Improving sentiment analysis on clinical narratives by exploiting UMLS semantic types.

Sentiments associated with assessments and observations recorded in a clinical narrative can often indicate a patient's health status. To perform sentiment analysis on clinical narratives, domain-specific knowledge concerning meanings of medical terms is required. In this study, semantic types in the Unified Medical Language System (UMLS) are exploited to improve lexicon-based sentiment classification methods. For sentiment classification using SentiWordNet, the overall accuracy is improved from 0.582 to 0.710 by using logistic regression to determine appropriate polarity scores for UMLS 'Disorders' semantic types. For sentiment classification using a trained lexicon, when disorder terms in a training set are replaced with their semantic types, classification accuracies are improved on some data segments containing specific semantic types. To select an appropriate classification method for a given data segment, classifier combination is proposed. Using classifier combination, classification accuracies are improved on most data segments, with the overall accuracy of 0.882 being obtained.

Analysis of Optimal Sensor Positions for Activity Classification and Application on a Different Data Collection Scenario.

This paper focuses on optimal sensor positioning for monitoring activities of daily living and investigates different combinations of features and models on different sensor positions, i.e., the side of the waist, front of the waist, chest, thigh, head, upper arm, wrist, and ankle. Nineteen features are extracted, and the feature importance is measured by using the Relief-F feature selection algorithm. Eight classification algorithms are evaluated on a dataset collected from young subjects and a dataset collected from elderly subjects, with two different experimental settings. To deal with different sampling rates, signals with a high data rate are down-sampled and a transformation matrix is used for aligning signals to the same coordinate system. The thigh, chest, side of the waist, and front of the waist are the best four sensor positions for the first dataset (young subjects), with average accuracy values greater than 96%. The best model obtained from the first dataset for the side of the waist is validated on the second dataset (elderly subjects). The most appropriate number of features for each sensor position is reported. The results provide a reference for building activity recognition models for different sensor positions, as well as for data acquired from different hardware platforms and subject groups.

False alarm reduction in BSN-based cardiac monitoring using signal quality and activity type information.

False alarms in cardiac monitoring affect the quality of medical care, impacting on both patients and healthcare providers. In continuous cardiac monitoring using wireless Body Sensor Networks (BSNs), the quality of ECG signals can be deteriorated owing to several factors, e.g., noises, low battery power, and network transmission problems, often resulting in high false alarm rates. In addition, body movements occurring from activities of daily living (ADLs) can also create false alarms. This paper presents a two-phase framework for false arrhythmia alarm reduction in continuous cardiac monitoring, using signals from an ECG sensor and a 3D accelerometer. In the first phase, classification models constructed using machine learning algorithms are used for labeling input signals. ECG signals are labeled with heartbeat types and signal quality levels, while 3D acceleration signals are labeled with ADL types. In the second phase, a rule-based expert system is used for combining classification results in order to determine whether arrhythmia alarms should be accepted or suppressed. The proposed framework was validated on datasets acquired using BSNs and the MIT-BIH arrhythmia database. For the BSN dataset, acceleration and ECG signals were collected from 10 young and 10 elderly subjects while they were performing ADLs. The framework reduced the false alarm rate from 9.58% to 1.43% in our experimental study, showing that it can potentially assist physicians in diagnosing a vast amount of data acquired from wireless sensors and enhance the performance of continuous cardiac monitoring.

Automatic fall monitoring: a review.

Falls and fall-related injuries are major incidents, especially for elderly people, which often mark the onset of major deterioration of health. More than one-third of home-dwelling people aged 65 or above and two-thirds of those in residential care fall once or more each year. Reliable fall detection, as well as prevention, is an important research topic for monitoring elderly living alone in residential or hospital units. The aim of this study is to review the existing fall detection systems and some of the key research challenges faced by the research community in this field. We categorize the existing platforms into two groups: wearable and ambient devices; the classification methods are divided into rule-based and machine learning techniques. The relative merit and potential drawbacks are discussed, and we also outline some of the outstanding research challenges that emerging new platforms need to address.

Toward continuous ambulatory monitoring using a wearable and wireless ECG- recording system: a study on the effects of signal quality on arrhythmia detection.

Five well-known arrhythmia classification algorithms were compared in this paper based on the recommendations in AAMI standard. They are C4.5, k-Nearest Neighbor, Multilayer Perceptron, PART, and Support Vector Machine, respectively, with inputs related to heartbeat intervals and ECG morphological features. They were evaluated on three independent datasets, including the MIT-BIH arrhythmia database, a collection of ECG signals acquired from healthy subjects by the wireless Body Sensor Network (BSN) nodes, and a third dataset captured also by the BSN nodes. Results showed the overall accuracy on the MIT-BIH arrhythmia database was approximately 99.04%, with high sensitivity, specificity, and selectivity. When tested with ECG signals acquired from the human subjects, which were partially deteriorated due to several factors, e.g., motion artifacts and data transmission problems, the overall accuracy of 94.19% and that of 81.22% were obtained for static activities and dynamic activities, respectively. In addition, the effects of the signal quality from these human subjects on false alarms were investigated. When false alarms occurring in signal segments with low quality were excluded, the number of false detections reduced from 14.17% to 8.65%. When evaluated on signals generated by the patient simulator, which included several types of premature ventricular contraction without artifacts from body movements, a high classification accuracy was also observed.