scholarly journals A Spasticity Assessment Method for Voluntary Movement using Data Fusion and Machine Learning

2020 ◽  
Author(s):  
yan chen ◽  
Song Yu ◽  
Qing Cai ◽  
Shuangyuan Huang ◽  
Ke Ma ◽  
...  
Keyword(s):  
Machine Learning ◽  
Classification Accuracy ◽  
Voluntary Movement ◽  
Muscle Activation ◽  
Evaluation Method ◽  
Rating Scale ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Measurement Unit ◽  
Stroke Patients

Abstract Background: Spasticity is a common complication of stroke. Effective spasticity management can improve patients' recovery efficiency and reduce patients' pain. The present clinical spasticity rating scale exhibits subjectivity and a ceiling effect, which makes it difficult to evaluate spasm objectively and to clinically analyze the pathological mechanism of spasticity. The sensor-based quantitative evaluation method is an effective substitute for the clinical spasm rating scale, but currently, it mainly focuses on the spasm evaluation of passive motion. The study of spasmodic state under active exercise can provide a basis for treatment and rehabilitation training, but the evaluation method of spasmodic state under active exercise has not yet been established. Therefore, we combine inertial measurement unit (IMU) and surface electromyography (sEMG) to test the feasibility of assessing spasticity patterns in stroke patients during voluntary movement. Methods: Nine stroke patients with varying degrees of spasticity and four healthy subjects performed isometric elbow exercises. sEMG and kinematics signals were recorded for all participants. The Empirical Mode Decomposition (EMD) algorithm and double threshold algorithms were used to separate sEMG of involuntary muscle activation from voluntary activation. Then, feature extraction and feature fusion were performed. Four common machine learning algorithms are used to monitor and evaluate spasticity patterns. The validity of the proposed method is verified by comparing the classification accuracy of four machine learning models. Results: Cross-validation yielded high classification accuracies (F1-score>0.88) for all four machine learning classifiers in assessing spasticity patterns. The highest detection performance was obtained using the Random Forest algorithm (average accuracy = 0.979; macro-F1 = 0.976). Conclusions: We present a novel method for assessing post-stroke spasticity based on voluntary movement and machine learning. Good classification performance verifies the feasibility of evaluating spasticity patterns by our method. Reliable classification accuracy achieved by the machine learning algorithms indicated the potential to evaluate spasticity patterns using IMU and sEMG when stroke survivors perform voluntary movements.

scholarly journals Improving Automatic Control of an Ankle-Foot Prosthesis Using Machine Learning Algorithms

2017 ◽  
Author(s):  
Stuart R. Fairhurst ◽  
Sara R. Koehler-McNicholas ◽  
Billie C. S. Slater ◽  
Eric A. Nickel ◽  
Karl A. Koester ◽  
...  
Keyword(s):  
Machine Learning ◽  
Lower Limb ◽  
Learning Algorithms ◽  
User Interaction ◽  
Standing Position ◽  
Machine Learning Algorithms ◽  
Measurement Unit ◽  
Smartphone App ◽  
Automatic Mode ◽  
Va Health Care System

Most commercially available lower-limb prostheses are designed for walking, not for standing. The Minneapolis VA Health Care System has developed a bimodal prosthetic ankle-foot system with distinct modes for walking and standing [1]. With this device, a prosthesis user can select standing or walking mode in order to maximize standing stability or walking functionality, depending on the activity and context. Additionally, the prosthesis was designed to allow for an “automatic mode” to switch between standing and walking modes based on readings from an onboard Inertial Measurement Unit (IMU) without requiring user interaction to manually switch modes. A smartphone app was also developed to facilitate changing between walking, standing and automatic modes. The prosthesis described in [1] was used in a pilot study with 18 Veterans with lower-limb amputations to test static, dynamic, and functional postural stability. As part of the study, 17 Veterans were asked for qualitative feedback on the bimodal ankle-foot system (Table 1). The majority of participants (82%) expressed an interest in having an automatic mode. The participants also indicated that the automatic mode would need to reach walking mode on their first step and to lock the ankle quickly once the standing position was achieved. When asked about how they wanted to control the modes of the prosthesis, 82% wanted to use a physical switch and only 12% wanted to use a smartphone app. The results indicated that the following major design changes would be needed: 1) A fast and accurate automatic mode 2) A physical switch for mode changes This paper describes the use of machine learning algorithms to create an improved automatic mode and the use of stakeholder feedback to design a physical switch for the bimodal ankle-foot system.

scholarly journals Question terminology and representation for question type classification

Terminology ◽  
2004 ◽  
Vol 10 (1) ◽  
pp. 153-168 ◽  
Author(s):  
Noriko Tomuro
Keyword(s):  
Machine Learning ◽  
Classification Accuracy ◽  
Cross Validation ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Question Type ◽  
Semantic Features ◽  
Feature Sets ◽  
Fixed Expressions ◽  
Type Classification

Question terminology is a set of terms which appear in keywords, idioms and fixed expressions commonly observed in questions. This paper investigates ways to automatically extract question terminology from a corpus of questions and represent them for the purpose of classifying by question type. Our key interest is to see whether or not semantic features can enhance the representation of strongly lexical nature of question sentences. We compare two feature sets: one with lexical features only, and another with a mixture of lexical and semantic features. For evaluation, we measure the classification accuracy made by two machine learning algorithms, C5.0 and PEBLS, by using a procedure called domain cross-validation, which effectively measures the domain transferability of features.

scholarly journals Detection of Unilateral Arm Paresis after Stroke by Wearable Accelerometers and Machine Learning

Sensors ◽  
2021 ◽  
Vol 21 (23) ◽  
pp. 7784
Author(s):  
Johan Wasselius ◽  
Eric Lyckegård Finn ◽  
Emma Persson ◽  
Petter Ericson ◽  
Christina Brogårdh ◽  
...  
Keyword(s):  
Machine Learning ◽  
Deep Learning ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Accelerometer Data ◽  
Learning Models ◽  
Stroke Patients ◽  
Convolutional Network ◽  
Stroke Treatment ◽  
Data Windows

Recent advances in stroke treatment have provided effective tools to successfully treat ischemic stroke, but still a majority of patients are not treated due to late arrival to hospital. With modern stroke treatment, earlier arrival would greatly improve the overall treatment results. This prospective study was performed to asses the capability of bilateral accelerometers worn in bracelets 24/7 to detect unilateral arm paralysis, a hallmark symptom of stroke, early enough to receive treatment. Classical machine learning algorithms as well as state-of-the-art deep neural networks were evaluated on detection times between 15 min and 120 min. Motion data were collected using triaxial accelerometer bracelets worn on both arms for 24 h. Eighty-four stroke patients with unilateral arm motor impairment and 101 healthy subjects participated in the study. Accelerometer data were divided into data windows of different lengths and analyzed using multiple machine learning algorithms. The results show that all algorithms performed well in separating the two groups early enough to be clinically relevant, based on wrist-worn accelerometers. The two evaluated deep learning models, fully convolutional network and InceptionTime, performed better than the classical machine learning models with an AUC score between 0.947–0.957 on 15 min data windows and up to 0.993–0.994 on 120 min data windows. Window lengths longer than 90 min only marginally improved performance. The difference in performance between the deep learning models and the classical models was statistically significant according to a non-parametric Friedman test followed by a post-hoc Nemenyi test. Introduction of wearable stroke detection devices may dramatically increase the portion of stroke patients eligible for revascularization and shorten the time to treatment. Since the treatment effect is highly time-dependent, early stroke detection may dramatically improve stroke outcomes.

scholarly journals A Comparison of Human against Machine-Classification of Spatial Audio Scenes in Binaural Recordings of Music

2020 ◽  
Vol 10 (17) ◽  
pp. 5956
Author(s):  
Sławomir K. Zieliński ◽  
Hyunkook Lee ◽  
Paweł Antoniuk ◽  
Oskar Dadan
Keyword(s):  
Machine Learning ◽  
Classification Accuracy ◽  
Screening Test ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Support Vector ◽  
Spatial Audio ◽  
Extreme Gradient Boosting ◽  
Music Ensemble

The purpose of this paper is to compare the performance of human listeners against the selected machine learning algorithms in the task of the classification of spatial audio scenes in binaural recordings of music under practical conditions. The three scenes were subject to classification: (1) music ensemble (a group of musical sources) located in the front, (2) music ensemble located at the back, and (3) music ensemble distributed around a listener. In the listening test, undertaken remotely over the Internet, human listeners reached the classification accuracy of 42.5%. For the listeners who passed the post-screening test, the accuracy was greater, approaching 60%. The above classification task was also undertaken automatically using four machine learning algorithms: convolutional neural network, support vector machines, extreme gradient boosting framework, and logistic regression. The machine learning algorithms substantially outperformed human listeners, with the classification accuracy reaching 84%, when tested under the binaural-room-impulse-response (BRIR) matched conditions. However, when the algorithms were tested under the BRIR mismatched scenario, the accuracy obtained by the algorithms was comparable to that exhibited by the listeners who passed the post-screening test, implying that the machine learning algorithms capability to perform in unknown electro-acoustic conditions needs to be further improved.

scholarly journals Comparison of classical machine learning algorithms in the task of handwritten digits classification

2021 ◽  
Vol 21 ◽  
pp. 279-286
Author(s):  
Oleksandr Voloshchenko ◽  
Małgorzata Plechawska-Wójcik
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Random Forest ◽  
Classification Accuracy ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Learning Speed ◽  
Host Machine ◽  
Speed Prediction ◽  
Handwritten Digit

The purpose of this paper is to compare classical machine learning algorithms for handwritten number classification. The following algorithms were chosen for comparison: Logistic Regression, SVM, Decision Tree, Random Forest and k-NN. MNIST handwritten digit database is used in the task of training and testing the above algorithms. The dataset consists of 70,000 images of numbers from 0 to 9. The algorithms are compared considering such criteria as the learning speed, prediction construction speed, host machine load, and classification accuracy. Each algorithm went through the training and testing phases 100 times, with the desired KPIs retained at each iteration. The results were averaged to reach reliable outcomes.

scholarly journals Classification of Control and Neurodegenerative Disease Subjects Using Tree Based Classifiers

2020 ◽  
pp. 63-73
Author(s):  
Syed Ahsin Ali Shah ◽  
Nazneen Habib ◽  
Wajid Aziz ◽  
Ehsan Ullah Khan ◽  
Malik Sajjad Ahmed Nadeem
Keyword(s):  
Machine Learning ◽  
Early Detection ◽  
Classification Accuracy ◽  
Neural Control ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Machine Learning Techniques ◽  
Pharmacological Interventions ◽  
Non Invasive

Background: The medical researchers are developing different non-invasive methods for early detection of Neurodegenerative Diseases (NDDs) when pharmacological interventions are still possible to further prevent the disease progression. The NDDs are associated with the degradation in the complex gait dynamics and motor activity. The classification of gait data using machine learning techniques can assist the physicians for early diagnosis of the neural disorder when clinical manifestation of the diseases is not yet apparent. Aims: The present study was undertaken to classify the control and NDD subjects using decision trees based classifiers (Random Forest (RF), J48 and REPTree). Methodology: The data used in the study comprises of 16 control, 20 Huntington’s Disease (HD), 15 Parkinson’s Disease (PD), and 13 Amyotrophic Lateral Sclerosis (ALS) subjects, which were taken from publicly available database from Physionet. The age range of control subjects was 20-74, HD subjects was 36-70, PD subjects was 44-80, and ALS subjects was 29-71. There were 13 attributes associated with the data. Important features/attributes of the data were selected using correlation feature selection - subset evaluation (cfs) method. Three tree based machine learning algorithms (RF, J48 and REPTree) were used to classify the control and NDD subjects. The performance of classifiers were evaluated using Precision, Recall, F-Measure, MAE and RMSE. Results: In order to evaluate the performance of tree based classifiers, two different settings of data i.e. complete features and selected features were used. In classifying control vs HD subjects, RF provides the robust separation with classification accuracy of 84.79% using complete features and 83.94% using selected features. While in classifying control vs PD subjects, and control vs ALS subjects, RF also provides the best separation with classification accuracy of 86.51% and 94.95% respectively using complete features and 85.19% and 93.64% respectively using selected features. Conclusion: The variability analysis of physiological signals provides a valuable non-invasive tool for quantifying the system of dynamics of healthy subjects and to examine the alternations in the controlling mechanism of these systems with aging and disease. It is concluded that selected features encode adequate information about neural control of the gait. Moreover, the selected features along with tree based machine learning algorithms can play a vital for early detection of NDDs, when pharmacological interventions are still possible.

A feature-centric spam email detection model using diverse supervised machine learning algorithms

2020 ◽  
Vol 38 (3) ◽  
pp. 633-657
Author(s):  
Ammara Zamir ◽  
Hikmat Ullah Khan ◽  
Waqar Mehmood ◽  
Tassawar Iqbal ◽  
Abubakker Usman Akram
Keyword(s):  
Machine Learning ◽  
Sentiment Analysis ◽  
Classification Accuracy ◽  
Research Study ◽  
Learning Algorithms ◽  
Machine Learning Algorithms ◽  
Supervised Machine Learning ◽  
Content Type ◽  
Detection Model ◽  
Proposed Model

Purpose This research study proposes a feature-centric spam email detection model (FSEDM) based on content, sentiment, semantic, user and spam-lexicon features set. The purpose of this study is to exploit the role of sentiment features along with other proposed features to evaluate the classification accuracy of machine learning algorithms for spam email detection. Design/methodology/approach Existing studies primarily exploits content-based feature engineering approach; however, a limited number of features is considered. In this regard, this research study proposed a feature-centric framework (FSEDM) based on existing and novel features of email data set, which are extracted after pre-processing. Afterwards, diverse supervised learning techniques are applied on the proposed features in conjunction with feature selection techniques such as information gain, gain ratio and Relief-F to rank most prominent features and classify the emails into spam or ham (not spam). Findings Analysis and experimental results indicated that the proposed model with sentiment analysis is competitive approach for spam email detection. Using the proposed model, deep neural network applied with sentiment features outperformed other classifiers in terms of classification accuracy up to 97.2%. Originality/value This research is novel in this regard that no previous research focuses on sentiment analysis in conjunction with other email features for detection of spam emails.

Sign in / Sign up

Export Citation Format

Share Document