Smart Health最新文献

Correctly recognizing the behaviors of children with Autism Spectrum Disorder (ASD) is of vital importance for the diagnosis of Autism and timely early intervention. However, the observation and recording during the treatment from the parents of autistic children may not be accurate and objective. In such cases, automatic recognition systems based on computer vision and machine learning (in particular deep learning) technology can alleviate this issue to a large extent. Existing human action recognition models can now achieve impressive performance on challenging activity datasets, e.g., daily activity, and sports activity. However, problem behaviors in children with ASD are very different from these general activities, and recognizing these problem behaviors via computer vision is less studied. In this paper, we first evaluate a strong baseline for action recognition, i.e., Video Swin Transformer, on two autism behaviors datasets (SSBD and ESBD) and show that it can achieve high accuracy and outperform the previous methods by a large margin, demonstrating the feasibility of vision-based problem behaviors recognition. Moreover, we propose language-assisted training to further enhance the action recognition performance. Specifically, we develop a two-branch multimodal deep learning framework by incorporating the ”freely available” language description for each type of problem behavior. Experimental results demonstrate that incorporating additional language supervision can bring an obvious performance boost for the autism problem behaviors recognition task as compared to using the video information only (i.e., 3.49% improvement on ESBD and 1.46% on SSBD). Our code and model will be publicly available for reproducing the results.

Background

The COVID-19 pandemic highlighted the challenges of providing quality healthcare to vulnerable populations, especially older adults who are disproportionately affected by health service disruptions. Increasingly, mobile health (mHealth) is used for remote healthcare service delivery in this group; however, a variety of factors may limit its adoption.

Aims

To identify the prevalence of mobile device usage among older adults (65yrs+) who present to acute hospitals and explore their willingness to use mHealth.

Methods

A cross-sectional study was conducted using convenience sampling to recruit adults over 65 years to complete a 28 question, 5-point-Likert questionnaire developed using the Unified Theory of Acceptance and Use of Technology (UTAUT).

Results

This study included 119 older adults. Fifty-three participants (44.5%) did not own a smartphone, and 53 (44.5%) had never used one. Sixty-six participants (55.5%) indicated an intention to use mHealth while 53 (44.5%) were either ambivalent or had no intention to use it. Smartphone owners were significantly more likely to use mHealth (OR:3.27, CI:1.53–6.95) than non-owners. Participants showed high self-efficacy (median = 4.0) and expected mHealth to perform well (median = 3.67) with minimal effort (median = 3.33). Within this cohort, intention to use is predicted by age (β = 0.163, p = 0.03), performance expectancy (β = 0.329, p = 0.01), effort expectancy (β = 0.231, p = 0.01) and subjective health status (β = −0.171, p = 0.01).

Conclusions

Many older adults attending acute hospitals remain disinclined in mHealth. This is associated with minimal experience to mobile devices. Empowering older adults to benefit from the increasingly digital landscape of healthcare will require uncovering creative ways to engage them in programs that increase their use of mHealth services.

Today's the blend of information technology and medicine has been improved patient's life. People can monitor their health status without the aid of a healthcare specialized; healthcare is now ubiquitous and don't limit in the waiting room. Unfortunately, there are rarely worked done on women's healthcare, especially pregnancy healthcare monitoring. In this research, a number of articles that have been specially presented about new measurement systems for daily life and health monitoring systems for pregnant women are investigated. Then, a separate overview of these research was presented based on the type of device used and an explanation of their structure. Finally, a model was designed and proposed to test comprehensive systems for monitoring the health of pregnant women. The proposed model is designed based on wearable and environmental sensors that collect daily medical data from pregnant women.

Older adults are prone to forgetfulness and varying degrees of cognitive impairment, which can lead to not taking medication on time, taking the wrong medication or the wrong dose, all of which can negatively affect a person’s health and recovery from illness. Existing medication reminders, like mobile apps and pill boxes, are neither age-friendly nor designed to minimize the burden of documenting medication adherence. In this paper, we present a Medication Adherence Management System (MAMS) for elders, which is based on a companion robot and a wearable device. The MAMS addresses the key issues of safe medication management: medication reminders, medication confirmation, and medication history recording. Human subject tests were conducted to evaluate the performance, acceptability and usability of the MAMS. Results from 35 human subjects showed that the average scores of the convenience, usefulness, and adoptability of the proposed MAMS were 8.17, 8.49, and 8.23 out of 10, respectively. The System Usability Scale (SUS) scores for the MAMS, the robot, and the wearable device are 75.29, 78.60 and 76.40, respectively. We believe the MAMS has potential use in future in-home geriatric care.

Smartphones have become an indispensable part of our everyday life, influencing various aspects of our routines, from wake-up alarms to managing daily life activities. Nowadays, almost every smartphone has a built-in accelerometer sensor. Motivated by the notable increase in smartphone usage in our everyday life, in this research, we focus on harnessing the potential of smartphone accelerometers to recognize human daily life activities, aiming to leverage the usability and convenience of smartphones. We used smartphone accelerometer data from data collection to daily life activity recognition. To accomplish this, we first collected the smartphone's accelerometer data while performing five activities of daily living (ADLs) namely: moving downstairs, upstairs, running, standing, and walking, from 25 volunteers through a mobile application. After this, we extracted 15 statistical features from the smartphone's accelerometer data to efficiently classify the five referred ADLs. We then applied data pre-processing techniques, i.e., data cleaning and feature extraction. Afterward, we trained nine commonly used machine learning models to recognize five ADLs. Finally, we evaluated and compared the performance of all nine ML models to recognize each activity and analyzed the performance of these trained ML models to identify all five ADLs. The evaluated results revealed that the Adaboost (AB) classifier outperformed all other ML models with 100% area under the curve (AUC), precision, recall, accuracy, and F1-score for recognizing the five ADLs.

Activity classification has become a vital feature of wearable health tracking devices. As innovation in this field grows, wearable devices worn on different parts of the body are emerging. To perform activity classification on a new body location, labeled data corresponding to the new locations are generally required, but this is expensive to acquire. In this work, we present an innovative method to leverage an existing activity classifier, trained on Inertial Measurement Unit (IMU) data from a reference body location (the source domain), in order to perform activity classification on a new body location (the target domain) in an unsupervised way, i.e. without the need for classification labels at the new location. Specifically, given an IMU embedding model trained to perform activity classification at the source domain, we train an embedding model to perform activity classification at the target domain by replicating the embeddings at the source domain. This is achieved using simultaneous IMU measurements at the source and target domains. The replicated embeddings at the target domain are used by a classification model that has previously been trained on the source domain to perform activity classification at the target domain. We have evaluated the proposed methods on three activity classification datasets PAMAP2, MHealth, and Opportunity, yielding high F1 scores of 67.19%, 70.40% and 68.34%, respectively when the source domain is the wrist and the target domain is the torso.

Objectives

Parkinson's disease (PD) is a neurodegenerative chronic disorder with multiple motor and non-motor symptoms. As PD has no ultimate cure, physicians aim to delay PD complications, especially those that degrade the patient's quality of life such as motor symptoms and dyskinesia. Patients' lack of adherence to prescribed medication is a major challenge for physicians, especially for patients suffering from chronic conditions. The Centers for Disease Control and Prevention (CDC) estimates that medication non-adherence causes 30 to 50 percent of chronic disease treatment failures and 125,000 deaths per year in the USA (U.S. Foods and Drugs Administration (FDA) “Why You Need to Take Your Medications as Prescribed or Instructed” https://www.fda.gov/drugs/special-features/why-you-need-take-your-medications-prescribed-or-instructed. June 2021). In PD patients particularly, adherence varies between 10% and 67% (Straka et al., 2019Straka, Igor, et al. "Adherence to pharmacotherapy in patients with Parkinson's disease taking three and more daily doses of medication." Frontiers in neurology 10 (2019): 799).

Objective

The goal of this work is to remotely determine whether PD patients have taken their medication, by analyzing gait data gathered from their smartphone sensors. Using this approach, physicians can track the level of medication adherence of their PD patients.

Methodology

Using data from the mPower study (Bot et al., 2016), we selected 152 PD patients who recorded at least 3 walks before and 3 after taking medications and 304 healthy controls (HC) who recorded 3 walks at minimum. We extracted each subject's gait cycle from their accelerometer and gyroscope sensors data. The sensor data corresponding to gait cycles were fed to DeePaMed; a multilayer Conventional Neural Network (CNN), crafted for patches of gait strides. DeePaMed classified 30 s of a walk as either PD patient “On” vs. “Off” medication, or if the gait data belongs to an HC.

Results

Our DeePaMed model was able to discriminate PD patients on-vs off-medication and baseline HC walk with an accuracy of 98.2%. The accuracy of our CNN model surpassed that of traditional Machine Learning methods by over 17%. We also found that our model performed best with inputs containing a minimum of 10 full gait strides.

Conclusion

Medication non-adherence can be accurately predicted using smartphone sensing of the motor symptoms of PD, suggesting that PD patients’ medication response and non-adherence can be monitored remotely via smartphone-based measures.

Non-invasive blood glucose estimation is an extensively researched area since current gold-standard invasive glucose monitoring methods present numerous inconveniences and challenges in terms of comfort and cost. We present the design, development, and validation of an Internet of Medical Things (IoMT) based wearable device for non-invasive and real-time measurement of blood glucose. This paper presents a diabetic health monitoring platform architecture that consists of (a) a user-worn photoplethysmography (PPG) device, (b) a smart analytics cloud that deploys models for blood glucose estimation, and (c) an end-to-end mobile/web application for monitoring diabetes patients. Blood glucose computation is achieved using a novel light-weight 1-dimensional input-reinforced deep neural network architecture, which we call as GlucoNet. This captures both long and short, temporal and spatial features from the PPG signal. The training and validation of the model were conducted on a dataset of 283 participants which demonstrated a mean absolute percentage error (MAPE) of 17.8% ($\pm$ 12.8%) wherein 100% of predictions fall in the clinically acceptable zones A and B of the Clarke-error grid. The lightweight model is also deployed on edge devices for real-time and offline blood glucose measurement. We report a clinical outcome deployment study and insights from 20,000＋ glucose measurements obtained from another 600 patients. To our knowledge, this is the largest reported work employing a non-calibrated, non-invasive, demography, and time-of-food agnostic IoMT glucose monitoring system that does not require any feature engineering and is capable of running on edge devices.

Malnutrition is a serious and prevalent health problem in the older population, and especially in hospitalised or institutionalised subjects. Accurate and early risk detection is essential for malnutrition management and prevention. M-health services empowered with Artificial Intelligence (AI) may lead to important improvements in terms of a more automatic, objective, and continuous monitoring and assessment. Moreover, the latest Explainable AI (XAI) methodologies may make AI decisions interpretable and trustworthy for end users.

This paper presents a novel AI framework for early and explainable malnutrition risk detection based on heterogeneous m-health data. We performed an extensive model evaluation including both subject-independent and personalised predictions, and the obtained results indicate Random Forest (RF) and Gradient Boosting as the best performing classifiers, especially when incorporating body composition assessment data. We also investigated several benchmark XAI methods to extract global model explanations. Model-specific explanation consistency assessment indicates that each selected model privileges similar subsets of the most relevant predictors, with the highest agreement shown between SHapley Additive ExPlanations (SHAP) and feature permutation method. Furthermore, we performed a preliminary clinical validation to verify that the learned feature-output trends are compliant with the current evidence-based assessment.

Wireless Body Area Network (WBAN) ensures a high-quality healthcare service to patients by providing remote and relentless monitoring of their health conditions. Nevertheless, the patients’ health-related data are very sensitive and require security and privacy while transmitting through WBAN to maximize its benefit. User authentication is one of the primary mechanisms to protect critical data, which verifies the identities of entities involved in data transmission. Hence, in the case of health data, every entity engaged in the data transfer process over WBAN needs to be authenticated. In the literature, an end-to-end user authentication mechanism covering each communicating party must be included. Besides, most of the existing user authentication mechanisms are designed assuming that the patient’s mobile phone is trusted. However, a patient’s mobile phone can be stolen or compromised by various malware, therefore, can behave maliciously. To address these limitations, this paper proposes an end-to-end user authentication and session key agreement scheme between sensors and medical experts where the patient’s mobile phone is semi-trusted. We present a formal security analysis using BAN logic and an informal security analysis of the proposed scheme. Both studies reveal that the proposed methodology is robust against well-known security attacks. We analyze the performance of the proposed scheme by collecting real data in practical deployments and find that our scheme achieves comparable efficiency in computation, communication, and energy usage overheads concerning state-of-the-art methods. Besides, the NS-3 simulation exhibits that our proposed scheme also preserves a satisfactory network performance.