Automated TB classification using ensemble of deep architectures

Multimedia Tools and Applications - Tuberculosis (TB) is an infectious disease that mainly affects the lung region. Its initial screening is mostly performed using chest radiograph, which is also...     

Histogram Matched Chest X-Rays Based Tuberculosis Detection Using CNN

Tuberculosis (TB) is a severe infection that mostly affects the lungs and kills millions of people’s lives every year. Tuberculosis can be diagnosed using chest X-rays (CXR) and data-driven deep learning (DL) approaches. Because of its better automated feature extraction capability, convolutional neural networks (CNNs) trained on natural images are particularly effective in image categorization. A combination of 3001 normal and 3001 TB CXR images was gathered for this study from different accessible public datasets. Ten different deep CNNs (Resnet50, Resnet101, Resnet152, InceptionV3, VGG16, VGG19, DenseNet121, DenseNet169, DenseNet201, MobileNet) are trained and tested for identifying TB and normal cases. This study presents a deep CNN approach based on histogram matched CXR images that does not require object segmentation of interest, and this coupled methodology of histogram matching with the CXRs improves the accuracy and detection performance of CNN models for TB detection. Furthermore, this research contains two separate experiments that used CXR images with and without histogram matching to classify TB and non-TB CXRs using deep CNNs. It was able to accurately detect TB from CXR images using pre-processing, data augmentation, and deep CNN models. Without histogram matching the best accuracy, sensitivity, specificity, precision and F1-score in the detection of TB using CXR images among ten models are 99.25%, 99.48%, 99.52%, 99.48% and 99.22% respectively. With histogram matching the best accuracy, sensitivity, specificity, precision and F1-score are 99.58%, 99.82%, 99.67%, 99.65% and 99.56% respectively. The proposed methodology, which has cutting-edge performance, will be useful in computer-assisted TB diagnosis and aids in minimizing irregularities in TB detection in developing countries.     

Sub-microliter scale in-gel loop-mediated isothermal amplification (LAMP) for detection of Mycobacterium tuberculosis

Tuberculosis (TB), caused by Mycobacterium tuberculosis, is a common human disease that is prevalent in resource-deprived areas of the world. Current detection techniques for TB require expensive conventional instruments in a laboratory setting, preventing accessible and low cost diagnosis of the disease. Using a loop-mediated isothermal amplification (LAMP) assay, we have amplified and detected TB in a 6 × 8 semisolid polyacrylamide gel post array using an inexpensive prototype instrument. Each post contains 670 nL of volume, minimizing the need for large quantities of reagents. Amplified DNA is detected via fluorescence of the dye LCGreen Plus+, which is polymerized into the gel along with other reagents. The prototype device contains a Peltier element for heating, a diode laser as an excitation source, and a CCD camera for detecting fluorescence in real-time. About 12 Mycobacterium tuberculosis genomes per gel post can be detected within 75 min of amplification. This sensitivity is similar to that obtained by conventional methods using a commercial thermocycler. We achieved comparable LAMP amplification when the template is added externally or when the template is polymerized in the gel. This rapid isothermal amplification technology, with its simple thermal requirements, has the potential to be integrated into micro-devices and serves as a model for implementing future low-cost point of care diagnostics.     

Semantic interoperability of controlled vocabularies in medicine: A case study of the International Statistical Classification of Diseases ‘Tuberculosis’ subset

Controlled vocabularies have been promoted for the achievement of semantic interoperability in e-health. However, the current implementations of healthcare information systems struggle with the remaining encoded semantics inside each particular data model. Multilevel modelling has been proposed for overcoming the challenges of semantic interoperability in healthcare, with different approaches for binding domain models to standard vocabularies; however, proofs of concept are still needed. This paper presents the fundamentals of knowledge modelling with standard vocabularies using the Multilevel Healthcare Information Modelling (MLHIM) specifications. The implementation of one term subset (‘Tuberculosis’) of the 10th Revision of the International Classification of Diseases (ICD-10) in MLHIM domain models, the Concept Constraint Definitions (CCD), is described, using the Brazilian mortality and hospital information systems as use cases. Technical details of the semantic validation of data instances generated according to the XML instances that include this ICD-10 term set, as well as the correspondent migration of the original databases to MLHIM-compliant databases, are presented.     

数据挖掘在肺结核疾病智能决策中的应用研究

针对单一数据挖掘方法对肺结核疾病诊断效率低、准确性不高的问题,本研究对北京市昌平区结核病防治所,北京市结核病控制研究所的1203例肺结核病人档案资料构建了电子档案,采用粗糙集和决策树结合方法建立肺结核疾病诊断模型,并对其准确性进行评估,在此基础上构建肺结核疾病诊断系统;在研究中,使用粗糙集和决策树相结合的方法进行属性约简,约简掉冗余属性57个,剩余属性22个,得到决策规则7条,模型准确率为83.46%;与未未约简的方法相比,决策规则减少128%,模型准确率基本保持不变;研究结果表明:使用该组合算法,在保证模型准确率的同时,降低了算法的时间和空间复杂性,提高了挖掘效率,为临床诊断提供了一定的借鉴。     

Homology modelling,docking, pharmacophore and site directed mutagenesis analysis to identify the critical amino acid residue of PknI from Mycobacterium tuberculosis

Tuberculosis is caused by Mycobacterium tuberculosis, an intracellular pathogen. PknI is one of the 11 functional Serine/Threonine Protein Kinases which is predicted to regulate the cell division of M. tuberculosis. In order to find newer drugs and vaccine we need to understand the pathogenesis of the disease. We have used the bioinformatics approach to identify the functionally active residues of PknI and to confirm the same with wet lab experiments. In the current study, we have created homology model for PknI and have done comparative structural analysis of PknI with other kinases. Molecular docking studies were done with a library of kinase inhibitors and T95 was found as the potent inhibitor for PknI. Based on structure based pharmacophore analysis of kinase substrate complexes, Lys 41 along with Asp90, Val92 and Asp96 were identified as functionally important residues. Further, we used site directed mutagenesis technique to mutate Lys 41 to Met resulting in defective cell division of Mycobacterium smegmatis mc². Overall, the proposed model together with its binding features gained from pharmacophore docking studies helped in identifying ligand inhibitor specific to PknI which was confirmed by laboratory experiments.     

QM/MM investigation of the reaction rates of substrates of 2,3-dimethylmalate lyase: A catabolic protein isolated from Aspergillus niger

Aspergillus niger is an industrially important microorganism used in the production of citric acid. It is a common cause of food spoilage and represents a health issue for patients with compromised immune systems. Recent studies on Aspergillus niger have revealed details on the isocitrate lyase (ICL) superfamily and its role in catabolism, including (2R, 3S)-dimethylmalate lyase (DMML). Members of this and related lyase super families are of considerable interest as potential treatments for bacterial and fungal infections, including Tuberculosis. In our efforts to better understand this class of protein, we investigate the catalytic mechanism of DMML, studying five different substrates and two different active site metals configurations using molecular dynamics (MD) and hybrid quantum mechanics/molecular mechanics (QM/MM) calculations. We show that the predicted barriers to reaction for the substrates show good agreement with the experimental k_cat values. This results help to confirm the validity of the proposed mechanism and open up the possibility of developing novel mechanism based inhibitors specifically for this target.     

Computer-Aided Diagnosis for Tuberculosis Classification with Water Strider Optimization Algorithm

Computer-aided diagnosis (CAD) models exploit artificial intelligence (AI) for chest X-ray (CXR) examination to identify the presence of tuberculosis (TB) and can improve the feasibility and performance of CXR for TB screening and triage. At the same time, CXR interpretation is a time-consuming and subjective process. Furthermore, high resemblance among the radiological patterns of TB and other lung diseases can result in misdiagnosis. Therefore, computer-aided diagnosis (CAD) models using machine learning (ML) and deep learning (DL) can be designed for screening TB accurately. With this motivation, this article develops a Water Strider Optimization with Deep Transfer Learning Enabled Tuberculosis Classification (WSODTL-TBC) model on Chest X-rays (CXR). The presented WSODTL-TBC model aims to detect and classify TB on CXR images. Primarily, the WSODTL-TBC model undergoes image filtering techniques to discard the noise content and U-Net-based image segmentation. Besides, a pre-trained residual network with a two-dimensional convolutional neural network (2D-CNN) model is applied to extract feature vectors. In addition, the WSO algorithm with long short-term memory (LSTM) model was employed for identifying and classifying TB, where the WSO algorithm is applied as a hyperparameter optimizer of the LSTM methodology, showing the novelty of the work. The performance validation of the presented WSODTL-TBC model is carried out on the benchmark dataset, and the outcomes were investigated in many aspects. The experimental development pointed out the betterment of the WSODTL-TBC model over existing algorithms.     

Dual-side see-through integral imaging 3D display system based on lens array holographic optical element

We propose a dual-side see-through integral imaging 3D display based on a lens array holographic optical element (LAHOE). The display system consists of two projectors and a LAHOE. The LAHOE is fabricated according to the theory of reflective volume holograms. An interference pattern is formed by a plane-wave beam and a spherical-wave array beam, and the interference pattern is recorded into a photopolymer material. Thus, the LAHOE is formed. When a reference beam is projected into the LAHOE, the wavefront of a convex micro-lens array is reconstructed. When a phase conjugation reference beam is projected into the LAHOE, the wavefront of a concave micro-lens array is reconstructed. So, the LAHOE possesses the optical properties of a convex micro-lens array and a concave micro-lens array. The proposed display is developed using integral imaging principle and it realizes dual-side see-through integral imaging 3D display.     

Minimal states and maximum free energies of materials with memory

Definitions are given of the maximum and minimum free energy associated with a given state of a material with memory. Also, the concept of a minimal state is introduced. These concepts are then explored in detail for a specific isothermal model, where the stress is given by a non-linear elastic part and a memory part which is a linear functional of the strain tensor history. It is shown that the equivalence class constituting a minimal state is a singleton except where only isolated singularities occur in the Fourier transform of the relaxation tensor derivative. If the minimal state is not a singleton, then the maximum free energy is less than the work function and is a function of the minimal state. An explicit expression is given for the maximum free energy.     

Sensor Planning for 3D Object Search

In this paper, we provide a systematic study of the task of sensor planning for object search. The search agent's knowledge of object location is encoded as a discrete probability density which is updated whenever a sensing action occurs. Each sensing action of the agent is defined by a viewpoint, a viewing direction, a field-of-view, and the application of a recognition algorithm. The formulation casts sensor planning as an optimization problem: the goal is to maximize the probability of detecting the target with minimum cost. This problem is proved to be NP-Complete, thus a heuristic strategy is favored. To port the theoretical framework to a real working system, we propose a sensor planning strategy for a robot equipped with a camera that can pan, tilt, and zoom. In order to efficiently determine the sensing actions over time, the huge space of possible actions with fixed camera position is decomposed into a finite set of actions that must be considered. The next action is then selected from among these by comparing the likelihood of detection and the cost of each action. When detection is unlikely at the current position, the robot is moved to another position for which the probability of target detection is the highest.     

Fault tolerant control for joint structure in PEMS high speed maglev train

The PEMS high speed maglev train, which features a permanent magnet inside an electromagnet, is a new kind of maglev train for long distance intercity transportation. The joint structure, which consists of two single levitation sub‐systems, is the fundamental levitation unit. Two kinds of faults are considered and corresponding fault tolerant control strategies are proposed. The first fault condition is when a gap sensor that is part of a single levitation system is faulty. For this kind of fault, a fault tolerant control strategy based on signal reconfiguration is proposed. The second fault condition is when the whole of a single levitation sub‐system is faulty. Under this condition, a faulty model is firstly established, then a fault tolerant control strategy is designed. When this kind of fault is detected, a switch from the normal controller to the fault tolerant controller can make the faulty system stable.     

混合关系中的函数依赖及其相关规则

在实际应用中有这样一类关系数据库,其中数据项在某个属性上的取值本身又是一个关系。讨论这种混合关系中的函数依赖及其相应的相关规则具有一定的理论意义和实际应用价值。给出这种混合关系的形式定义、混合关系中4种类型的函数依赖关系以及相应的4种类型的相关规则;并给出混合关系到一般关系的转换,通过这个转换说明了混合关系与普通关系的联系及其差别。给出一个实际应用中的例子,来说明混合关系中的函数依赖和相关规则的形式多样性。     

K(Knowledge)-net: building up and its dynamics

The essence of intelligence is to use certain abilities to obtain knowledge, to use that knowledge, and to operate with that knowledge. New knowledge learned by a human is often related to old existing knowledge, and sometimes we could have more conceptual knowledge based on old knowledge. So, the knowledge in the brain exists in a related structural form, and this structure is dynamic, and therefore is evolvable. Based on the understanding of the real process of learning by a human being, we discuss how to make a model to describe the dynamic structure of knowledge. This model is also a principle of artificial brain design. Most of the knowledge a child learns is from natural language and perception information, and we define this as semantic knowledge. The model to describe the process and structure of knowledge growing in a network form is called a K-net. It is a dynamic network with two main dynamics: one is new knowledge added, and the other is aggregating knowledge existing in the network with some probability. Under these very natural conditions, we found that the network is originally a simple random net, and then some characteristics of a complex network gradually appear when more new knowledge is added and aggregated. A more interesting phenomenon is the appearance of a random hierarchical structure. Does this mean emergence?     

Generalized theory of uncertainty (GTU)—principal concepts and ideas

Uncertainty is an attribute of information. The path-breaking work of Shannon has led to a universal acceptance of the thesis that information is statistical in nature. Concomitantly, existing theories of uncertainty are based on probability theory. The generalized theory of uncertainty (GTU) departs from existing theories in essential ways. First, the thesis that information is statistical in nature is replaced by a much more general thesis that information is a generalized constraint, with statistical uncertainty being a special, albeit important case. Equating information to a generalized constraint is the fundamental thesis of GTU. Second, bivalence is abandoned throughout GTU, and the foundation of GTU is shifted from bivalent logic to fuzzy logic. As a consequence, in GTU everything is or is allowed to be a matter of degree or, equivalently, fuzzy. Concomitantly, all variables are, or are allowed to be granular, with a granule being a clump of values drawn together by a generalized constraint. And third, one of the principal objectives of GTU is achievement of NL-capability, that is, the capability to operate on information described in natural language. NL-capability has high importance because much of human knowledge, including knowledge about probabilities, is described in natural language. NL-capability is the focus of attention in the present paper. The centerpiece of GTU is the concept of a generalized constraint. The concept of a generalized constraint is motivated by the fact that most real-world constraints are elastic rather than rigid, and have a complex structure even when simple in appearance. The paper concludes with examples of computation with uncertain information described in natural language.     

Control design based on a linear state function observer

A control design method based on a linear state function observer is proposed. The method is a semi-inverse design procedure in that the control law is not designed before the observer system, but is a result of the observer design. However, the observer design is not completely independent of the control design, but seeks to yield a feedback signal that is close to a prescribed control law. First, the observer design problem is considered as the reconstruction of a linear function of the state vector. The linear state function to be reconstructed is the given control law. Then, based on the derivation for linear state function observers, the observer design is formulated as a parameter optimization problem. The optimization objective is to generate a matrix that is close to the given feedback gain matrix. Based on that matrix, the form of the observer and a new control law can be determined. The semi-inverse design procedure can yield a reduced-order observer with dimension considerably smaller than that of the system. Two numerical examples are used to demonstrate the proposed design procedure.     

A type system for logic programs

A theory for a type system for logic programs is developed which addressesthe question of well-typing, type inference, and compile-time and run-time type checking. A type is a recursively enumerable set of ground atoms, which is tuple-distributive. The association of a type to a program is intended to mean that only ground atoms that are elements of the type may be derived from the program. A declarative definition of well-typed programs is formulated, based on an intuitive approach related to the fixpoint semantics of logic programs. Whether a program is well typed is undecidable in general. We define a restricted class of types, called regular types, for which type checking is decidable. Regular unary logic programs are proposed as a specification language for regular types. An algorithm for type-checking a logic program with respect to a regular type definition is described, and its complexity is analyzed. Finally, the practicality of the type system is discussed, and some examples are shown. The type system has been implemented in FCP for FCP and is incorporated in the Logix system.     

基于通用比较模型的t-可诊断系统的特征化及并行诊断算法

文中介绍了一种系统级故障诊断模型－－通用比较模型,该模型处理器作为自身的比较器,综合经典的ＰＭＣ模型和Ｍａｅｎｇ／Ｍａｌｅｋ模型的优点,基于该报多处理器系统的ｔ＿可诊断性问题,给出了ｔ＿可诊断系统的特征化,证明了一个系统成为ｔ－可诊断系统的新的充分必要条件,其次,证明在通用比较模型中,确定故障顺集的问题等价于求解一个超图的最小横切集（Ｍｉｎｉｍｕｍ ｔｒａｖｅｒｓａｌ）,该超图是根据多处理器的通信