Narrow Results

Spatial and intensity normalizations are nowadays a prerequisite for neuroimaging analysis. Influenced by voxel-wise and other univariate comparisons, where these corrections are key, they are commonly applied to any type of analysis and imaging modalities. Nuclear imaging modalities such as PET-FDG or FP-CIT SPECT, a common modality used in Parkinson’s disease diagnosis, are especially dependent on intensity normalization. However, these steps are computationally expensive and furthermore, they may introduce deformations in the images, altering the information contained in them. Convolutional neural networks (CNNs), for their part, introduce position invariance to pattern recognition, and have been proven to classify objects regardless of their orientation, size, angle, etc. Therefore, a question arises: how well can CNNs account for spatial and intensity differences when analyzing nuclear brain imaging? Are spatial and intensity normalizations still needed? To answer this question, we have trained four different CNN models based on well-established architectures, using or not different spatial and intensity normalization preprocessings. The results show that a sufficiently complex model such as our three-dimensional version of the ALEXNET can effectively account for spatial differences, achieving a diagnosis accuracy of 94.1% with an area under the ROC curve of 0.984. The visualization of the differences via saliency maps shows that these models are correctly finding patterns that match those found in the literature, without the need of applying any complex spatial normalization procedure. However, the intensity normalization — and its type — is revealed as very influential in the results and accuracy of the trained model, and therefore must be well accounted.

In the neural network literature, many preprocessing techniques, such as feature de-correlation, input unbiasing and normalization, are suggested to accelerate multilayer perceptron training. In this paper, we show that a network trained with an original data set and one trained with a linear transformation of the original data will go through the same training dynamics, as long as they start from equivalent states. Thus preprocessing techniques may not be helpful and are merely equivalent to using a different weight set to initialize the network. Theoretical analyses of such preprocessing approaches are given for conjugate gradient, back propagation and the Newton method. In addition, an efficient Newton-like training algorithm is proposed for hidden layer training. Experiments on various data sets confirm the theoretical analyses and verify the improvement of the new algorithm.

Unsupervised image segmentation is a fundamental but challenging problem in computer vision. In this paper, we propose a novel unsupervised segmentation algorithm, which could find diverse applications in pattern recognition, particularly in computer vision. The algorithm, named Two-stage Fuzzy c-means Hybrid Approach (TFHA), adaptively clusters image pixels according to their multichannel Gabor responses taken at multiple scales and orientations. In the first stage, the fuzzy c-means (FCM) algorithm is applied for intelligent estimation of centroid number and initialization of cluster centroids, which endows the novel segmentation algorithm with adaptivity. To improve the efficiency of the algorithm, we utilize the Gray Level Co-occurrence Matrix (GLCM) feature extracted at the hyperpixel level instead of the pixel level to estimate centroid number and hyperpixel-cluster memberships, which are used as initialization parameters of the following main clustering stage to reduce the computational cost while keeping the segmentation performance in terms of accuracy close to original one. Then, in the second stage, the FCM algorithm is utilized again at the pixel level to improve the compactness of the clusters forming final homogeneous regions. To examine the performance of the proposed algorithm, extensive experiments were conducted and experimental results show that the proposed algorithm has a very effective segmentation results and computational behavior, decreases the execution time and increases the quality of segmentation results, compared with the state-of-the-art segmentation methods recently proposed in the literature.

This paper introduces an effective method for signature separation from nonhomogeneous noisy background. It also introduces a solution to the problem of simulated signature verification in off-line systems. Extraction of shape and density features and the effectiveness of using each and both of them are discussed in the light of experimental results.

In order to highlight the interesting problems and actual results on the state of the art in optical character recognition (OCR), this paper describes and compares preprocessing, feature extraction and postprocessing techniques for commercial reading machines.

Problems related to handwritten and printed character recognition are pointed out, and the functions and operations of the major components of an OCR system are described.

Historical background on the development of character recognition is briefly given and the working of an optical scanner is explained.

The specifications of several recognition systems that are commercially available are reported and compared.

In general, nonlinear shape normalization methods for binary images have been used in order to compensate for the shape distortions of handwritten characters. However, in most document image analysis and recognition systems, a gray-scale image is first captured and digitized using a scanner or a video camera, then a binary image is extracted from the original gray-scale image using a certain extraction technique. This binarization process may remove some useful information of character images such as topological features, and introduce noises to character background. These errors are accumulated in nonlinear shape normalization step and transferred to the following feature extraction or recognition step. They may eventually cause incorrect recognition results.

In this paper, we propose nonlinear shape normalization methods for gray-scale handwritten Oriental characters in order to minimize the loss of information caused by binarization and compensate for the shape distortions of characters. Two-dimensional linear interpolation technique has been extended to nonlinear space and the extended interpolation technique has been adopted in the proposed methods to enhance the quality of normalized images.

In order to verify the efficiency of the proposed methods, the recognition rate, the processing time and the computational complexity of the proposed algorithms have been considered. The experimental results demonstrate that the proposed methods are efficient not only to compensate for the shape distortions of handwritten Oriental characters but also to maintain the information in gray-scale Oriental characters.

This paper shows the added value of using the existing specific domain knowledge to generate new derivated variables to complement a target dataset and the benefits of including these new variables into further data analysis methods. The main contribution of the paper is to propose a methodology to generate these new variables as a part of preprocessing, under a double approach: creating 2nd generation knowledge-driven variables, catching the experts criteria used for reasoning on the field or 3rd generation data-driven indicators, these created by clustering original variables. And Data Mining and Artificial Intelligence techniques like Clustering or Traffic light Panels help to obtain successful results. Some results of the project INSESS-COVID19 are presented, basic descriptive analysis gives simple results that even though they are useful to support basic policy-making, especially in health, a much richer global perspective is acquired after including derivated variables. When 2nd generation variables are available and can be introduced in the method for creating 3rd generation data, added value is obtained from both basic analysis and building new data-driven indicators.

The $k$-plex is a relaxation of the classical clique structure, which is usually used to characterize the cohesiveness of a subgroup in social networks. In this paper, we propose a new notion called the $k$-plex influence to depict the outward connection of a $k$-plex. The influence of a $k$-plex is defined as the number of outside vertices adjacent to the vertices of the $k$-plex. With the new notion, we can distinguish different $k$-plexes and find more pivotal subgroups in networks. We propose an exact algorithm to find the $k$-plex with the maximum influence. The algorithm implements a branch-and-bound approach, in which we integrate a novel upper bound and an effective preprocessing strategy. We conducted experiments to evaluate the performance of the algorithm and compared it with the general mixed integer linear programming approach. The results show that both the preprocessing strategy and the upper bound can effectively reduce the search space, and the proposed branch-and-bound algorithm can solve the problem in massive graphs effectively and has a significant performance advantage over the general mixed integer linear programming approach.

This review compares different methods to identify differentially expressed genes in two-sample cDNA arrays. A two-sample experiment is a commonly used design to compare relative mRNA abundance between two different samples. This simple design is customarily used by biologists as a first screening before relying on more complex designs. Statistical techniques are quite well developed for such simple designs. For the identification of differentially expressed genes, four methods were described and compared: a fold test, a t-test (Long et al., 2001), SAM (Tusher et al., 2001) and an ANOVA-based bootstrap method (Kerr and Churchill, 2001). Mutual comparison of these methods clearly illustrates each method's advantages and pitfalls. Our analyses showed that the most reliable predictions are made by the combined use of different methods, each of which is based on a different statistic. The ANOVA-based bootstap method used in this study performed rather poorly in identifying differentially expressed genes.

We propose that the Magno (M)-channel filter, belonging to the extended classical receptive field (ECRF) model, provides us with "vision at a glance", by performing smoothing with edge preservation. We compare the performance of the M-channel filter with the well-known bilateral filter in achieving such "vision at a glance" which is akin to image preprocessing in the computer vision domain. We find that at higher noise levels, the M-channel filter performs better than the bilateral filter in terms of reducing noise while preserving edge details. The M-channel filter is also significantly simpler and therefore faster than the bilateral filter. Overall, the M-channel filter enables us to model, simulate and arrive at a better understanding of some of the initial mechanisms in visual pathway, while simultaneously providing a fast, biologically inspired algorithm for digital image preprocessing.

In recent years, biometric authentication systems have remained a hot research topic, as they can recognize or authenticate a person by comparing their data to other biometric data stored in a database. Fingerprints, palm prints, hand vein, finger vein, palm vein, and other anatomic or behavioral features have all been used to develop a variety of biometric approaches. Finger vein recognition (FVR) is a common method of examining the patterns of the finger veins for proper authentication among the various biometrics. Finger vein acquisition, preprocessing, feature extraction, and authentication are all part of the proposed intelligent deep learning-based FVR (IDL-FVR) model. Infrared imaging devices have primarily captured the use of finger veins. Furthermore, a region of interest extraction process is carried out in order to save the finger part. The shark smell optimization algorithm is used to tune the hyperparameters of the bidirectional long–short-term memory model properly. Finally, an authentication process based on Euclidean distance is performed, which compares the features of the current finger vein image to those in the database. The IDL-FVR model surpassed the earlier methods by accomplishing a maximum accuracy of 99.93%. Authentication is successful when the Euclidean distance is small and vice versa.

Facial Expression (FE) encompasses information concerning the emotional together with the physical state of a human. In the last few years, FE Recognition (FER) has turned out to be a propitious research field. FER is the chief processing technique for non-verbal intentions, and also it is a significant and propitious computer vision together with the artificial intelligence field. As a novel machine learning theory, Deep Learning (DL) not only highlights the depth of the learning model but also emphasizes the significance of Feature Learning (FL) for the network model, and it has made several research achievements in FER. Here, the present research states are examined typically from the latest FE extraction algorithm as well as the FER centered on DL. The research on classifiers gathered from recent papers discloses a more powerful as well as reliable comprehending of the peculiar traits of classifiers for research fellows. At the ending of the survey, few problems in addition to chances that are required to be tackled in the upcoming future are presented.

Electrocardiogram (ECG) signal is a non-invasive method, used to diagnose the patients with cardiac abnormalities. The subjective evaluation of interval and amplitude of ECG by physician can be tedious, time consuming, and susceptible to observer bias. ECG signals are generated due to the excitation of many cardiac myocytes and hence resultant signals are non-linear in nature. These subtle changes can be well represented and discriminated in transform and non-linear domains. In this paper, performance of Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD) methods are compared for automated diagnosis of five classes namely Non-ectopic (N), Supraventricular ectopic (S), Ventricular ectopic (V), Fusion (F) and Unknown (U) beats. Six different approaches: (i) Principal Components (PCs) on DCT, (ii) Independent Components (ICs) on DCT, (iii) PCs on DWT, (iv) ICs on DWT, (v) PCs on EMD and (vi) ICs on EMD are employed in this work. Clinically significant features are selected using ANOVA test ($p<0.0001$) and fed to k-Nearest Neighbor (k-NN) classifier. We have obtained a classification accuracy of 99.77% using ICs on DWT method. Consistency of performance is evaluated using Cohen’s kappa statistic. Developed approach is robust, accurate and can be employed for mass diagnosis of cardiac healthcare.

Peptide sequencing plays a fundamental role in proteomics. Tandem mass spectrometry, being sensitive and efficient, is one of the most commonly used techniques in peptide sequencing. Many computational models and algorithms have been developed for peptide sequencing using tandem mass spectrometry. In this paper, we investigate general issues in de novo sequencing, and present results that can be used to improve current de novo sequencing algorithms. We propose a general preprocessing scheme that performs binning, pseudo-peak introduction, and noise removal, and present theoretical and experimental analyses on each of the components. Then, we study the antisymmetry problem and current assumptions related to it, and propose a more realistic way to handle the antisymmetry problem based on analysis of some datasets. We integrate our findings on preprocessing and the antisymmetry problem with some current models for peptide sequencing. Experimental results show that our findings help to improve accuracies for de novo sequencing.

The identification of haplotypes, which encode SNPs in a single chromosome, makes it possible to perform a haplotype-based association test with disease. Given a set of genotypes from a population, the process of recovering the haplotypes, which explain the genotypes, is called haplotype inference (HI). We propose an improved preprocessing method for solving the haplotype inference by pure parsimony (HIPP), which excludes a large amount of redundant haplotypes by detecting some groups of haplotypes that are dispensable for optimal solutions. The method uses only inclusion relations between groups of haplotypes but dramatically reduces the number of candidate haplotypes; therefore, it causes the computational time and memory reduction of real HIPP solvers. The proposed method can be easily coupled with a wide range of optimization methods which consider a set of candidate haplotypes explicitly. For the simulated and well-known benchmark datasets, the experimental results show that our method coupled with a classical exact HIPP solver run much faster than the state-of-the-art solver and can solve a large number of instances that were so far unaffordable in a reasonable time.

Current healthcare applications commonly incorporate the Internet of Things (IoT) and cloud computing ideas. IoT devices provide massive amounts of patient data in the healthcare industry. These data stored in the cloud are analyzed using mobile devices’ built-in storage and processing power. The Internet of Medical Healthcare Things (IoMHT) integrates health monitoring components including sensors and medical equipment to remotely monitor patient records in order to provide more intelligent and sophisticated healthcare services. In this research, we analyze one of the deadliest illnesses with a high fatality rate worldwide, the chronic kidney disease (CKD), to provide the finest healthcare services possible to users of e-health and m-health applications by presenting the IoTC services based on healthcare delivery system for the prediction and observation of CKD with its severity level. The suggested architecture gathers patient data from linked IoT devices and saves it in the cloud alongside real-time data, pertinent medical records that are collected from the UCI Machine Learning Repository, and relevant medical documents. We further use a Deep Neural Network (DNN) classifier to predict CKD and its severity. To boost the effectiveness of the DNN classifier, a Particle Swarm Optimization (PSO)-based feature selection technique is also applied. We compare the performance of the proposed model using different classification measures utilizing different classifiers. A Quick Flower Pollination Algorithm (QFPA)- (DNN)-based IoT and cloud-based CKD diagnosis model, is presented in this paper. The CKD diagnosis steps in the QFPA- DNN model involve data gathering, preparation, feature selection and classification stages.

Accurately identifying the various types of knee osteoarthritis aids in an accurate diagnosis. The unique kind and severity of osteoarthritis enable medical specialists to offer the best management and treatment plans. Knee osteoarthritis greatly affects the living style of people by causing higher anxiety, mental issues, and health issues. Early treatment is possible because of early prediction, which may improve patient outcomes. Individuals may be able to prevent or postpone the development of knee osteoarthritis symptoms. An efficient categorization method for knee osteoarthritis employing the Military Scrutolf optimization-tuned deep Convolutional Neural Network (MSO-DCNN) and the advancement of study into this crippling disorder and the improvement of diagnosis, therapy, resource allocation, and disease monitoring are all made possible by the CNN classifier. The preprocessing of the data, which is carried out in three parts and involves the Circular Fourier Transform, Histogram Equalization, and Multivariate Linear Function, also contributes significantly to the success of this study. The proposed MSO technique, which improves convergence time and fine-tunes the classifier’s weight and Bias parameters, was built utilizing the features of military dogs and scrutolf to assist in getting increased seeking and hunting qualities. The MSO-tuned DCNN classifier’s adjusted weights and bias to give more effective desired classification results without using up more time or storage. By examining the performance measures and comparing the existing techniques to the MDO-based DCNN, the suggested MSO-DCNN improved based on TP accuracy by 1.33%, f1 measure by 2.9%, precision by 0.8%, and recall by 2.905%.

Echocardiogram is the test which uses ultrasound to visualize the various heart-related diseases. In order to improve the pattern recognition accuracy, the Frost-Filterative Fuzzified Gravitational Search-based Shift Invariant Deep Structure Feature Learning (FFFGS-SIDSFL) technique is introduced. The FFFGS-SIDSFL technique takes the echocardiogram videos as input for pattern recognition. The input echocardiogram videos are partitioned into frames. At first, the enhanced frost filtering technique is applied to a frame for removing the speckle noise and increase the quality of image. Second, an optimal combination of the feature selection is performed by applying Stochastic Gradient Learning Fuzzified Gravitational Search algorithm. The fuzzy triangular membership function is applied to enhance the Gravitational Search algorithm. Followed by, the different statistical features such as texture, shape, size and intensity are extracted. Finally, the Gaussian activation function at the output unit is used for matching the learned feature vector with the training feature vector. The matching results provide the accurate pattern recognition. Experimental measurement is conducted for analyzing the performance of FFFGS-SIDSFL technique against the two state-of-the-art methods with different metrics, such as Peak signal to noise ratio, pattern recognition accuracy, computational time, and complexity with respect to a diverse number of electrocardiogram images. Based on this observation, the FFFGS-SIDSFL technique provides the better performance in terms of higher accuracy results than the two other existing approaches. As a future work, a distributed pattern recognition scheme that uses IoT with blockchain as event monitoring is proposed.

In general, nonlinear shape normalization methods for binary images have been used in order to compensate for the shape distortions of handwritten characters. However, in most document image analysis and recognition systems, a gray-scale image is first captured and digitized using a scanner or a video camera, then a binary image is extracted from the original gray-scale image using a certain extraction technique. This binarization process may remove some useful information of character images such as topological features, and introduce noises to character background. These errors are accumulated in nonlinear shape normalization step and transferred to the following feature extraction or recognition step. They may eventually cause incorrect recognition results.

In this paper, we propose nonlinear shape normalization methods for gray-scale handwritten Oriental characters in order to minimize the loss of information caused by binarization and compensate for the shape distortions of characters. Two-dimensional linear interpolation technique has been extended to nonlinear space and the extended interpolation technique has been adopted in the proposed methods to enhance the quality of normalized images.

In order to verify the efficiency of the proposed methods, the recognition rate, the processing time and the computational complexity of the proposed algorithms have been considered. The experimental results demonstrate that the proposed methods are efficient not only to compensate for the shape distortions of handwritten Oriental characters but also to maintain the information in gray-scale Oriental characters.

This paper is about the analysis of sets of constraints, with no further assumptions. We explore the relationship between the minimal representation problem and a certain set covering problem of Boneh. This provides a framework that shows the connection between minimal representations, irreducible infeasible systems, minimal infeasibility sets, as well as other attributes of the preprocessing of mathematical programs. The framework facilitates the development of preprocessing algorithms for a variety of mathematical programs. As some such algorithms require random sampling, we present results to identify those sets of constraints for which all information can be sampled with nonzero probability.