Towards an autonomous human chromosome classification system using Competitive Support Vector Machines Teams (CSVMT)

In broad terms, karyotyping is the process of examination and classification of human chromosome images to diagnose genetic diseases and disorders. It requires time consuming manual examination of cell images by a cytogeneticist to distinguish chromosome classes from each other. Thus, a reliable autonomous human chromosome classification system not only saves time and money but also reduces errors due to the inadequate knowledge level of the expert. Human cell contains 23 pairs of chromosome, 22 autosomes and a pair of sex chromosomes. Hence, we face a multi-class classification task which represents a challenging case for any sort of classifier. In this work, to solve this classification problem, we propose a novel methodology consisting two stages: (i) data preparation and training, and (ii) testing. To determine the most informative content of the dataset several preliminary experiments are conducted and a Principal Component Analysis is done. Then, a single Support Vector Machine (SVM_ij) is trained to separate a pair of classes, (i,j) where a numerical optimization method Pattern Search (PS), is employed to find the optimal parameters for the SVM_ij. Considering 22 pairs of autosomes, 22 × 22 experts are trained and optimized. The cluster of experts, we obtain is named as Competitive SVM Teams (CSVMTs) where each SVM_ij competes with the others to label a new classification instance. The final output of the classifier is determined by majority voteing. The results obtained on Copenhagen dataset proves the merit of the algorithm as correct classification rates (CRR) on train and test samples are 99.55% and 97.84% respectively, which are higher than any accuracy rate achieved so far in the related literature.     

A review of thresholding strategies applied to human chromosome segmentation

Karyotype analysis is a widespread procedure in cytogenetics to assess the presence of genetic defects by the visualization of the structure of chromosomes. The procedure is lengthy and repetitive and an effective automatic analysis would greatly help the cytogeneticist routine work. Still, automatic segmentation and the full disentangling of chromosomes are open issues. The first step in every automatic procedure is the thresholding step, which detect blobs that represent either single chromosomes or clusters of chromosomes. The better the thresholding step, the easier is the subsequent disentanglement of chromosome clusters into single entities. We implemented eleven thresholding methods, i.e. the ones that appear in the literature as the best performers, and compared their performance in segmenting chromosomes and chromosome clusters in cytogenetic Q-band images. The images are affected by the presence of hyper- or hypo-fluorescent regions and by a contrast variability between the stained chromosomes and the background. A thorough analysis of the results highlights that, although every single algorithm shows peculiar strong/weak points, Adaptive Threshold and Region Based Level Set have the overall best performance. In order to provide the scientific community with a public dataset, the data and manual segmentation used in this paper are available for public download at http://bioimlab.dei.unipd.it.     

A modular framework for the automatic classification of chromosomes in Q-band images

The manual analysis of the karyogram is a complex and time-consuming operation, as it requires meticulous attention to details and well-trained personnel. Routine Q-band laboratory images show chromosomes that are randomly rotated, blurred or corrupted by overlapping and dye stains. We address here the problem of robust automatic classification, which is still an open issue. The proposed method starts with an improved estimation of the chromosome medial axis, along which an established set of features is then extracted. The following novel polarization stage estimates the chromosome orientation and makes this feature set independent on the reading direction along the axis. Feature rescaling and normalizing techniques take full advantage of the results of the polarization step, reducing the intra-class and increasing the inter-class variances. After a standard neural network based classification, a novel class reassignment algorithm is employed to maximize the probability of correct classification, by exploiting the constrained composition of the human karyotype. An average 94% of correct classification was achieved by the proposed method on 5474 chromosomes, whose images were acquired during laboratory routine and comprise karyotypes belonging to slightly different prometaphase stages. In order to provide the scientific community with a public dataset, all the data we used are publicly available for download.     

Designing fuzzy inference systems from data: Aninterpretability-oriented review

Fuzzy inference systems (FIS) are widely used for process simulation or control. They can be designed either from expert knowledge or from data. For complex systems, FIS based on expert knowledge only may suffer from a loss of accuracy. This is the main incentive for using fuzzy rules inferred from data. Designing a FIS from data can be decomposed into two main phases: automatic rule generation and system optimization. Rule generation leads to a basic system with a given space partitioning and the corresponding set of rules. System optimization can be done at various levels. Variable selection can be an overall selection or it can be managed rule by rule. Rule base optimization aims to select the most useful rules and to optimize rule conclusions. Space partitioning can be improved by adding or removing fuzzy sets and by tuning membership function parameters. Structure optimization is of a major importance: selecting variables, reducing the rule base and optimizing the number of fuzzy sets. Over the years, many methods have become available for designing FIS from data. Their efficiency is usually characterized by a numerical performance index. However, for human-computer cooperation another criterion is needed: the rule interpretability. An implicit assumption states that fuzzy rules are by nature easy to be interpreted. This could be wrong when dealing with complex multivariable systems or when the generated partitioning is meaningless for experts. The paper analyzes the main methods for automatic rule generation and structure optimization. They are grouped into several families and compared according to the rule interpretability criterion. For this purpose, three conditions for a set of rules to be interpretable are defined     

A fuzzy inference and categorization approach for supplier selection using compensatory and non-compensatory decision rules

Fuzzy set theory has been used as an approach to deal with uncertainty in the supplier selection decision process. However, most studies limit applications of fuzzy set theory to outranking potential suppliers, not including a qualification stage in the decision process, in which non-compensatory types of decision rules can be used to reduce the set of potential suppliers. This paper presents a supplier selection decision method based on fuzzy inference that integrates both types of approaches: a non-compensatory rule for sorting in qualification stages and a compensatory rule for ranking in the final selection. Fuzzy inference rules model human reasoning and are embedded in the system, which is an advantage when compared to approaches that combine fuzzy set theory with multicriteria decision making methods. Fuzzy inference combined with a fuzzy rule-based classification method is used to categorize suppliers in qualification stages. Classes of supplier performance can be represented by linguistic terms, which allow decision makers to deal with subjectivity and to express qualification requirements in linguistic formats. Implementation of the proposed method and techniques were analyzed and discussed using an illustrative case. Three defuzzification operators were used in the final selection, yielding the same ranking. Factorial design was applied to test consistency and sensitivity of the inference rules. The findings reinforce the argument that including stages of qualification based on fuzzy inference and categorization makes this method especially useful for selecting from a large set of potential suppliers and also for first time purchase.     

A rule-based computer scheme for centromere identification and polarity assignment of metaphase chromosomes

Automatic centromere identification and polarity assignment are two key factors in the automatic karyotyping of human chromosomes. A multi-stage rule-based computer scheme has been investigated to automatically detect centomeres and determine polarities for both abnormal and normal metaphase chromosomes. The scheme first implements a modified thinning algorithm to identify the medial axis of a chromosome and extracts three feature profiles. Based on a set of pre-optimized classification rules, the scheme adaptively identifies the centromere and then assigns corresponding polarity. An image dataset of 2287 chromosomes acquired from 24 abnormal and 26 normal Giemsa metaphase cells is utilized to optimize and test the scheme. The overall accuracy is 91.4% for centromere identification and 97.4% for polarity assignment. The experimental results demonstrate that our scheme can be successfully applied to diverse chromosomes, which include those severely bent and abnormal chromosomes extracted from cancer cells.     

Low level image segmentation: an expert system

A major problem in robotic vision is the segmentation of images of natural scenes in order to understand their content. This paper presents a new solution to the image segmentation problem that is based on the design of a rule-based expert system. General knowledge about low level properties of processes employ the rules to segment the image into uniform regions and connected lines. In addition to the knowledge rules, a set of control rules are also employed. These include metarules that embody inferences about the order in which the knowledge rules are matched. They also incorporate focus of attention rules that determine the path of processing within the image. Furthermore, an additional set of higher level rules dynamically alters the processing strategy. This paper discusses the structure and content of the knowledge and control rules for image segmentation.     

Joint segmentation and pairing of multispectral chromosome images

Automated segmentation of touching or overlapping chromosomes in a metaphase image is a critical step for computer-aided chromosomes analysis. Conventional chromosome imaging methods acquire single-band grayscale images, and such a limitation makes the separation of touching or overlapping chromosomes challenging. In the multiplex fluorescence in situ hybridization (M-FISH) technique, each class of chromosomes can bind with a different combination of fluorophores. The M-FISH technique results in multispectral chromosome images, which has distinct spectral signatures. This paper presents a novel automated chromosome analysis method to combine the pixel-level geometric and multispectral information with decision-level pairing information. Our chromosome segmentation method uses the geometric and spectral information to partition the chromosome cluster into three regions. There will be ambiguity when combining these regions into separated chromosomes by using only spectral and geometric information. Then a graph–theoretical pairing method is introduced to resolve any remaining ambiguity of the aforementioned segmentation process. Experimental results demonstrate that the proposed joint segmentation and pairing method outperforms conventional grayscale and multispectral segmentation methods in separating touching and overlapping chromosomes.     

The expert system for assessing customer satisfaction on fragrance notes: Using artificial neural networks

The objective of this research is to assess customer satisfaction on fragrance notes with the aid of the expert system program developed by the authors by using artificial neural networks. The expert system’s role is in the preparation to capture the knowledge of the experts and the data from the customer’s requirements. The system has the capability to compile the collected data and form the appropriate rules for choosing fragrance notes for the products. In order to identify the hidden pattern of the customer’s needs, the artificial neural networks technique has been applied to classify the fragrance notes based upon a list of selected information. Moreover, the expert system program has the capability to make decisions in ranking the scores of the fragrance notes presented in the selection. In addition, to validate the approach, the expert system program has been tested with a variety of customer types and the results indicate that an average of 59.42% is correctly predicted from customer group selection process. The recommendations and limitations are also presented.     

Fuzzy inference system for follicle detection in ultrasound images of ovaries

The ovarian ultrasound imaging is an effective tool in infertility treatment. Monitoring the follicles is especially important in human reproduction. Periodic measurements of the size and shape of follicles over several days are the primary means of evaluation by physicians. Today monitoring the follicles is done by non-automatic means with human interaction. This work can be very demanding and inaccurate and, in most of the cases, means only an additional burden for medical experts. To improve the performance of follicle detection in ultrasound images of ovaries, we develop a new algorithm using fuzzy logic. The proposed method employs contourlet transform for despeckling the ultrasound images of ovaries, active contours without edge method for segmentation and fuzzy logic for classification. The follicles in an ovary are characterized by seven geometric features which are used as inputs to the fuzzy logic block of the Fuzzy Inference System. The output of the fuzzy logic block is a follicle class or non follicle class. The fuzzy-knowledge-base consists of a set of physically interpretable if-then rules providing physical insight into the process. The experimentation has been done using sample ultrasound images of ovaries and the results are compared with the inferences drawn by interval based classifier and also those drawn by the medical expert. The experimental results demonstrate the efficacy of the proposed method.     

A novel evolutionary approach for optimizing content-based image indexing algorithms.

Optimization of content-based image indexing and retrieval (CBIR) algorithms is a complicated and time-consuming task since each time a parameter of the indexing algorithm is changed, all images in the database should be indexed again. In this paper, a novel evolutionary method called evolutionary group algorithm (EGA) is proposed for complicated time-consuming optimization problems such as finding optimal parameters of content-based image indexing algorithms. In the new evolutionary algorithm, the image database is partitioned into several smaller subsets, and each subset is used by an updating process as training patterns for each chromosome during evolution. This is in contrast to genetic algorithms that use the whole database as training patterns for evolution. Additionally, for each chromosome, a parameter called age is defined that implies the progress of the updating process. Similarly, the genes of the proposed chromosomes are divided into two categories: evolutionary genes that participate to evolution and history genes that save previous states of the updating process. Furthermore, a new fitness function is defined which evaluates the fitness of the chromosomes of the current population with different ages in each generation. We used EGA to optimize the quantization thresholds of the wavelet-correlogram algorithm for CBIR. The optimal quantization thresholds computed by EGA improved significantly all the evaluation measures including average precision, average weighted precision, average recall, and average rank for the wavelet-correlogram method.     

Data resource selection in distributed visual information systems

With the advances in multimedia databases and the popularization of the Internet, it is now possible to access large image and video repositories distributed throughout the world. One of the challenging problems in such access is how the information in the respective databases can be summarized to enable an intelligent selection of relevant database sites based on visual queries. This paper presents an approach to solve this problem based on image content-based indexing of a metadatabase at a query distribution server. The metadatabase records a summary of the visual content of the images in each database through image templates and statistical features characterizing the similarity distributions of the images. The selection of the databases is done by searching the metadatabase using a ranking algorithm that uses the query's similarity to a template and the features of the databases associated with the template. Two selection approaches, termed mean-based and histogram-based approaches, are presented. The database selection mechanisms have been implemented in a metaserver, and extensive experiments have been performed to demonstrate the effectiveness of the database selection approaches     

Feature Selection vs Theory Reformulation: A Study of Genetic Refinement of Knowledge-based Neural Networks

Expert classification systems have proven themselves effective decision makers for many types of problems. However, the accuracy of such systems is often highly dependent upon the accuracy of a human expert's domain theory. When human experts learn or create a set of rules, they are subject to a number of hindrances. Most significantly experts are, to a greater or lesser extent, restricted by the tradition of scholarship which has preceded them and by an inability to examine large amounts of data in a rigorous fashion without the effects of boredom or frustration. As a result, human theories are often erroneous or incomplete. To escape this dependency, machine learning systems have been developed to automatically refine and correct an expert's domain theory. When theory revision systems are applied to expert theories, they often concentrate on the reformulation of the knowledge provided rather than on the reformulation or selection of input features. The general assumption seems to be that the expert has already selected the set of features that will be most useful for the given task. That set may, however, be suboptimal. This paper studies theory refinement and the relative benefits of applying feature selection versus more extensive theory reformulation.     

面向工业过程的图像生成及其应用研究综述

在面向工业过程的计算机视觉研究中, 智能感知模型能否实际应用取决于其对复杂工业环境的适应能力. 由于可利用的工业图像数据集存在分布不均、多样性不足和干扰严重等问题, 如何生成符合多工况分布的期望训练集是提高感知模型性能的关键. 为解决上述问题, 以城市固废焚烧(Municipal solid wastes incineration, MSWI)过程为背景, 综述目前面向工业过程的图像生成及其应用研究, 为进行面向工业图像的感知建模提供支撑. 首先, 梳理面向工业过程的图像生成定义和流程以及其应用需求; 随后, 分析在工业领域中具有潜在应用价值的图像生成算法; 接着, 从工业过程图像生成、生成图像评估和应用等视角进行现状综述; 然后, 对下一步研究方向进行讨论与分析; 最后, 对全文进行总结并指出未来挑战.     

用于提高谷歌图像搜索结果的二分类器在线学习方法

对于基于关键词的图像检索,利用检索结果的视觉相似性学习二分类器有望成为改善检索结果的最有效途径之一. 为改善搜索引擎的搜索结果,本文提出一种算法框架并且基于此框架着重研究训练数据选择这一关键问题. 训练数据选择过程由两个阶段组成：1）训练数据初始化以开始分类器学习过程;2）分类器迭代学习过程中的动态数据选择. 对于初始训练数据的选择,我们探讨了基于聚类和基于排序两种方法,并且对比了自动训练数据选择与人工标注的结果. 对于动态数据选择,我们比较了支持向量机和基于最大最小后验伪概率的贝叶斯分类器的分类效果. 组合上述两个阶段的不同方法,我们得到了8种不同的算法,并将其用于谷歌搜索引擎进行基于关键词的图像检索. 实验结果证明,如何从含有噪声的搜索结果中选择训练数据是搜索结果改善的关键问题. 实验显示我们的方法能够有效的改善谷歌搜索的结果,尤其是排序在前的结果. 尽早为用户提供更相关的结果能够更大程度的减少用户逐个翻页查看结果的工作. 另外,如何使自动训练数据选择与人工标注媲美仍是需要继续研究的一个问题.     

Visual change detection on tunnel linings

We describe an automated system for detecting, localising, clustering and ranking visual changes on tunnel surfaces. The system is designed to provide assistance to expert human inspectors carrying out structural health monitoring and maintenance on ageing tunnel networks. A three-dimensional tunnel surface model is first recovered from a set of reference images using Structure from Motion techniques. New images are localised accurately within the model and changes are detected versus the reference images and model geometry. We formulate the problem of detecting changes probabilistically and evaluate the use of different feature maps and a novel geometric prior to achieve invariance to noise and nuisance sources such as parallax and lighting changes. A clustering and ranking method is proposed which efficiently presents detected changes and further improves the inspection efficiency. System performance is assessed on a real data set collected using a low-cost prototype capture device and labelled with ground truth. Results demonstrate that our system is a step towards higher frequency visual inspection at a reduced cost.     

Reverse engineering expert visual observations: From fixations to the learning of spatial filters with a neural-gas algorithm

Human beings can become experts in performing specific vision tasks, for example, doctors analysing medical images, or botanists studying leaves. With sufficient knowledge and experience, people can become very efficient at such tasks. When attempting to perform these tasks with a machine vision system, it would be highly beneficial to be able to replicate the process which the expert undergoes. Advances in eye-tracking technology can provide data to allow us to discover the manner in which an expert studies an image. This paper presents a first step towards utilizing these data for computer vision purposes. A growing-neural-gas algorithm is used to learn a set of Gabor filters which give high responses to image regions which a human expert fixated on. These filters can then be used to identify regions in other images which are likely to be useful for a given vision task. The algorithm is evaluated by learning filters for locating specific areas of plant leaves.     

Opening the black box of neural networks for remote sensing image classification

Neural networks, which make no assumption about data distribution, have achieved improved image classification results compared to traditional methods. Unfortunately, a neural network is generally perceived as being a ‘black box’. It is extremely difficult to document how specific classification decisions are reached. Fuzzy systems, on the other hand, have the capability to represent classification decisions explicitly in the form of fuzzy ‘if-then’ rules. However, the construction of a knowledge base, especially the fine-tuning of the fuzzy set parameters of the fuzzy rules in a fuzzy expert system, is a tedious and subjective process. This research has developed a new, improved neuro-fuzzy image classification system based on the synergism between neural networks and fuzzy expert systems. It incorporates the best of both technologies and compensates for the shortcomings of each. The learning algorithms of neural networks developed here are used to automate the derivation of fuzzy set parameters for the fuzzy ‘if-then’ rules in a fuzzy expert system. The rules obtained, in symbolic form, facilitate the understanding of the neural network based image classification system. In addition, the image classification accuracy obtained from the improved neuro-fuzzy system was significantly superior to those of the back-propagation based neural network and the maximum likelihood approaches.