Soft Transducer for Patient’s Vitals Telemonitoring with Deep Learning-Based Personalized Anomaly Detection

This work addresses the design, development and implementation of a 4.0-based wearable soft transducer for patient-centered vitals telemonitoring. In particular, first, the soft transducer measures hypertension-related vitals (heart rate, oxygen saturation and systolic/diastolic pressure) and sends the data to a remote database (which can be easily consulted both by the patient and the physician). In addition to this, a dedicated deep learning algorithm, based on a Long-Short-Term-Memory Autoencoder, was designed, implemented and tested for providing an alert when the patient’s vitals exceed certain thresholds, which are automatically personalized for the specific patient. Furthermore, a mobile application (EcO2u) was developed to manage the entire data flow and facilitate the data fruition; this application also implements an innovative face-detection algorithm that ensures the identity of the patient. The robustness of the proposed soft transducer was validated experimentally on five individuals, who used the system for 30 days. The experimental results demonstrated an accuracy in anomaly detection greater than 93%, with a true positive rate of more than 94%.

Automated histopathological evaluation of pterygium using artificial intelligence

PurposeThis study aimed to evaluate the efficacy of a new automated method for the evaluation of histopathological images of pterygium using artificial intelligence.MethodsAn in-house software for automated grading of histopathological images was developed. Histopathological images of pterygium (400 images from 40 patients) were analysed using our newly developed software. Manual grading (I–IV), labelled based on an established scoring system, served as the ground truth for training the four-grade classification models. Region of interest segmentation was performed before the classification of grades, which was achieved by the combination of expectation-maximisation and k-nearest neighbours. Fifty-five radiomic features extracted from each image were analysed with feature selection methods to examine the significant features. Five classifiers were evaluated for their ability to predict quantitative grading.ResultsAmong the classifier models applied for automated grading in this study, the bagging tree showed the best performance, with a 75.9% true positive rate (TPR) and 75.8% positive predictive value (PPV) in internal validation. In external validation, the method also demonstrated reproducibility, with an 81.3% TPR and 82.0% PPV for the average of four classification grades.ConclusionsOur newly developed automated method for quantitative grading of histopathological images of pterygium may be a reliable method for quantitative analysis of histopathological evaluation of pterygium.

Heart Disease Prediction Using Glowworm Swarm Optimization and Support Vector Machine Classifier

At present, health disorder is growing day by way of the day due to existence lifestyle, hereditary. Particularly, heart disease has ended up greater frequent these days. Heart disorder prognosis technique is very quintessential and integral trouble for the patient's health. Besides, it will help out to limit disorder to a larger distinctive level. The role of using strategy like machine learning and algorithm such as heart disease diagnosis using Data Mining(DM) techniques is very significant. In the previous system, the Fuzzy Extreme Learning Machine (FELM) was proposed to predict heart disease, ensuring an accurate and timely diagnosis. However, it only achieves 87.14 % of accuracy. To improve the classification accuracy, the proposed system designed an Improved Step Adjustment based Glowworm Swarm Optimization Algorithm with Weighted Feature based Support Vector Machine (ISAGSO-WFSVM) for Heart disease diagnosis. This proposed venture utilizes the dataset of heart disease for input. Using the Improved Step Adjustment based Glowworm Swarm Optimization Algorithm (ISAGSO) to enhance the true positive rate, optimal features are then selected. Finally, with the aid of the Weighted Feature based Support Vector Machine (WFSVM) classifier, classification is carried out relying selected features. In the proposed method, better performance obtained and that is validated through the experimental results in terms of precision, accuracy, recall and f-measures

Heat Stroke Prevention in Hot Specific Occupational Environment Enhanced by Supervised Machine Learning with Personalized Vital Signs

Recently, wet-bulb globe temperature (WBGT) has attracted a lot of attention as a useful index for measuring heat strokes even when core body temperature cannot be available for the prevention. However, because the WBGT is only valid in the vicinity of the WBGT meter, the actual ambient heat could be different even in the same room owing to ventilation, clothes, and body size, especially in hot specific occupational environments. To realize reliable heat stroke prevention in hot working places, we proposed a new personalized vital sign index, which is combined with several types of vital data, including the personalized heat strain temperature (pHST) index based on the temperature/humidity measurement to adjust the WBGT at the individual level. In this study, a wearable device was equipped with the proposed pHST meter, a heart rate monitor, and an accelerometer. Additionally, supervised machine learning based on the proposed personalized vital index was introduced to improve the prevention accuracy. Our developed system with the proposed vital sign index achieved a prevention accuracy of 85.2% in a hot occupational experiment in the summer season, where the true positive rate and true negative rate were 96.3% and 83.7%, respectively.

Preoperative Diagnosis of Hepatocellular Carcinoma Patients with Bile Duct Tumor Thrombus using Deep Learning Method

Preoperative diagnosis of bile duct tumor thrombus (BDTT) is clinically important as the surgical prognosis of hepatocellular carcinoma (HCC) patients with BDTT is significantly different from that of patients without BDTT. The current diagnosis of BDTT is usually based on identifying dilated bile ducts (DBDs) on medical images (eg., CT and MRI images). However, it is easy for doctors to ignore DBDs when reporting imaging scan results, leading to a high misdiagnosis rate in practice. The aim of the present study was to develop an artificial intelligence (AI) pipeline for diagnosing HCC patients with BDTT using medical images. The proposed AI pipeline includes two stages. First, the object detection neural network Faster R-CNN is adopted to identify DBDs; then, an HCC patient is diagnosed to have BDTT if the proportion of images with at least one identified DBD exceeds some threshold value. The proposed AI pipeline was applied to a real dataset consisting of 2,611 CT images collected from 34 HCC patients (16 with BDTT and 18 without BDTT). The average true positive rate for identifying DBDs per patient was 0.92, while the patient-level true positive rate for diagnosing BDTT was 0.94. The area under ROC curve for patient-level diagnosis of BDTT was 0.92 (95% CI: 0.83, 1.00), compared with 0.71 (95% CI: 0.52, 0.89) by random forest based on preoperative clinical variables. These results demonstrated that the proposed AI pipeline is successful in the diagnosis of BDTT. The automatic detection of DBDs is a key step in early diagnosis of HCC patients with BDTT, and is helpful in the treatment and prognosis of these patients.

Weighting Methods for Rare Event Identification From Imbalanced Datasets

In machine learning, we often face the situation where the event we are interested in has very few data points buried in a massive amount of data. This is typical in network monitoring, where data are streamed from sensing or measuring units continuously but most data are not for events. With imbalanced datasets, the classifiers tend to be biased in favor of the main class. Rare event detection has received much attention in machine learning, and yet it is still a challenging problem. In this paper, we propose a remedy for the standing problem. Weighting and sampling are two fundamental approaches to address the problem. We focus on the weighting method in this paper. We first propose a boosting-style algorithm to compute class weights, which is proved to have excellent theoretical property. Then we propose an adaptive algorithm, which is suitable for real-time applications. The adaptive nature of the two algorithms allows a controlled tradeoff between true positive rate and false positive rate and avoids excessive weight on the rare class, which leads to poor performance on the main class. Experiments on power grid data and some public datasets show that the proposed algorithms outperform the existing weighting and boosting methods, and that their superiority is more noticeable with noisy data.

Exploring Author Profiling for Fake News Detection

<div><div><div><p>The proliferation of online media allows for the rapid dissemination of unmoderated news, unfortunately including fake news. The extensive spread of fake news poses a potent threat to both individuals and society. This paper focuses on designing author profiles to detect authors who are primarily engaged in publishing fake news articles. We build on the hypothesis that authors who write fake news repeatedly write only fake news articles, at least in short-term periods. Fake news authors have a distinct writing style compared to real news authors, who naturally want to maintain trustworthiness. We explore the potential to detect fake news authors by designing authors’ profiles based on writing style, sentiment, and co-authorship patterns. We evaluate our approach using a publicly available dataset with over 5000 authors and 20000 articles. For our evaluation, we build and compare different classes of supervised machine learning models. We find that the K-NN model performed the best, and it could detect authors who are prone to writing fake news with an 83% true positive rate with only a 5% false positive rate.</p></div></div></div>

Considerations on the region of interest in the ROC space

Receiver Operating Characteristic curves have been widely used to represent the performance of diagnostic tests. The corresponding area under the curve, widely used to evaluate their performance quantitatively, has been criticized in several respects. Several proposals have been introduced to improve area under the curve by taking into account only specific regions of the Receiver Operating Characteristic space, that is, the plane to which Receiver Operating Characteristic curves belong. For instance, a region of interest can be delimited by setting specific thresholds for the true positive rate or the false positive rate. Different ways of setting the borders of the region of interest may result in completely different, even opposing, evaluations. In this paper, we present a method to define a region of interest in a rigorous and objective way, and compute a partial area under the curve that can be used to evaluate the performance of diagnostic tests. The method was originally conceived in the Software Engineering domain to evaluate the performance of methods that estimate the defectiveness of software modules. We compare this method with previous proposals. Our method allows the definition of regions of interest by setting acceptability thresholds on any kind of performance metric, and not just false positive rate and true positive rate: for instance, the region of interest can be determined by imposing that [Formula: see text] (also known as the Matthews Correlation Coefficient) is above a given threshold. We also show how to delimit the region of interest corresponding to acceptable costs, whenever the individual cost of false positives and false negatives is known. Finally, we demonstrate the effectiveness of the method by applying it to the Wisconsin Breast Cancer Data. We provide Python and R packages supporting the presented method.

Exploring Author Profiling for Fake News Detection

<div><div><div><p>The proliferation of online media allows for the rapid dissemination of unmoderated news, unfortunately including fake news. The extensive spread of fake news poses a potent threat to both individuals and society. This paper focuses on designing author profiles to detect authors who are primarily engaged in publishing fake news articles. We build on the hypothesis that authors who write fake news repeatedly write only fake news articles, at least in short-term periods. Fake news authors have a distinct writing style compared to real news authors, who naturally want to maintain trustworthiness. We explore the potential to detect fake news authors by designing authors’ profiles based on writing style, sentiment, and co-authorship patterns. We evaluate our approach using a publicly available dataset with over 5000 authors and 20000 articles. For our evaluation, we build and compare different classes of supervised machine learning models. We find that the K-NN model performed the best, and it could detect authors who are prone to writing fake news with an 83% true positive rate with only a 5% false positive rate.</p></div></div></div>

Exploring Author Profiling for Fake News Detection

<div><div><div><p>The proliferation of online media allows for the rapid dissemination of unmoderated news, unfortunately including fake news. The extensive spread of fake news poses a potent threat to both individuals and society. This paper focuses on designing author profiles to detect authors who are primarily engaged in publishing fake news articles. We build on the hypothesis that authors who write fake news repeatedly write only fake news articles, at least in short-term periods. Fake news authors have a distinct writing style compared to real news authors, who naturally want to maintain trustworthiness. We explore the potential to detect fake news authors by designing authors’ profiles based on writing style, sentiment, and co-authorship patterns. We evaluate our approach using a publicly available dataset with over 5000 authors and 20000 articles. For our evaluation, we build and compare different classes of supervised machine learning models. We find that the K-NN model performed the best, and it could detect authors who are prone to writing fake news with an 83% true positive rate with only a 5% false positive rate.</p></div></div></div>