Medical Ultrasound / Diagnostic Sonography / Ultrasonography
Mahsa Arab; Ali Fallah; Saeid Rashidi; Maryam Mehdizadeh Dastjerdi; Nasrin Ahmadinejad
Volume 17, Issue 2 , September 2023, , Pages 140-150
Abstract
Breast cancer stands as the most prevalent form of cancer among women, with over 80% of early-stage breast abnormalities being benign. Timely detection is therefore crucial for prompt intervention. Ultrasound Radio Frequency (US RF) signals represent a non-invasive, and real-time screening method for ...
Read More
Breast cancer stands as the most prevalent form of cancer among women, with over 80% of early-stage breast abnormalities being benign. Timely detection is therefore crucial for prompt intervention. Ultrasound Radio Frequency (US RF) signals represent a non-invasive, and real-time screening method for breast cancer, offering advantages in tissue differentiation and cost-effectiveness without requiring additional equipment. This research aims to present an intelligent approach for the classification of benign, suspicious, and malignant breast lesions based on effective features extracted from the time series. The dataset, registered as USRFTS, comprises 170 instances recorded from 88 patients. The proposed methodology encompasses four key phases: pre-processing, feature extraction, feature selection, and classification. In the pre-processing phase, B-mode images are reconstructed from US RF time series, and a radiologist manually selects the Region of Interest (ROI) in each image. Subsequently, diverse features in the time and frequency domains are extracted from each ROI during the feature extraction stage. The ant colony method is employed for the selection of impactful features. The dataset is then subjected to evaluation using classifiers such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Decision Tree (DT), Linear Discriminant Analysis (LDA), and a reference classification method (RCM). The results demonstrate a maximum classification accuracy of 94.95% for two classes and 93.33% for three classes
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Parastoo Sadeghinia; Hamed Danandeh Hesar
Volume 16, Issue 3 , December 2022, , Pages 271-287
Abstract
Phonocardiography (PCG) signals provide valuable information about the heart valves .These auditory signals can be useful in the early diagnosis of heart diseases. Automatic heart sound classification has a promising potential in the field of heart pathology. In this research, a new method based on machine ...
Read More
Phonocardiography (PCG) signals provide valuable information about the heart valves .These auditory signals can be useful in the early diagnosis of heart diseases. Automatic heart sound classification has a promising potential in the field of heart pathology. In this research, a new method based on machine learning techniques is proposed for discriminating normal and abnormal heart sounds. In this method, first, the heart sounds are segmented into 4 main parts: S1, S2, systole and diastole segments. From these segments, statistical and time-frequency features are extracted for classification. Before classification, the distinctive features are selected using two approaches. In the first approach, the feature selection is accomplished using particle swarm optimization algorithm (PSO). In the second approach, we use Sequential Forward Feature Selection (SFFS) method. The proposed method was evaluated on the Physionet 2016 Challenge database using 10-fold cross-validation method. In this database, the number of normal and abnormal PCG signals are not balanced. Therefore, in this paper, the synthetic minority over-sampling technique (SMOTE) is applied to produce balanced data. The evaluation results showed that the proposed method can distinguish the normal heart sounds from abnormal ones with accuracy of 98/03% and sensitivity and specificity of 97.64%, 98.43%respectively.
Biomedical Image Processing / Medical Image Processing
Sina Shamekhi
Volume 16, Issue 2 , September 2022, , Pages 95-113
Abstract
Intuitive examination of retinal layers in Spectral-Domain Optical Coherence Tomography (SD-OCT) images is one of the main methods used by physicians to diagnose retinal diseases. This method faces challenges such as noise and image complexity and the proximity of retinal layers. In recent years, the ...
Read More
Intuitive examination of retinal layers in Spectral-Domain Optical Coherence Tomography (SD-OCT) images is one of the main methods used by physicians to diagnose retinal diseases. This method faces challenges such as noise and image complexity and the proximity of retinal layers. In recent years, the automatic diagnosis of retinal diseases has become an important clinical issue in computer vision. In this research, a new method for efficient multi-class automatic classification of SD-OCT images has been proposed. This method consists of five stages, preprocessing, layer recognition, feature extraction, and image classification. Examination of the shape of the RNFL layer and IS/OS junction as a clinical method is influential in physicians' decisions to diagnose retinal diseases. Therefore, in this study, inspired by this clinical diagnosis method, the RNFL layer, and the IS/OS junction have been detected by a new method based on the Frangi vessel enhancement algorithm and the gradient of the image. Then, by extracting and selecting several efficient features from the curves of the layers, an algorithm based on the ensemble decision tree has been proposed for classifying SD-OCT images of the retina and presented as a MATLAB application. The proposed method has been evaluated using images of two well-known databases of Duke and Kermany. Based on the results, precision, sensitivity, specificity, accuracy, miss rate and F1-score of the proposed method in Duke database were equal to 98.7, 98.8, 99.4, 99.1, 1.3, and 98.7, respectively, and in Kermany database were 96.8, 96.7, 98.9, 98.4, 3.2 and 96.7 respectively. The results show the superiority of the proposed method compared to other comparative methods. In summary, the use of efficient features of retinal effective layers and a powerful algorithm for classification has improved the performance of the proposed method compared to previous more complex methods.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Faezeh Daneshmand-Bahman; Ateke Goshvarpour
Volume 16, Issue 2 , September 2022, , Pages 115-131
Abstract
Anxiety disorders are one of the most common and debilitating mental disorders worldwide. On the other hand, since 2019, with the outbreak of Covid-19, anxiety has increased among people, especially the medical staff. Currently, anxiety is diagnosed (when the symptoms are severe enough) using a questionnaire ...
Read More
Anxiety disorders are one of the most common and debilitating mental disorders worldwide. On the other hand, since 2019, with the outbreak of Covid-19, anxiety has increased among people, especially the medical staff. Currently, anxiety is diagnosed (when the symptoms are severe enough) using a questionnaire by a specialist. To resolve this shortcoming, researchers have recently paid attention to the use of brain signals. Consequently, the present study aimed to diagnose anxiety using brain signals. The novelty of this study is the use of the Chebyshev chaotic map for the first time in biological signal analysis. It used the DASPS database, which includes a 14-channel electroencephalogram (EEG) of 23 people (10 men and 13 women, with a mean age of 30 years). The self-assessment manikin scores were used to divide anxiety into two and four levels. First, the data were normalized. Then, the chaotic map was reconstructed and divided into 128 strips. The density of points in each of the strips was calculated. Two indicators were considered as features, (1) maximum density and (2) its corresponding sample. Finally, features were applied to Support Vector Machines (SVM) and k-Nearest Neighbors (K-NN) in 5 ways, (1) feature 1 of all channels, (2) feature1 mapping of all channels using principal component analysis (PCA), (3) feature 2 of all channels, (4) feature 2 mapping of all channels using PCA and (5) each feature - each channel separately. The results show a maximum accuracy of 93.75% for diagnosing two levels of anxiety and 96.15% for diagnosing four levels of anxiety. In addition, K-NN outperformed SVM. Accordingly, the proposed algorithm can be introduced as a suitable approach for diagnosing anxiety.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Hamid Shafaatfar; Mehdi Taghizadeh; Morteza Valizadeh; Mohamad Hossein Fatehi
Volume 16, Issue 2 , September 2022, , Pages 147-158
Abstract
Automatic detection of cardiac arrhythmias is very important for the successful treatment of heart disease and machine learning is used for this purpose. To correctly classify arrhythmic classes, it is important to extract the appropriate features to distinguish between different classes. In this paper, ...
Read More
Automatic detection of cardiac arrhythmias is very important for the successful treatment of heart disease and machine learning is used for this purpose. To correctly classify arrhythmic classes, it is important to extract the appropriate features to distinguish between different classes. In this paper, a deep convolutional neural network is used to extract the feature. Due to the fact that the heart rates of different patients are very different, arrhythmia classes will have many intra-class changes. To reduce intra-class changes, each patient’s heart rate is mapped with a dedicated function to increase its resemblance to the heart rate of one of the training patient data’s. The proposed specific mapping reduces intra-class changes and significantly increases the classification accuracy of cardiac arrhythmias. To prove the effectiveness of the proposed method, its results were compared with several new studies based on three criteria for accuracy, sensitivity and specificity and on the same data set. The accuracy obtained is about 96.24%, which shows the better performance of the proposed method compared to other works.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Maryam Dorvashi; Neda Behzadfar; Ghazanfar Shahgholian
Volume 14, Issue 2 , July 2020, , Pages 109-119
Abstract
Consumption of alcohol contributes to disorders in brain. In this study, in order to detect the consumption of alcohol, electroencephalogram (EEG) signal of 20 participants (10 alcoholic and 10 control subjects) recorded by 64 channels was investigated. Frequency and non-frequency features of EEG signal ...
Read More
Consumption of alcohol contributes to disorders in brain. In this study, in order to detect the consumption of alcohol, electroencephalogram (EEG) signal of 20 participants (10 alcoholic and 10 control subjects) recorded by 64 channels was investigated. Frequency and non-frequency features of EEG signal including power spectrum of signal, permutation entropy, approximate entropy, Katz fractal dimension and Petrosion fractal dimension were extracted to analyses the EEG signal. Statistical analysis was used to investigate the significant differences between the alcohol and control groups. The Davis-Bouldin (DB) criterion was used to select the best channel distinguishing between the alcoholic and non-alcoholic EEG signal. Results showed that between frequency features, power of lower2 alpha frequency decreased in alcoholic individuals and regarding the DB criterion, the CP3 channel (DB=1.7638) showed the best discrimination between the alcohol and control groups. Also, among the non-frequency features, the Katz fractal dimension increased in the control group and FP2 channel (DB = 0.862) had the best discrimination. Eventually, power of Lower2-alpha frequency band and Katz fractal dimension fed into the nearest neighbor classifier (KNN), 71% and 93% accuracy were achieved, respectively. According to the results, it can be concluded that the best feature and channel discriminating between alcohol and control groups is the Katz fractal dimension and FP2 channel.
Abolfazl Tabatabaei; Vali Derhami; Razieh Sheikhpour; Mohammad-Reza Pajoohan
Volume 13, Issue 4 , December 2019, , Pages 337-348
Abstract
Feature selection is a well-known preprocessing technique in machine learning, data mining and especially bioinformatics microarray analysis with a high-dimension, low-sample-size (HDLSS) data. The diagnosis of genes responsible for disease using microarray data is an important issue to promoting knowledge ...
Read More
Feature selection is a well-known preprocessing technique in machine learning, data mining and especially bioinformatics microarray analysis with a high-dimension, low-sample-size (HDLSS) data. The diagnosis of genes responsible for disease using microarray data is an important issue to promoting knowledge about the mechanism of disease and improves the way of dealing with the disease. In feature selection methods based on information theory, which cover a wide range of feature selection methods, the concept of entropy is used to define criteria for relevance, redundancy and complementarity. In this paper, we propose a new relevancy criterion based on the concept of pure continuity rather than the concept of entropy. In the proposed method, to control and reduce redundancy, the relevancy between a feature and each class is separately examined, while in most of the filter methods the value of a feature is measured based on its relation to the entire class. This solution allows us to identify the most efficient features (genes) of each class separately, while identifying common features (genes) is also possible. Discretization is another challenge in some available techniques. Using a homomorphism transformation in proposed method avoids engaging with discretization complexities, while taking advantages of it. Seven types of cancer microarrays with three types of classification models (e.g. NB, KNN and SVM) are used to establish a comparison between the proposed method and other relevant methods. The results confirm the efficiency of the proposed method in the term of accuracy and number of selected genes as two parameters of classification.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Raheleh Davoodi; Mohammad Hasan Moradi
Volume 12, Issue 1 , June 2018, , Pages 25-39
Abstract
Depression is one of the most common mental disorders in the current century where early diagnosis can result in better treatment. One of the depression diagnostic methods is the analysis of the brain electrical signals. In this paper, we are seeking for a method to distinguish among the levels of the ...
Read More
Depression is one of the most common mental disorders in the current century where early diagnosis can result in better treatment. One of the depression diagnostic methods is the analysis of the brain electrical signals. In this paper, we are seeking for a method to distinguish among the levels of the depression. The proposed model is a deep rule-based system based on the stacked principle and focuses on the interpretability of the rules alongside high accuracy. Fuzzy systems have the proper capability in the classification of medical data with various levels of uncertainty. Moreover, in the recent years, deep learning has been taken considerable attention in the field of Artificial Intelligence. In this paper, we aim to benefit from capabilities of both fields. The proposed architecture employs a robust fuzzy clustering approach that can determine an appropriate number of clusters in each layer, unsupervised and a hierarchical stacked structure to transfer the interpretable trained rules from the previous layers with the same linguistic labels to the next layer. The interpretability is due to the presence of the input space into the consequent ones. The presence of the output of the previous layer’s rules at the input space of the next parts equals to a fuzzy system with non-linear consequent or the certainty factor in a fuzzy system with linear consequent. EEG data were preprocessed and time, frequency and nonlinear features such as recurrent plot were extracted and selected and after that were employed in the proposed system. The proposed system was compared with common classifiers like Neural Net, Support Vector Machine, Naive Bayes, Decision Tree and Linear Discriminant Analysis. Accuracy results for the test data in 30 folds (49.01% in comparison to 41.42%, 40.47%, 40.01%, 38.35% and 40.28% respectively) demonstrate the considerable performance of the proposed system.
Bioinformatics / Biomedical Informatics / Medical Informatics / Health Informatics
Hossein Bankikoshki; Seyed Ali Seyyedsalehi; Fatemeh Zare Mirakabad
Volume 11, Issue 3 , September 2017, , Pages 219-230
Abstract
The use of genomic nucleotide sequences as biochemical signals in machine learning methods is possible by converting these sequences into numerical codes. This conversion results in an unrealistic increase in the dimension of the data and encounters some data analysis operations such as visualization ...
Read More
The use of genomic nucleotide sequences as biochemical signals in machine learning methods is possible by converting these sequences into numerical codes. This conversion results in an unrealistic increase in the dimension of the data and encounters some data analysis operations such as visualization and feature extraction with constraints. Therefore, one should use the dimensionality reduction technics in order to return the data to its real dimension. In this study, a deep autoencoder neural network has been used to reduce the dimension of binding site sequence data on the human genome. In order to determine whether the information of real data is preserved in compressed data, we perform a two-class classification using a support vector machine. The results show that information is almost entirely preserved in compression. Then, compressed data is used for visualization as well as feature selection by analysis of variance. The results show that the first, the tenth and eighth positions in the sequences are the most informative positions. While the majority of the previous works deal with gene expression data of microarrays and compare a few dimension reduction algorithms, this paper for the first time uses an autoencoder on nucleotide sequence data and provides a comprehensive comparison between the performance of the dimension reduction technics and machine learning algorithms.
Biomedical Image Processing / Medical Image Processing
Amir Ehsan Lashkari; Fatemeh Pak; Mohammad Firouzmand
Volume 9, Issue 1 , April 2015, , Pages 71-84
Abstract
Breast cancer is the most common type of cancer among women. The important key to treat the breast cancer is early detection of it because according to many pathological studies more 80% of all abnormalities are still benign at primary stages; so in recent years, many studies and extensive research done ...
Read More
Breast cancer is the most common type of cancer among women. The important key to treat the breast cancer is early detection of it because according to many pathological studies more 80% of all abnormalities are still benign at primary stages; so in recent years, many studies and extensive research done to early detection of breast cancer with higher precision and accuracy. Infra-red breast thermography is an imaging technique based on recording temperature distribution patterns of breast tissue. Compared with breast mammography technique, thermography is more suitable technique because it is noninvasive, non-contact, passive and free ionizing radiation. In this paper, a full automatic high accuracy technique for classification of suspicious areas in thermogram images with the aim of assisting physicians in early detection of breast cancer has been presented. Proposed algorithm consists of four main steps: pre-processing & segmentation, feature extraction, feature selection and classification. At the first step, using full automatic operation, region of interest (ROI) determined and the quality of image improved. Using thresholding and edge detection techniques, both right and left breasts separated from each other. Then relative suspected areas become segmented and image matrix normalized due to the uniqueness of each person's body temperature. At feature extraction stage, 23 features, including statistical, morphological, frequency domain, histogram and Gray Level Co-occurrence Matrix (GLCM) based features are extracted from segmented right and left breast obtained from step 1. To achieve the best features, feature selection methods such as mRMR, SFS, SBS, SFFS, SFBS and GA have been used at step 3. Finally to classify and TH labeling procedures, different classifiers such as AdaBoost, SVM, kNN, NB and PNN are assessed to find the best suitable one. The results obtained on native database showed the best and significant performance of the proposed algorithm in comprise to the similar studies. According to experimental results, mRMR combined with AdaBoost with the maximum accuracy of 92%, and SFFS combined with AdaBoost with a maximum accuracy of 88%, are the best combination of feature selection and classifier for evaluation of the right and left breast images respectively.
Biomedical Image Processing / Medical Image Processing
Hadi Sabahi; Hamid Soltanian Zadeh; Lisa Scarpace; Tom Mikkelsen
Volume 5, Issue 4 , June 2011, , Pages 289-295
Abstract
In this paper, we propose a method to predict the outcome of Bevacizumab therapy on Glioblastoma Multiform (GBM) tumors. The method uses diffusion anisotropy indices (DAI) and spatial information to predict the treatment response of each tumor voxel. These DAIs are Fractional Anisotropy, Mean Diffusivity, ...
Read More
In this paper, we propose a method to predict the outcome of Bevacizumab therapy on Glioblastoma Multiform (GBM) tumors. The method uses diffusion anisotropy indices (DAI) and spatial information to predict the treatment response of each tumor voxel. These DAIs are Fractional Anisotropy, Mean Diffusivity, Relative Anisotropy, and Volume Ratio, extracted from Diffusion Tensor Imaging (DTI) data before treatment. The spatial information is considered as the distance of each tumor voxel from the tumor center, extracted from pre-treatment post-contrast T1-weighted Magnetic Resonance Images (pc-T1-MRI). DAIs and spatial information of each tumor voxel are considered as feature vector. DTI and pc-T1-MRI are gathered before and after the treatment of seven GBM patients. First, DAIs of all brain voxels and the distance of each tumor voxel from the tumor center are calculated. Second, the method registers pretreatment DAI maps and post-treatment pc-T1-MRI to pre-treatment pc-T1-MRI. Next, the tumor is segmented using thresholding technique from pc-T1-MRI. Then, Gd-enhanced voxels of the pre- and posttreatment pc-T1-MRI are compared to label the feature vectors. Three classifiers were evaluated, including Support Vector Machine, K-Nearest Neighbor, and Artificial Neural Network. Classification results show a preference for K-Nearest Neighbor based on well-established performance measures.
Zahra Amini; Vahid Abootalebi; Mohammad Taghi Sadeghi
Volume 4, Issue 4 , June 2010, , Pages 293-306
Abstract
The aim of this paper is to design a pattern recognition based system to detect P300 component in multi-channel electroencephalogram (EEG) trials. This system has two main blocks, feature extraction and classification. In feature extraction block, in addition to conventional features namely morphological, ...
Read More
The aim of this paper is to design a pattern recognition based system to detect P300 component in multi-channel electroencephalogram (EEG) trials. This system has two main blocks, feature extraction and classification. In feature extraction block, in addition to conventional features namely morphological, frequency and wavelet features, some new features included intelligent segmentation, common spatial pattern (CSP) and combined features (CSP + Segmentation) have also been used. Three criteria were used for evaluation and selection of a feature set by choosing a subset of the original features that contains most of essential information. Firstly, a statistical analysis has been applied for evaluating the fitness of each feature in discriminating between target and non target signals. Secondly, each of these six groups of features was evaluated by a Linear Discriminant Analysis (LDA) classifier. Furthermore by using Stepwise Linear Discriminant Analysis (SWLDA), the best set of features was selected. Among these six feature vectors, intelligent segmentation was seen to be most efficient in classification of these signals. In classification phase, two linear classifiers -LDA and SWLDA- were used. The algorithm was described here has tested with dataset II from the BCI competition 2005. In this research, the best result for P300 detection is 97.05% .This result have proven to be more accurate than the results of previous works carried out in this filed.
Biomedical Signal Processing / Medical Signal Processing / Biosignal Processing
Mohammad Mehdi Ramezani; Ahmad Reza Sharafat
Volume 4, Issue 2 , June 2010, , Pages 123-134
Abstract
In this paper, we propose a novel approach for classification of surface electromyogram (sEMG) signal with a view to controlling myoelectric prosthetic devices. The sEMG signal generated during isometric contraction is modeled by a stochastic process whose probability density function (PDF) is non- Gaussian ...
Read More
In this paper, we propose a novel approach for classification of surface electromyogram (sEMG) signal with a view to controlling myoelectric prosthetic devices. The sEMG signal generated during isometric contraction is modeled by a stochastic process whose probability density function (PDF) is non- Gaussian for low levels of applied force. Since the PDF of ambient noise is assumed to be Gaussian, we extract correntropy features, as they contain information on non-Gaussian components (the sEMG signal) only; and utilize the linear discriminant analysis (LDA) to classify the sEMG signal using correntropy features. Our proposed method has lower classification error and requires much less computations as compared to other existing advanced methods.
Majid Ghoshuni; Mohammad Ali Khalilzadeh; Ali Moghimi
Volume 1, Issue 4 , June 2007, , Pages 251-267
Abstract
Episodic memory is the explicit recollection of incidents occurred at a particular time and place in One’s Personal Past. In This Study, Detection of Episodic Memory Activity In Event Related Potentials (ERPs) was done. ERPs were recorded while the subjects made old/new recognition judgments on ...
Read More
Episodic memory is the explicit recollection of incidents occurred at a particular time and place in One’s Personal Past. In This Study, Detection of Episodic Memory Activity In Event Related Potentials (ERPs) was done. ERPs were recorded while the subjects made old/new recognition judgments on the new unstudied meaningless pictures and the old pictures which had been presented at the study phase. In order to extract the features correlated with the episodic memory activity, time and time-frequency features were extracted from ERPs. Wavelet method was implemented for feature extraction in time-frequency. Independent sample test has was for detection of the separable degree the between old/new ERPs. Furthermore, by using stepwise linear discriminate analysis, ERP signals were classified to old and new classes. Ultimately for better classification between old/new ERPs, Multilayer Perceptron was implemented, and for best feature selection, genetic algorithm was used. In the best results, by using time domain features extracted from Pz channel, 100% accuracy in the training and test data was obtained.
Biomedical Image Processing / Medical Image Processing
Jamal Esmaeilpour; Sattar Mirzakouchaki; Jalil Seyfali Harsini; Abdorrahim Kadkhoda Mohammadi
Volume 1, Issue 3 , June 2007, , Pages 167-176
Abstract
In this paper, the role of Vector Quantizer Neural Network in classification of six types of ECG signals has been investigated using the features that extracted from Daubechies6 Wavelet transformation. The six types of signals are: normal beat, left bundle branch block beat, right bundle branch block ...
Read More
In this paper, the role of Vector Quantizer Neural Network in classification of six types of ECG signals has been investigated using the features that extracted from Daubechies6 Wavelet transformation. The six types of signals are: normal beat, left bundle branch block beat, right bundle branch block beat, premature ventricular contraction paced beat and fusion of paced and normal beats. The required data were obtained from the MIT/BIH arrhythmia databases. By using the annotation files of the databases, the patterns of these six types of ECG signals were separated. Then, for better feature extraction, filtering and scaling on the patterns were applied. We used the energies of the last five detailed signals obtained from the exerting the Wavelet transformation in six levels, as the pattern features for Vector Quantizer Network training and testing. From each class, five hundred patterns were used for network training and one hundred patterns for testing. The results indicated %93.1 accuracy for six classes and above %94.3 for lesser than six classes. Then the rate of similarity and dissimilarity of the classes were considered. Finally, the results of this method were compared with some other methods in terms of accuracy.