PLOS digital health最新文献

Ensuring privacy in distributed machine learning while computing the Area Under the Curve (AUC) is a significant challenge because pooling sensitive test data is often not allowed. Although cryptographic methods can address some of these concerns, they may compromise either scalability or accuracy. In this paper, we present two privacy-preserving solutions for secure AUC computation across multiple institutions: (1) an exact global AUC method that handles ties in prediction scores and scales linearly with the number of samples, and (2) an approximation method that substantially reduces runtime while maintaining acceptable accuracy. Our protocols leverage a combination of homomorphic encryption (modified Paillier), symmetric and asymmetric cryptography, and randomized encoding to preserve the confidentiality of true labels and model predictions. We integrate these methods into the Personal Health Train (PHT)-meDIC platform, a distributed machine learning environment designed for healthcare, to demonstrate their correctness and feasibility. Results using both real-world and synthetic datasets confirm the accuracy of our approach: the exact method computes the true AUC without revealing private inputs, and the approximation provides a balanced trade-off between computational efficiency and precision. All relevant code and data is publicly available at https://github.com/PHT-meDIC/PP-AUC, facilitating straightforward adoption and further development within broader distributed learning ecosystems.

Black women are disproportionally likely to contract sexually transmitted infections (STIs) including HIV compared to women of other races and ethnicities. It is possible that mobile health (referred to as "mHealth") strategies, including mobile applications, designed for Black women could provide sexual health support and reduce STI/HIV transmission. We sought to explore acceptability of mHealth strategies among Black women and to identify if preferences varied by age or HIV vulnerability. We surveyed 213 Black women aged 14-64 attending a family planning clinic in Chicago. We asked about mHealth use, desired sources of sexual health information, and mHealth application (app) features. Responses were analyzed as dichotomous variables, with age categorized as ≤24 years of age or ≥25 years of age and HIV vulnerability score categorized as low (<2) or high (≥2). HIV vulnerability was determined based on affirmative answers to the following questions: having had condomless sex (either vaginal or anal) in the past three months, having had an abortion in the past 12 months, having received STI treatment in the past three months, and having had ≥ 2 sex partners in the last three months. Odds ratios and 95% confidence intervals (OR 95% CI) were created using logistic regression models. The majority of participants were interested in using technology as part of their sexual health care (84.5%) and were likely to download an mHealth app (74.7%). Many questions about desirability and interest in app features did not differ by age or HIV vulnerability category. Black women ≥25 years had 7.3 times the odds of rating the inclusion of short videos as an important part of the mHealth app (OR 7.3 95% CI (1.7, 32.4)). Within this population, interest in using a sexual health app was high, suggesting an openness to app development for both sexual health as well as specifically for pre-exposure prophylaxis.

[This corrects the article DOI: 10.1371/journal.pdig.0000665.].

Gait measurements are a central component of functional assessments and risk stratification before surgery. Various sensors can measure gait metrics, but none are routinely integrated into surgical workflows because they are too challenging to implement at scale in clinical situations. In this manuscript, we report the development and validation of a rapidly-deployable, low footprint, entirely contactless sensing system, called GroundCode, that is explicitly integrated within a surgical workflow. GroundCode combines the Microsoft Kinect with seven floor-mounted single-axis accelerometers, overcoming the weaknesses of each individual sensor technology and providing both robust spatiotemporal resolution (Kinect) and high-fidelity footstep detection and quantification (floor accelerometers). We show that GroundCode-derived gait speed and cadence are highly precise measurements (>90%), and we validate them against two standard clinical gait measurements relevant to pre-surgical evaluations - stopwatch time and six-minute walk test distance. We show that GroundCode-derived gait metrics identify various surgical risk factors, like age, sex, and frailty. In addition, we show that preoperative gait is associated with postoperative quality of recovery. Importantly, we designed this system to be deployed by non-technical personnel and performed this study in a non-laboratory setting, providing proof-of-principle that GroundCode can be used in various real-world environments. We conclude that GroundCode provides highly robust gait measurements in real-world settings with possible applications spanning clinical diagnosis, risk stratification, and digital biomarker development.

Dysarthria, characterised by slurred speech, is a hallmark of many neurological disorders and brain trauma. Clinical assessment requires an audio-visual investigation by a trained healthcare expert, who evaluates criteria such as respiration, phonation, articulation, resonance, and prosody during speech. Quantitative assessment of dysarthria is challenging due to its complexity, variability, and the subjective nature of human-observation-based scoring methods. We present a novel machine-learning framework using transformers for stratifying and monitoring patient speech. Our framework integrates a wav2vec 2.0 model, pre-trained on raw speech data from healthy individuals. To reduce reliance on speaker-specific characteristics and effectively manage the intrinsic intra-class variability of dysarthric speech, we employ a contrastive learning strategy with a multi-task objective: cross-entropy loss for classifying dysarthria severity, and triplet margin loss to ensure latent embeddings are grouped by severity rather than by speaker. This Speaker-Agnostic Latent Regularisation (SALR) framework provides an objective, accessible, and cost-effective alternative to traditional assessments. On the UA-Speech dataset, SALR achieved 70.5% accuracy and 59.2% F1 using leave-one-subject-out cross-validation-a 16.5% absolute (30% relative) improvement over prior benchmarks. Explainability analysis indicates that our multi-task objective enhances the ordinal structure of the latent space, reducing dependence on speaker-specific cues and demonstrating robustness and generalisability. In conclusion, this proof-of-concept study demonstrates the potential of the SALR framework for speaker-independent dysarthria severity classification, with potential implications for broader clinical applications in automated dysarthria assessments.

Wearable technology can use gesture detection to identify smoking behavior and provide real-time feedback. Receiving notifications when smoking occurs may help increase awareness of smoking behavior to help promote change. The current study sought to examine the feasibility and preliminary efficacy of using a smartband for real-time smoking feedback as an adjunct to standard tobacco treatment in an outpatient hospital setting. We enrolled 38 adults (age M = 57.4, SD = 8.5, 63% female, race/ethnicity: 16% Hispanic, 68% White, 24% Black, 5% Multiracial) who smoked cigarettes daily (M = 17.2, SD = 10.9 cigarettes per day). All received standard tobacco treatment and participants were randomized to a control group (n = 20) or experimental group (receiving real-time smoking notifications from a smartband, n = 18) for 8 weeks. Participants wore the smartband on average for 45.6 (SD = 17.0) days out of the 56 days of treatment and 83.3% said they would recommend the smartband to others to help them quit smoking, indicating high adherence and satisfaction. Measures of smoking behavior favored the experimental group, although differences were not statistically significant. Rates of biochemically confirmed 7-day point-prevalence abstinence were 11% and 5% for the experimental and control groups, respectively. Those in the experimental group reported more percent days smoke-free (M = 12.4%, SD = 27.2% vs. control M = 6.9%, SD = 14.6%, cohen's d = .26) and had larger reductions in cigarettes smoked per day (CPD) (mean change in CPD = 10.2, SD = 12.2 vs. control mean change in CPD = 7.7, SD = 6.5, cohen's d = .26) during treatment. Findings support the feasibility of using smartband technology for smoking monitoring with adults from an outpatient hospital setting and show promise for improving cessation outcomes above and beyond standard tobacco treatment. Additional large-scale clinical trials are needed.

Mental health literacy has largely been studied via vignettes and surveys. Capturing the reality of the mental health literacy dimensions in a natural setting is an important step for moving towards a more actionable phase for mental health literacy. This study aims to identify the frequency patterns of the four mental health literacy dimensions reflected in the mental health-related tweets specific to information technology professionals. 15,782 tweets from October 2018 to October 2022 were collected from information technology-specific accounts. Content analysis, specifically a multi-class text classification approach, was used to analyze and interpret the tweets and categorize them into themes based on the mental health literacy construct. Tweets on "Knowledge and beliefs about risk factors and causes, self-treatments/interventions, and professional help available" were the most common (n = 6,179), and tweets on "ability to recognize specific disorders" (n = 196) were the least common. The ease of sharing content on X (formerly Twitter) could be leveraged to increase mental health awareness via targeted educational material on how to recognize specific disorders, seek help, and therefore improve mental health. Integrating mental health literacy information with the content being shared by well-established organizations in the information technology sector could help to enhance mental health literacy among information technology professionals.

Data for healthcare applications are typically customized for specific purposes but are often difficult to access due to high costs and privacy concerns. Rather than prepare separate datasets for individual applications, we propose a novel approach: building a general-purpose generative model applicable to virtually any type of healthcare application. This generative model encompasses a broad range of human attributes, including age, sex, anthropometric measurements, blood components, physical performance metrics, and numerous healthcare-related questionnaire responses. To achieve this goal, we integrated the results of multiple clinical studies into a unified training dataset and developed a generative model to replicate its characteristics. The model can estimate missing attribute values from known attribute values and generate synthetic datasets for various applications. Our analysis confirmed that the model captures key statistical properties of the training dataset, including univariate distributions and bivariate relationships. We demonstrate the model's practical utility through multiple real-world applications, illustrating its potential impact on predictive, preventive, and personalized medicine.