scholarly journals Explaining Machine Learning Models for Clinical Gait Analysis

2022 ◽  
Vol 3 (2) ◽  
pp. 1-27
Djordje Slijepcevic ◽  
Fabian Horst ◽  
Sebastian Lapuschkin ◽  
Brian Horsak ◽  
Anna-Maria Raberger ◽  

Machine Learning (ML) is increasingly used to support decision-making in the healthcare sector. While ML approaches provide promising results with regard to their classification performance, most share a central limitation, their black-box character. This article investigates the usefulness of Explainable Artificial Intelligence (XAI) methods to increase transparency in automated clinical gait classification based on time series. For this purpose, predictions of state-of-the-art classification methods are explained with a XAI method called Layer-wise Relevance Propagation (LRP). Our main contribution is an approach that explains class-specific characteristics learned by ML models that are trained for gait classification. We investigate several gait classification tasks and employ different classification methods, i.e., Convolutional Neural Network, Support Vector Machine, and Multi-layer Perceptron. We propose to evaluate the obtained explanations with two complementary approaches: a statistical analysis of the underlying data using Statistical Parametric Mapping and a qualitative evaluation by two clinical experts. A gait dataset comprising ground reaction force measurements from 132 patients with different lower-body gait disorders and 62 healthy controls is utilized. Our experiments show that explanations obtained by LRP exhibit promising statistical properties concerning inter-class discriminativity and are also in line with clinically relevant biomechanical gait characteristics.

2020 ◽  
Nalika Ulapane ◽  
Karthick Thiyagarajan ◽  
sarath kodagoda

<div>Classification has become a vital task in modern machine learning and Artificial Intelligence applications, including smart sensing. Numerous machine learning techniques are available to perform classification. Similarly, numerous practices, such as feature selection (i.e., selection of a subset of descriptor variables that optimally describe the output), are available to improve classifier performance. In this paper, we consider the case of a given supervised learning classification task that has to be performed making use of continuous-valued features. It is assumed that an optimal subset of features has already been selected. Therefore, no further feature reduction, or feature addition, is to be carried out. Then, we attempt to improve the classification performance by passing the given feature set through a transformation that produces a new feature set which we have named the “Binary Spectrum”. Via a case study example done on some Pulsed Eddy Current sensor data captured from an infrastructure monitoring task, we demonstrate how the classification accuracy of a Support Vector Machine (SVM) classifier increases through the use of this Binary Spectrum feature, indicating the feature transformation’s potential for broader usage.</div><div><br></div>

Diagnostics ◽  
2021 ◽  
Vol 11 (3) ◽  
pp. 574
Gennaro Tartarisco ◽  
Giovanni Cicceri ◽  
Davide Di Pietro ◽  
Elisa Leonardi ◽  
Stefania Aiello ◽  

In the past two decades, several screening instruments were developed to detect toddlers who may be autistic both in clinical and unselected samples. Among others, the Quantitative CHecklist for Autism in Toddlers (Q-CHAT) is a quantitative and normally distributed measure of autistic traits that demonstrates good psychometric properties in different settings and cultures. Recently, machine learning (ML) has been applied to behavioral science to improve the classification performance of autism screening and diagnostic tools, but mainly in children, adolescents, and adults. In this study, we used ML to investigate the accuracy and reliability of the Q-CHAT in discriminating young autistic children from those without. Five different ML algorithms (random forest (RF), naïve Bayes (NB), support vector machine (SVM), logistic regression (LR), and K-nearest neighbors (KNN)) were applied to investigate the complete set of Q-CHAT items. Our results showed that ML achieved an overall accuracy of 90%, and the SVM was the most effective, being able to classify autism with 95% accuracy. Furthermore, using the SVM–recursive feature elimination (RFE) approach, we selected a subset of 14 items ensuring 91% accuracy, while 83% accuracy was obtained from the 3 best discriminating items in common to ours and the previously reported Q-CHAT-10. This evidence confirms the high performance and cross-cultural validity of the Q-CHAT, and supports the application of ML to create shorter and faster versions of the instrument, maintaining high classification accuracy, to be used as a quick, easy, and high-performance tool in primary-care settings.

Ahmad Iwan Fadli ◽  
Selo Sulistyo ◽  
Sigit Wibowo

Traffic accident is a very difficult problem to handle on a large scale in a country. Indonesia is one of the most populated, developing countries that use vehicles for daily activities as its main transportation.  It is also the country with the largest number of car users in Southeast Asia, so driving safety needs to be considered. Using machine learning classification method to determine whether a driver is driving safely or not can help reduce the risk of driving accidents. We created a detection system to classify whether the driver is driving safely or unsafely using trip sensor data, which include Gyroscope, Acceleration, and GPS. The classification methods used in this study are Random Forest (RF) classification algorithm, Support Vector Machine (SVM), and Multilayer Perceptron (MLP) by improving data preprocessing using feature extraction and oversampling methods. This study shows that RF has the best performance with 98% accuracy, 98% precision, and 97% sensitivity using the proposed preprocessing stages compared to SVM or MLP.

Sensors ◽  
2021 ◽  
Vol 21 (21) ◽  
pp. 7417
Alex J. Hope ◽  
Utkarsh Vashisth ◽  
Matthew J. Parker ◽  
Andreas B. Ralston ◽  
Joshua M. Roper ◽  

Concussion injuries remain a significant public health challenge. A significant unmet clinical need remains for tools that allow related physiological impairments and longer-term health risks to be identified earlier, better quantified, and more easily monitored over time. We address this challenge by combining a head-mounted wearable inertial motion unit (IMU)-based physiological vibration acceleration (“phybrata”) sensor and several candidate machine learning (ML) models. The performance of this solution is assessed for both binary classification of concussion patients and multiclass predictions of specific concussion-related neurophysiological impairments. Results are compared with previously reported approaches to ML-based concussion diagnostics. Using phybrata data from a previously reported concussion study population, four different machine learning models (Support Vector Machine, Random Forest Classifier, Extreme Gradient Boost, and Convolutional Neural Network) are first investigated for binary classification of the test population as healthy vs. concussion (Use Case 1). Results are compared for two different data preprocessing pipelines, Time-Series Averaging (TSA) and Non-Time-Series Feature Extraction (NTS). Next, the three best-performing NTS models are compared in terms of their multiclass prediction performance for specific concussion-related impairments: vestibular, neurological, both (Use Case 2). For Use Case 1, the NTS model approach outperformed the TSA approach, with the two best algorithms achieving an F1 score of 0.94. For Use Case 2, the NTS Random Forest model achieved the best performance in the testing set, with an F1 score of 0.90, and identified a wider range of relevant phybrata signal features that contributed to impairment classification compared with manual feature inspection and statistical data analysis. The overall classification performance achieved in the present work exceeds previously reported approaches to ML-based concussion diagnostics using other data sources and ML models. This study also demonstrates the first combination of a wearable IMU-based sensor and ML model that enables both binary classification of concussion patients and multiclass predictions of specific concussion-related neurophysiological impairments.

Noviyanti Santoso ◽  
Wahyu Wibowo ◽  
Hilda Hikmawati

In the data mining, a class imbalance is a problematic issue to look for the solutions. It probably because machine learning is constructed by using algorithms with assuming the number of instances in each balanced class, so when using a class imbalance, it is possible that the prediction results are not appropriate. They are solutions offered to solve class imbalance issues, including oversampling, undersampling, and synthetic minority oversampling technique (SMOTE). Both oversampling and undersampling have its disadvantages, so SMOTE is an alternative to overcome it. By integrating SMOTE in the data mining classification method such as Naive Bayes, Support Vector Machine (SVM), and Random Forest (RF) is expected to improve the performance of accuracy. In this research, it was found that the data of SMOTE gave better accuracy than the original data. In addition to the three classification methods used, RF gives the highest average AUC, F-measure, and G-means score.

2020 ◽  
Christopher Zhou ◽  
William Grumbles ◽  
Thomas Cundari

Six machine learning models (random forest, neural network, support vector machine, k-nearest neighbors, Bayesian ridge regression, least squares linear regression) were trained on a dataset of 3d transition metal-methyl and -methane complexes to predict p<i>K<sub>a</sub></i>(C–H), a property demonstrated to be important in catalytic activity and selectivity. Results illustrate that the machine learning models are quite promising, with RMSE metrics ranging from 4.6 to 8.8 p<i>K<sub>a</sub></i> units, despite the relatively modest amount of data available to train on. Importantly, the machine learning models agreed that (a) conjugate base properties were more impactful than those of the corresponding conjugate acid, and (b) the energy of the highest occupied molecular orbital conjugate base was the most significant input feature in the prediction of p<i>K<sub>a</sub></i>(C–H). Furthermore, results from additional testing conducted using an external dataset of Sc-methyl complexes demonstrated the robustness of all models, with RMSE metrics ranging from 1.5 to 6.6 p<i>K<sub>a</sub></i> units. In all, this research demonstrates the potential of machine learning models in organometallic catalyst development.

Muhammad Azam Zia ◽  
Zhongbao Zhang ◽  
Guangda Li ◽  
Haseeb Ahmad ◽  

Prediction of rising stars has become a core issue in data mining and social networks. Prediction of rising venues could unveil rapidly emerging research venues in citation network. The aim of this research is to predict the rising venues. First, we presented five effective prediction features along with their mathematical formulations for extracting rising venues. The underlying features are composed by incorporating the citation count, publications, cited to and cited by information at venue level. For prediction purpose, we employ four machine learning algorithms including Bayesian Network, Support Vector Machine, Multilayer Perceptron and Random Forest. Experimental results demonstrate that proposed features set are effective for rising venues prediction. Our empirical analysis spotlights the rising venues that demonstrate the continuous improvement over time and finally become the leading scientific venues.

Atmosphere ◽  
2019 ◽  
Vol 10 (5) ◽  
pp. 251 ◽  
Wael Ghada ◽  
Nicole Estrella ◽  
Annette Menzel

Rain microstructure parameters assessed by disdrometers are commonly used to classify rain into convective and stratiform. However, different types of disdrometer result in different values for these parameters. This in turn potentially deteriorates the quality of rain type classifications. Thies disdrometer measurements at two sites in Bavaria in southern Germany were combined with cloud observations to construct a set of clear convective and stratiform intervals. This reference dataset was used to study the performance of classification methods from the literature based on the rain microstructure. We also explored the possibility of improving the performance of these methods by tuning the decision boundary. We further identified highly discriminant rain microstructure parameters and used these parameters in five machine-learning classification models. Our results confirm the potential of achieving high classification performance by applying the concepts of machine learning compared to already available methods. Machine-learning classification methods provide a concrete and flexible procedure that is applicable regardless of the geographical location or the device. The suggested procedure for classifying rain types is recommended prior to studying rain microstructure variability or any attempts at improving radar estimations of rain intensity.

2018 ◽  
Vol 25 (7) ◽  
pp. 855-861 ◽  
Halil Kilicoglu ◽  
Graciela Rosemblat ◽  
Mario Malički ◽  
Gerben ter Riet

Abstract Objective To automatically recognize self-acknowledged limitations in clinical research publications to support efforts in improving research transparency. Methods To develop our recognition methods, we used a set of 8431 sentences from 1197 PubMed Central articles. A subset of these sentences was manually annotated for training/testing, and inter-annotator agreement was calculated. We cast the recognition problem as a binary classification task, in which we determine whether a given sentence from a publication discusses self-acknowledged limitations or not. We experimented with three methods: a rule-based approach based on document structure, supervised machine learning, and a semi-supervised method that uses self-training to expand the training set in order to improve classification performance. The machine learning algorithms used were logistic regression (LR) and support vector machines (SVM). Results Annotators had good agreement in labeling limitation sentences (Krippendorff’s α = 0.781). Of the three methods used, the rule-based method yielded the best performance with 91.5% accuracy (95% CI [90.1-92.9]), while self-training with SVM led to a small improvement over fully supervised learning (89.9%, 95% CI [88.4-91.4] vs 89.6%, 95% CI [88.1-91.1]). Conclusions The approach presented can be incorporated into the workflows of stakeholders focusing on research transparency to improve reporting of limitations in clinical studies.

Sign in / Sign up

Export Citation Format

Share Document