: a high-throughput tool for analyzing FISH images

MOTIVATION: Fluorescence in situ hybridization (FISH) is used to study the organization and the positioning of specific DNA sequences within the cell nucleus. Analyzing the data from FISH images is a tedious process that invokes an element of subjectivity. Automated FISH image analysis offers savings in time as well as gaining the benefit of objective data analysis. While several FISH image analysis software tools have been developed, they often use a threshold-based segmentation algorithm for nucleus segmentation. As fluorescence signal intensities can vary significantly from experiment to experiment, from cell to cell, and within a cell, threshold-based segmentation is inflexible and often insufficient for automatic image analysis, leading to additional manual segmentation and potential subjective bias. To overcome these problems, we developed a graphical software tool called FISH Finder to automatically analyze FISH images that vary significantly. By posing the nucleus segmentation as a classification problem, compound Bayesian classifier is employed so that contextual information is utilized, resulting in reliable classification and boundary extraction. This makes it possible to analyze FISH images efficiently and objectively without adjustment of input parameters. Additionally, FISH Finder was designed to analyze the distances between differentially stained FISH probes. AVAILABILITY: FISH Finder is a standalone MATLAB application and platform independent software. The program is freely available from: http://code.google.com/p/fishfinder/downloads/list.     

基于面向对象的QuickBird遥感影像林隙分割与分类

传统的实地调查和人工解译方法已经不能满足区域尺度的林隙获取,高空间分辨率遥感影像的出现为区域尺度的林隙获取提供了可能.本研究采用QuickBird高空间分辨率光学遥感影像,结合面向对象分类技术对福建省三明市将乐县将乐国有林场进行林隙分割与分类.在面向对象分类过程中,采用10种尺度(10~100,步长为10)对QuickBird遥感影像进行分割,应用参考对象相交面积(RA_or)和分割对象相交面积(RA_os)进行分割结果评价.对每个尺度分割结果应用16个光谱特征,采用向量机分类器(SVM)进行林隙、非林隙和其他类型分类.结果表明:通过RA_or和RA_os等值法获得最优分割尺度参数为40.不同尺度参数之间的分类总精度最高相差22%.在最优尺度下,应用SVM分类器对林隙、非林隙和其他类型分类的总精度高达88%(Kappa=0.82).采用高空间分辨率遥感数据并结合面向对象的方法,可以代替传统的实地调查和人工解译对区域尺度的林隙进行识别分类.     

Super-resolution segmentation of imaging mass spectrometry data: Solving the issue of low lateral resolution

In the last decade, imaging mass spectrometry has seen incredible technological advances in its applications to biological samples. One computational method of data mining in this field is the spatial segmentation of a sample, which produces a segmentation map highlighting chemically similar regions. An important issue for any imaging mass spectrometry technology is its relatively low spatial or lateral resolution (i.e. a large size of pixel) as compared with microscopy. Thus, the spatial resolution of a segmentation map is also relatively low, that complicates its visual examination and interpretation when compared with microscopy data, as well as reduces the accuracy of any automated comparison. We address this issue by proposing an approach to improve the spatial resolution of a segmentation map. Given a segmentation map, our method magnifies it up to some factor, producing a super-resolution segmentation map. The super-resolution map can be overlaid and compared with a high-res microscopy image. The proposed method is based on recent advances in image processing and smoothes the "pixilated" region boundaries while preserving fine details. Moreover, it neither eliminates nor splits any region. We evaluated the proposed super-resolution segmentation approach on three MALDI-imaging datasets of human tissue sections and demonstrated the superiority of the super-segmentation maps over standard segmentation maps.     

Quantitative classification of life-style types in predaceous phytoseiid mites

Classification of species into different functional groups based on biological criteria has been a difficult problem in ecology. The difficulty mainly arises because natural classification patterns are not necessarily mutually exclusive. The more group characteristics overlap, the more difficult it is to identify the membership of a species in the overlapping portions of any two groups. In this paper, we present an application of discriminant analysis by creating classification models from life history and morphological data for two specialist and two generalist life-styles type of predaceous phytoseiid mites. Two stages can be distinguished in our method: life-style group membership assignment and trait variable evaluation. We use a Bayesian framework to create a classifier system to locate or assign species within a mixture of trait distributions. The method assumes that a mixture of trait distributions can represent the multiple dimensions of biological data. The mixture is most evident near the boundaries between groups. Because of the complexity of analytical solution, an iterative method is used to estimate the unknown means, variances, and mixing proportion between groups. We also developed a criterion based on information theory to evaluate model performance with different combinations of input variables and different hypotheses. We present a working example of our proposed methods. We apply these methods to the problem of selecting key species for inoculative release and for classical introductions of biological pest control agents.     

Semi-automatic 3D segmentation of carotid lumen in contrast-enhanced computed tomography angiography images

The atherosclerosis disease is one of the major causes of the death in the world. Atherosclerosis refers to the hardening and narrowing of the arteries by plaques. Carotid stenosis is a narrowing or constriction of carotid artery lumen usually caused by atherosclerosis. Carotid artery stenosis can increase risk of brain stroke. Contrast-enhanced Computed Tomography Angiography (CTA) is a minimally invasive method for imaging and quantification of the carotid plaques. Manual segmentation of carotid lumen in CTA images is a tedious and time consuming procedure which is subjected to observer variability. As a result, there is a strong and growing demand for developing computer-aided carotid segmentation procedures. In this study, a novel method is presented for carotid artery lumen segmentation in CTA data. First, the mean shift smoothing is used for uniformity enhancement of gray levels. Then with the help of three seed points, the centerlines of the arteries are extracted by a 3D Hessian based fast marching shortest path algorithm. Finally, a 3D Level set function is performed for segmentation. Results on 14 CTA volumes data show 85% of Dice similarity and 0.42 mm of mean absolute surface distance measures. Evaluation shows that the proposed method requires minimal user intervention, low dependence to gray levels changes in artery path, resistance to extreme changes in carotid diameter and carotid branch locations. The proposed method has high accuracy and can be used in qualitative and quantitative evaluation.     

基于近邻成分分析算法的原发性肝癌精确放疗后HBV再激活分类预测

原发性肝癌(PLC)患者在精确放疗后乙型肝炎病毒(HBV)再激活是一种常见并发症,及时的预测防护能降低发病率、死亡率。研究表明:多余的特征变量会影响HBV再激活的预测精度。通过提出基于近邻成分分析(NCA)的特征选择方法找出HBV再激活的危险因素及特征组合。之后分别建立经Bayes优化前后的支持向量机模型(SVM)对这些关键特征子集及初始特征集进行分类预测。实验结果表:明HBV DNA水平、KPS评分、分割方式、外放边界、V25、肿瘤分期TNM、ChildPugh等都是影响HBV再激活的危险因素。其中经NCA特征选择之后发现的V25是在乙型肝炎病毒再激活研究中首次提出的危险因素。10折交叉验证下特征组合HBV DNA水平、外放边界、V25的预测精度高达86.11%。支持向量机分类器可以很好的应用于乙型肝炎病毒再激活的研究,特征选择后的关键特征组合具有更优越的分类性能。     

Constructing a fuzzy rule-based system using the ILFN network and Genetic Algorithm.

In this paper, a method for automatic construction of a fuzzy rule-based system from numerical data using the Incremental Learning Fuzzy Neural (ILFN) network and the Genetic Algorithm is presented. The ILFN network was developed for pattern classification applications. The ILFN network, which employed fuzzy sets and neural network theory, equips with a fast, one-pass, on-line, and incremental learning algorithm. After trained, the ILFN network stored numerical knowledge in hidden units, which can then be directly interpreted into if then rule bases. However, the rules extracted from the ILFN network are not in an optimized fuzzy linguistic form. In this paper, a knowledge base for fuzzy expert system is extracted from the hidden units of the ILFN classifier. A genetic algorithm is then invoked, in an iterative manner, to reduce number of rules and select only discriminate features from input patterns needed to provide a fuzzy rule-based system. Three computer simulations using a simulated 2-D 3-class data, the well-known Fisher's Iris data set, and the Wisconsin breast cancer data set were performed. The fuzzy rule-based system derived from the proposed method achieved 100% and 97.33% correct classification on the 75 patterns for training set and 75 patterns for test set, respectively. For the Wisconsin breast cancer data set, using 400 patterns for training and 299 patterns for testing, the derived fuzzy rule-based system achieved 99.5% and 98.33% correct classification on the training set and the test set, respectively.