2024 · Health information science and systems · paywalled
A new multivariate blood glucose prediction method with hybrid feature clustering and online transfer learning
You et al.
The finding, in our words
A multivariate blood glucose prediction model combining mixed feature clustering and online transfer learning decreased root mean square prediction error by 4.2% at 30 minutes and 5.9% at 60 minutes compared to traditional methods. Evaluated across datasets including OhioT1DM, the approach also reduced model training time by around 5% while demonstrating improved predictive accuracy across horizons up to 90 minutes.
A paraphrase to the Library’s standard, never the abstract. The source is one link away and is always the authority.
Wearable-derived digital biomarkers have been accepted by regulatory agencies as endpoints in clinical trials for Duchenne muscular dystrophy, showing their potential to support decentralised diagnostics through patient-centric monitoring. This matters because it enables more sensitive and continuous assessment of disease progression outside clinical settings.
Ma et al., Neurology and therapy (paywalled) · source ↗
Wearable biochemical monitoring requires a complete measurement chain that integrates quality-controlled biofluid sampling, sensor fusion and metadata capture to convert raw sensor outputs into clinically valid digital biomarkers.
This narrative review found that metabolomic markers enhance prediction of cardiometabolic risk beyond classic indicators, and that personalised nutrition plans based on microbiome and clinical characteristics improve glycaemic control measured by continuous glucose monitoring, HbA1c and triglycerides more than a standard Mediterranean diet. The integration of wearable-derived digital biomarkers with multiomic data could support patient-centric clinical applications for cardiometabolic disease management.
Çelik et al., Journal of translational medicine · source ↗
Over-the-counter continuous glucose monitors demonstrated reasonable agreement with overall capillary blood glucose trends in individuals without diabetes, yielding an overall mean absolute relative difference of 12.9%. However, sensor accuracy decreased following carbohydrate consumption, suggesting caution is required when using these devices for diagnostic cut-offs or exact-value monitoring.
Prakash et al., Journal of diabetes science and technology (paywalled) · source ↗
AI-driven speech analysis shows high diagnostic performance for distinguishing Alzheimer's disease from healthy controls but is less reliable for mild cognitive impairment; clinical translation is limited by methodological bias and lack of external validation, indicating a need for large-scale, multicenter, longitudinal studies.
Mairesse et al., Journal of voice : official journal of the Voice Foundation (paywalled) · source ↗