A comparative study on recognizing human activities by applying diverse Machine Learning approaches.

This paper deals with the problem of identifying and recognizing everyday human activities. The main goal is to compare a variety of implemented classification models founded on diverse machine learning approaches; one that utilizes features extracted from the time and frequency domain and three others that take advantage of the attributes of the symbolic space in order to extract conclusions regarding the performance and the potential usefulness of each of them. To guarantee the impartiality of the comparison, we used the signals contained in a free accessible dataset, which are subjected to the same preprocessing, and divided into equal time-length windows. The Nearest Neighour classifier is applied to compare the four approaches.

Proprioceptive Disturbance in Chronic Neck Pain: Discriminate Validity and Reliability of Performance of the Clinical Cervical Movement Sense Test.

Chronic neck pain is associated with sensorimotor dysfunctions, which may develop symptoms, affect daily activities, and prevent recovery. Feasible, reliable, and valid objective methods for the assessment of sensorimotor functions are important to identify movement impairments and guide interventions. The aim of this study was to investigate the discriminative validity of a clinical cervical movement sense test, using a laser pointer and an automatic video-based scoring system. Individuals with chronic neck pain of idiopathic onset (INP), traumatic onset (TNP), and healthy controls (CON) were tested. Associations between movement sense and neck disability were examined and the repeatability of the test was investigated. A total of 106 participants (26 INP, 28 TNP, and 52 CON) were included in a cross-sectional study. Acuity, Speed, Time, and NormAcuity (i.e., normalized acuity by dividing acuity with movement time) were used as outcome measures. ANOVAs were used for group comparisons and Pearson correlations for associations between movement sense variables and neck disability index (NDI). Notably, 60 of the participants (30 CON, 17 INP, and 13 TNP) performed the test on a second occasion to explore test-retest reliability. Results revealed a reduced NormAcuity for both INP and TNP compared with CON (p < 0.05). The neck pain groups had similar Acuity but longer Time compared with CON. Among TNP, there was a fair positive correlation between Acuity and NDI, while there was a negative correlation between Acuity and NDI among INP. Reliability measures showed good to excellent ICC values between tests, but standard error of measurements (SEM) and minimal detectable change (MDC) scores were high. The results showed that NormAcuity is a valuable measure to identify disturbed cervical movement sense among INP and TNP. While Acuity was similar between the groups, different strategies, such as longer Time, to perform the task among neck patient groups were used. Few differences were identified between the neck pain groups, but altered strategies may exist. Reliability was acceptable, and the test is feasible to perform in the clinic. However, the technical complexity of the automated image analysis is a concern. Future developments will provide more feasible solutions.

A Machine Learning Approach for Human Activity Recognition.

Human Activity Recognition (HAR) is becoming a significant issue in modern times and directly impact the field of mobile health. Therefore, it is essential the designing of systems which are capable of recognizing properly the activities conducted by the individuals. In this work, we developed a system using the Internet of Things (IoT) and machine learning technologies in order to monitor and assist individuals in their daily life. We compared the data collected using a mobile application and a wearable device with built-in sensors (accelerometer and gyroscope) with the data of a publicly available dataset. By this way, we were able to validate our results and also investigate the functionality and applicability of the wearable device that we choose for the Human Activity Recognition problem. The classification results for the different types of activities presented using our dataset (99%) outperforms the results from the publicly database (97%).

Track My Health: An IoT Approach for Data Acquisition and Activity Recognition.

Human Activity Recognition (HAR) is an arisen research topic because of its usage of self-care and prevention issues. In our days, the advances of technology (smart-phones, smart-watches, tablets, wristbands) and achievements of Machine Learning provide great opportunities for in-depth research on HAR. Technological gadgets include many sensors that gather various, which in turn are input to machine learning techniques to derive useful information and results about human activities and health conditions. Activity Recognition is mainly based physical sensors attached to the human body, with wearable devices coming with built-in sensors such as the accelerometer, gyroscope. This work presents a system based on the Internet of Things (IoT), that monitoring essential vital signals. A mobile application has designed and developed to collect data from a wearable device with built-in sensors (accelerometer and gyroscope) for different human activities and store them for use in a database. The purpose of this work is to present the module of the system that is responsible for the data acquisition, processing and storage of signals that will feed then the Machine Learning module to identify the human health status.

Abnormal Microvascular Architecture, Fibrosis, and Pericyte Characteristics in the Calf Muscle of Peripheral Artery Disease Patients with Claudication and Critical Limb Ischemia.

Work from our laboratory documents pathological events, including myofiber oxidative damage and degeneration, myofibrosis, micro-vessel (diameter = 50-150 µm) remodeling, and collagenous investment of terminal micro-vessels (diameter ≤ 15 µm) in the calf muscle of patients with Peripheral Artery Disease (PAD). In this study, we evaluate the hypothesis that the vascular pathology associated with the legs of PAD patients encompasses pathologic changes to the smallest micro-vessels in calf muscle. Biopsies were collected from the calf muscle of control subjects and patients with Fontaine Stage II and Stage IV PAD. Slide specimens were evaluated by Quantitative Multi-Spectral and Fluorescence Microscopy. Inter-myofiber collagen, stained with Masson Trichrome (MT), was increased in Stage II patients, and more substantially in Stage IV patients in association with collagenous thickening of terminal micro-vessel walls. Evaluation of the Basement Membrane (BM) of these vessels reveals increased thickness in Stage II patients, and increased thickness, diameter, and Collagen I deposition in Stage IV patients. Coverage of these micro-vessels with pericytes, key contributors to fibrosis and BM remodeling, was increased in Stage II patients, and was greatest in Stage IV patients. Vascular pathology of the legs of PAD patients extends beyond atherosclerotic main inflow arteries and affects the entire vascular tree-including the smallest micro-vessels.

PortWeather: A Lightweight Onboard Solution for Real-Time Weather Prediction.

Maritime journeys significantly depend on weather conditions, and so meteorology has always had a key role in maritime businesses. Nowadays, the new era of innovative machine learning approaches along with the availability of a wide range of sensors and microcontrollers creates increasing perspectives for providing on-board reliable short-range forecasting of main meteorological variables. The main goal of this study is to propose a lightweight on-board solution for real-time weather prediction. The system is composed of a commercial weather station integrated with an industrial IOT-edge data processing module that computes the wind direction and speed forecasts without the need of an Internet connection. A regression machine learning algorithm was chosen so as to require the smallest amount of resources (memory, CPU) and be able to run in a microcontroller. The algorithm has been designed and coded following specific conditions and specifications. The system has been tested on real weather data gathered from static weather stations and onboard during a test trip. The efficiency of the system has been proven through various error metrics.

Investigating pH based evaluation of fetal heart rate (FHR) recordings.

Cardiotocography (CTG) is a standard tool for the assessment of fetal well-being during pregnancy and delivery. However, its interpretation is associated with high inter- and intra-observer variability. Since its introduction there have been numerous attempts to develop computerized systems assisting the evaluation of the CTG recording. Nevertheless these systems are still hardly used in a delivery ward. Two main approaches to computerized evaluation are encountered in the literature; the first one emulates existing guidelines, while the second one is more of a data-driven approach using signal processing and computational methods. The latter employs preprocessing, feature extraction/selection and a classifier that discriminates between two or more classes/conditions. These classes are often formed using the umbilical cord artery pH value measured after delivery. In this work an approach to Fetal Heart Rate (FHR) classification using pH is presented that could serve as a benchmark for reporting results on the unique open-access CTU-UHB CTG database, the largest and the only freely available database of this kind. The overall results using a very small number of features and a Least Squares Support Vector Machine (LS-SVM) classifier, are in accordance to the ones encountered in the literature and outperform the results of a baseline classification scheme proving the utility of using advanced data processing methods. Therefore the achieved results can be used as a benchmark for future research involving more informative features and/or better classification algorithms.

An ordinal classification approach for CTG categorization.

Evaluation of cardiotocogram (CTG) is a standard approach employed during pregnancy and delivery. But, its interpretation requires high level expertise to decide whether the recording is Normal, Suspicious or Pathological. Therefore, a number of attempts have been carried out over the past three decades for development automated sophisticated systems. These systems are usually (multiclass) classification systems that assign a category to the respective CTG. However most of these systems usually do not take into consideration the natural ordering of the categories associated with CTG recordings. In this work, an algorithm that explicitly takes into consideration the ordering of CTG categories, based on binary decomposition method, is investigated. Achieved results, using as a base classifier the C4.5 decision tree classifier, prove that the ordinal classification approach is marginally better than the traditional multiclass classification approach, which utilizes the standard C4.5 algorithm for several performance criteria.

A new clinical test for sensorimotor function of the hand - development and preliminary validation.

BACKGROUND: Sensorimotor disturbances of the hand such as altered neuromuscular control and reduced proprioception have been reported for various musculoskeletal disorders. This can have major impact on daily activities such as dressing, cooking and manual work, especially when involving high demands on precision and therefore needs to be considered in the assessment and rehabilitation of hand disorders. There is however a lack of feasible and accurate objective methods for the assessment of movement behavior, including proprioception tests, of the hand in the clinic today. The objective of this observational cross- sectional study was to develop and conduct preliminary validation testing of a new method for clinical assessment of movement sense of the wrist using a laser pointer and an automatic scoring system of test results. METHODS: Fifty physiotherapists performed a tracking task with a hand-held laser pointer by following a zig-zag pattern as accurately as possible. The task was performed with left and right hand in both left and right directions, with three trials for each hand movement. Each trial was video recorded and analysed with a specifically tailored image processing pipeline for automatic quantification of the test. The main outcome variable was Acuity, calculated as the percent of the time the laser dot was on the target line during the trial. RESULTS: The results showed a significantly better Acuity for the dominant compared to non-dominant hand. Participants with right hand pain within the last 12 months had a significantly reduced acuity (p < 0.05), and although not significant there was also a similar trend for reduced Acuity also for participants with left hand pain. Furthermore, there was a clear negative correlation between Acuity and Speed indicating a speed-accuracy trade off commonly found in manual tasks. The repeatability of the test showed acceptable intra class correlation (ICC2.1) values (0.68-0.81) and standard error of measurement values ranging between 5.0-6.3 for Acuity. CONCLUSIONS: The initial results suggest that the test may be a valid and feasible test for assessment of the movement sense of the hand. Future research should include assessments on different patient groups and reliability evaluations over time and between testers.

Wavelet analysis for detection of phasic electromyographic activity in sleep: influence of mother wavelet and dimensionality reduction.

Phasic electromyographic (EMG) activity during sleep is characterized by brief muscle twitches (duration 100-500ms, amplitude four times background activity). High rates of such activity may have clinical relevance. This paper presents wavelet (WT) analyses to detect phasic EMG, examining both Symlet and Daubechies approaches. Feature extraction included 1s epoch processing with 24 WT-based features and dimensionality reduction involved comparing two techniques: principal component analysis and a feature/variable selection algorithm. Classification was conducted using a linear classifier. Valid automated detection was obtained in comparison to expert human judgment with high (>90%) classification performance for 11/12 datasets.

A three class treatment of the FHR classification problem using latent class analysis labeling.

Electronic Fetal Monitoring in the form of cardiotocography is routinely used for fetal assessment both during pregnancy and delivery. However its interpretation requires a high level of expertise and even then the assessment is somewhat subjective as it has been proven by the high inter and intra-observer variability. Therefore the scientific community seeks for more objective methods for its interpretation. Along this path, presented work proposes a classification approach, which is based on a latent class analysis method that attempts to produce more objective labeling of the training cases, a step which is vital in a classification problem. The method is combined with a simple logistic regression approach under two different schemes: a standard multi-class classification formulation and an ordinal classification one. The results are promising suggesting that more effort should be put in this proposed approach.

Fully unsupervised M-FISH chromosome image characterization.

Chromosome analysis is an important and difficult task for clinical diagnosis and biological research. A color imaging technique, multiplex fluorescent in situ hybridization (M-FISH), has been developed to ease the analysis of the process. Using an M-FISH technique each chromosome class (1,2, …,22,X,Y) is stained with a unique color. However, significant variations between images are observed due to a number of factors such as uneven hybridization and spectral overlap among channels. These types of variations influence the pixel classification accuracy of image classification methods which are supervised and require a set of annotated images for training. In this paper, we present a fully unsupervised M-FISH chromosome image classification methodology. Our main contributions are 1) the assumption that the intensity of a chromosome pixel is sampled from multiple Gaussian components [Gaussian mixture model (GMM)] such that each component corresponds to one chromosome class, and 2) the initialization of the GMM model using the emission information of each chromosome class. This is feasible since prior to the M-FISH image acquirement, we already know which chromosome class is emitting to each of the five M-FISH image channels. The method has been tested on a large number of M-FISH images and an overall accuracy of 89.85% is reported. Our method is unsupervised and presents higher classification accuracy even when it is compared with common supervised based methods. Since the developed classification method does not require training data, it is highly convenient when ground truth does not exist.

A hybrid plaque characterization method using intravascular ultrasound images.

BACKGROUND: Intravascular ultrasound (IVUS) is an invasive imaging modality that provides high resolution cross-sectional images permitting detailed evaluation of the lumen, outer vessel wall and plaque morphology and evaluation of its composition. Over the last years several methodologies have been proposed which allow automated processing of the IVUS data and reliable segmentation of the regions of interest or characterization of the type of the plaque. OBJECTIVE: In this paper we present a novel methodology for the automated identification of different plaque components in grayscale IVUS images. METHODS: The proposed method is based on a hybrid approach that incorporates both image processing techniques and classification algorithms and allows classification of the plaque into three different categories: Hard Calcified, Hard-Non Calcified and Soft plaque. Annotations by two experts on 8 IVUS examinations were used to train and test our method. RESULTS: The combination of an automatic thresholding technique and active contours coupled with a Random Forest classifier provided reliable results with an overall classification accuracy of 86.14%. CONCLUSIONS: The proposed method can accurately detect the plaque using grayscale IVUS images and can be used to assess plaque composition for both clinical and research purposes.

Segmentation of microarray images using pixel classification-comparison with clustering-based methods.

OBJECTIVE: DNA microarray technology yields expression profiles for thousands of genes, in a single hybridization experiment. The quantification of the expression level is performed using image analysis. In this paper we introduce a supervised method for the segmentation of microarray images using classification techniques. The method is able to characterize the pixels of the image as signal, background and artefact. METHODS AND MATERIAL: The proposed method includes five steps: (a) an automated gridding method which provides a cell of the image for each spot. (b) Three multichannel vector filters are employed to preprocess the raw image. (c) Features are extracted from each pixel of the image. (d) The dimension of the feature set is reduced. (e) Support vector machines are used for the classification of pixels as signal, background, artefacts. The proposed method is evaluated using both real images from the Stanford microarray database and simulated images generated by a microarray data simulator. The signal and the background pixels, which are responsible for the quantification of the expression levels, are efficiently detected. RESULTS: A quality measure (qindex) and the pixel-by-pixel accuracy are used for the evaluation of the proposed method. The obtained qindex varies from 0.742 to 0.836. The obtained accuracy for the real images is about 98%, while the accuracies for the good, normal and bad quality simulated images are 96, 93 and 71%, respectively. The proposed classification method is compared to clustering-based techniques, which have been proposed for microarray image segmentation. This comparison shows that the classification-based method reports better results, improving the performance by up to 20%. CONCLUSIONS: The proposed method can be used for segmentation of microarray images with high accuracy, indicating that segmentation can be improved using classification instead of clustering. The proposed method is supervised and it can only be used when training data are available.

A novel semiautomated atherosclerotic plaque characterization method using grayscale intravascular ultrasound images: comparison with virtual histology.

Intravascular ultrasound (IVUS) virtual histology (VH-IVUS) is a new technique, which provides automated plaque characterization in IVUS frames, using the ultrasound backscattered RF-signals. However, its computation can only be performed once per cardiac cycle (ECG-gated technique), which significantly decreases the number of characterized IVUS frames. Also atherosclerotic plaques in images that have been acquired by machines, which are not equipped with the VH software, cannot be characterized. To address these limitations, we have developed a plaque characterization technique that can be applied in grayscale IVUS images. Our semiautomated method is based on a three-step approach. In the first step, the plaque area [region of interest (ROI)] is detected semiautomatically. In the second step, a set of features is extracted for each pixel of the ROI and in the third step, a random forest classifier is used to classify these pixels into four classes: dense calcium, necrotic core, fibrotic tissue, and fibro-fatty tissue. In order to train and validate our method, we used 300 IVUS frames acquired from virtual histology examinations from ten patients. The overall accuracy of the proposed method was 85.65% suggesting that our approach is reliable and may be further investigated in the clinical and research arena.

Enhancement of multichannel chromosome classification using a region-based classifier and vector median filtering.

Multichannel chromosome image acquisition is used for cancer diagnosis and research on genetic disorders. This type of imaging, apart from aiding the cytogeneticist in several ways, facilitates the visual detection of chromosome abnormalities. However, chromosome misclassification errors result from different factors, such as uneven hybridization, spectral overlap among fluors, and biochemical noise. In this paper, we enhance the chromosome classification accuracy by making use of a region Bayes classifier that increases the classification accuracy when compared to the already developed pixel-by-pixel classifier and by incorporating the vector median filtering approach for filtering of the image. The method is evaluated using a publicly available database that contains 183 six-channel chromosome sets of images. The overall improvement on the chromosome classification accuracy is 9.99%, compared to the pixel-by-pixel classifier without filtering. This improvement in the chromosome classification accuracy would allow subtle deoxyribonucleic acid abnormalities to be identified easily. The efficiency of the method might further improve by using features extracted from each region and a more sophisticated classifier.

Comparison of quantitative coronary angiography with intracoronary ultrasound. Can quantitative coronary angiography accurately estimate the severity of a luminal stenosis?

In this study we investigated the accuracy of monoplane and biplane quantitative coronary angiography in estimating the luminal dimensions, using intracoronary ultrasound as gold standard. Biplane angiography and intracoronary ultrasound were performed in 24 arterial segments. The end-diastolic intracoronary ultrasound frames were manually selected and segmented. In 2 end-diastolic X ray projections, quantitative coronary angiography was performed and a novel methodology was applied to register the segmented frames onto the processed angiographic images. The luminal areas determined by quantitative coronary angiography in 1 (monoplane) and 2 projections (mean) were compared with those determined by intracoronary ultrasound. The obtained correlation coefficients for the monoplane and mean estimations were 0.69 +/- 0.12 and 0.77 +/- 0.08, respectively. It would appear that by increasing the angle between the biplane projections, the correlation between intracoronary ultrasound and mean estimations improves. Our results provide evidence that orthogonal biplane angiography is more reliable and should be preferred to assess luminal dimensions.

A classification-based segmentation of cDNA microarray images using Support Vector Machines.

Microarray technology provides a tool for the simultaneous analysis of the expression level for an amount of genes. Microarray studies have been shown that image processing techniques can significantly influence microarray data precision. In this paper we propose a supervised method for the segmentation of microarray images based on classification techniques. Support Vector machine is employed to classify each pixel of the image into signal, background or artefacts. In addition, a preprocessing step is applied in order to filter the initial image. The proposed method is applied both to real and simulated images. The pixels of the image are classified in two classes for the real images and three classes for the simulated one. For this task, an informative set of features is used from both green and red channels. The results obtained indicate high accuracy (approximately 99%).

A watershed based segmentation method for multispectral chromosome images classification.

M-FISH (multicolor fluorescence in situ hybridization) is a recently developed cytogenetic technique for cancer diagnosis and research on genetic disorders which uses 5 fluors to label uniquely each chromosome and a fluorescent DNA stain. In this paper, an automated method for chromosome classification in M-FISH images is presented. The chromosome image is initially decomposed into a set of primitive homogeneous regions through the morphological watershed transform applied to the image intensity gradient magnitude. Each segmented area is then classified using a Bayes classifier. We have evaluated our methodology on a commercial available M-FISH database. The classifier was trained and tested on non-overlapping chromosome images and an overall accuracy of 89% is achieved. By introducing feature averaging on watershed basins, the proposed technique achieves substantially better results than previous methods at a lower computational cost.