Data-Driven Health Insights Using Wearable Technology

Google just released a foundation model that learns directly from incomplete wearable sensor data without any imputation. Wearable health data is notoriously fragmented. Existing AI models typically rely on imputation or discard incomplete samples altogether. But that’s not scalable to real-world, day-long multimodal sensor streams. 𝗟𝗦𝗠-𝟮 𝘄𝗶𝘁𝗵 𝗔𝗜𝗠 𝗶𝘀 𝗮 𝗳𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹 𝘁𝗼 𝗹𝗲𝗮𝗿𝗻 𝗿𝗼𝗯𝘂𝘀𝘁 𝗲𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝘀 𝗱𝗶𝗿𝗲𝗰𝘁𝗹𝘆 𝗳𝗿𝗼𝗺 𝗶𝗻𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲 𝘄𝗲𝗮𝗿𝗮𝗯𝗹𝗲 𝗱𝗮𝘁𝗮 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗶𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻.  1. Pretrained on 40M hours of day-long multimodal sensor data from 60,000+ people; none of which had 100% data completeness.  2. Introduced a dual-masking approach (Adaptive + Inherited) to model real-world missingness using learnable tokens instead of filling gaps.  3. Outperformed prior foundation models across 10 downstream tasks spanning classification (hypertension, anxiety), regression (age, BMI), and generative imputation.  4. Preserved physiological signal importance (e.g., removing nighttime signals dropped hypertension F1 by 5%), while removing daytime signals had near-zero effect.  5. Maintained performance even when key sensors were removed, showing 73% smaller degradation and +15% higher accuracy under missingness compared to LSM-1. It's cool that the authors position AIM as a 𝗿𝗲𝗽𝗿𝗲𝘀𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻 𝗹𝗲𝗮𝗿𝗻𝗲𝗿 𝗳𝗼𝗿 𝗵𝗲𝗮𝘃𝗶𝗹𝘆‐𝗺𝗶𝘀𝘀𝗶𝗻𝗴 𝘄𝗲𝗮𝗿𝗮𝗯𝗹𝗲 𝘀𝘁𝗿𝗲𝗮𝗺𝘀, contrasting it with prior work that either (i) treats missingness in simpler tabular data or (ii) tackles irregularly‑sampled EHR events. Also, cool to see the 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗰𝗵𝗼𝗶𝗰𝗲𝘀 𝗲𝗺𝗽𝗵𝗮𝘀𝗶𝘀𝗲 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝗹𝗲𝗻𝗴𝘁𝗵 rather than sheer parameter count:  1. A 25 M‑param ViT‑1D encoder (384‑d hidden, 12 × Encoder / 4 × Decoder) ingests an entire 24‑hour day (3 744 tokens) with a 2‑D positional scheme (time × sensor).  2. Union masking lets AIM keep sequence length high while pruning only the artificial‑mask tokens—avoiding the hard 𝐷‑=‑const assumption of classic MAE. Here's the awesome paper: https://lnkd.in/gcgmRYZp Congrats to Maxwell Xu, Girish Narayanswamy, Kumar Ayush, Xin L., Daniel McDuff, and co! Free access to the best LLMs and ask any medical questions here: labela.ai/healtholymp I post my takes on the latest developments in health AI – 𝗰𝗼𝗻𝗻𝗲𝗰𝘁 𝘄𝗶𝘁𝗵 𝗺𝗲 𝘁𝗼 𝘀𝘁𝗮𝘆 𝘂𝗽𝗱𝗮𝘁𝗲𝗱! Also, check out my health AI blog here: https://lnkd.in/g3nrQFxW

The integration of multidimensional health data is likely to drive the next frontier in consumer healthtech. However, the current ecosystem (and underlying infrastructure) is highly fragmented, with a few key groups emerging: 1. Physiology-driven Applications: These tools track vital metrics like heart rate, blood pressure, and glucose levels. While they adeptly provide insights and lifestyle recommendations based on physiological trends, they often miss the critical integration of biomechanical and psychological factors, limiting their holistic impact. 2. Biomechanics-driven Applications: These systems specialize in monitoring movement patterns, muscle engagement, and posture. Their proficiency in enhancing physical activities is commendable, yet they fall short in acknowledging the intertwined roles of performance, physiological health, and mental well-being. 3. Psychology-driven Applications: With a focus on mental health, these applications delve into stress levels, mood patterns, and cognitive functions. They offer valuable insights into mental health status and behavioral trends. However, their lack of integration with physiological and biomechanical data presents an incomplete picture of overall health. __________________________________________________________________________________ The future of consumer health lies in understanding the mind-body connection; and processing biological data in a way which supports this. This holistic representation is not just a nice-to-have; it's essential for truly personalized and accurate analysis. ➡ I'm curious to hear: Have you come across any apps or platforms in consumer health that successfully merge two or three of these insight types together... beyond just a shared interface?

The future of health is Personal. Enlightening talk Michael Snyder Stanford Genetics chair, Noosheen Hashemi, January AI CEO, & Alborz Mahdavi of Protomer Technologies & Eli Lilly and Company. Mike emphasized healthcare system should focus on proactively keeping people healthy. “Medicine should be individualized based on each person's unique baseline”. Average body temp is 97.5°F, not 98.6°F. Normal range varies widely, so what's healthy for one may not be for another. January uses personal baselines to track people's health and identify problems early on. It relies on new technologies such as #wearables, #AI, & #CGM & measures various factors affecting health, #genetics, #environment, #exercise, etc. Of 5,300 enrolled in his studies, 32 ended up with a diagnosis. Noosheen: “Putting together the entire picture of someone's health instead of individual biomarkers is important.” They can detect illness from a smartwatch with 80% accuracy. People react differently to glucose. It suggests that personalized approaches to managing metabolic health are needed. Wearables have gotten smaller, cheaper, & smarter since 1971 the first digital watch, Fitbit (now part of Google) measures steps & 2017 sleep, Apple Watch in 2018 added ECG, Aktiia in 2021 measures blood pressure, and Masimo in 2023 measures H20. #CGMs take fitness to health. They are getting smaller: #Dexom, Abbott #Libre, and Medtronic Diabetes. Implantable sensors will last 900 days & can read 20 different outlets. Interesting points:  - People react differently to the same glucose type  - Resting heart rate is a better health measurement than temperature  - January AI has helped people to identify precancerous conditions, heart defects, and other health problems before they had any symptoms. Dr. Snyder discovered his Lyme disease due to a faster heart rate before any systems.  - Workplace stress increases your resting heart rate  - 9% of people are diabetic, and 33% are #prediabetes in the US - The #Diabetes #endemic is worse than the #COVID Pandemic - #Microbiome explains only 20% of our reactions to food - 70% of people with #diabetes have #depression - 22% and 90% of diabetes and prediabetes people in the US don’t know it - 20% of the US population uses wearables Noosheen shared three categories of user-generated data companies: - #Food Logging: MyFitnessPal, Noom, WeightWatchers - #Emotion Tracking: Moodkie Interactive Apple, Daylio, How We Feel - #Health + #Fitness: Strava, Nike NRC, Lifesum The goal is to help people get a more complete picture of their health using deep data, making better health decisions, and living healthier lives. An incredible #innovation presented by Alborz Mahdavi silenced the audience. its next-gen #protein can sense molecular activators in the body with a tunable activity that can be controlled. e.g., #insulins that can sense sugar levels in the blood and automatically activate as needed throughout the day! WOW!

Utilizing patient data from sources such as RPM #digitalpathways, #wearables, core personal data, fitness apps, treatment history, and general #healthbehavior can substantially optimize healthcare delivery and create significant value. 💡 Insights into patterns of disease progression and treatment effectiveness enable personalized #pathways and #careplans. 💡Market data can help identify health trends and predict patient needs, facilitating proactive interventions. 💡#Wearables and fitness apps generate real-time health data, enabling continuous monitoring, earlier detection of potential health issues, and timely interventions. 💡Core personal data and treatment history can help identify risk factors and drive preventive care. Integrating and analyzing these diverse data sources can enable a holistic view of a patient’s #healthstatus and behavior. General health behavior data can provide insights into lifestyle factors that impact health outcomes and can be used to encourage healthy habits. For example, tracking a patient's diet, physical activity, and sleep patterns can provide valuable insights into their health and allow for personalized recommendations and interventions. By leveraging comprehensive and timely patient data, healthcare providers can deliver more effective, personalized, and timely care, ultimately improving #patientoutcomes and reducing #healthcarecosts.

We Have the 🛠️ Tools. The Potential 💡 Is Clear. Let’s Rethink ❤️🩹Cardiovascular Care ❤️🩹at Scale. A compelling review by Aline Pedroso, PhD and Rohan Khera in Nature Portfolio’s Cardiovascular Health. Great outline on how AI-powered wearables, PPG/ECG sensors, point-of-care ultrasound, and edge-AI models can and are transforming cardiovascular care—extending reach, reducing friction, and bringing precision to the front lines. 👉 Article: https://lnkd.in/eCNVj8_F Why this matters: ✅Community-based detection of arrhythmias and structural heart disease is feasible now. ✅Multimodal sensor + AI fusion improves prediction, risk stratification, and monitoring. ✅Cloud and edge tech enable privacy-preserving integration into clinical workflows. ✅Tools like AI-guided echocardiograms with GE HealthCare’s Caption Guidance (FDA-cleared for use by any medical professional) allow earlier, scalable echo screenings—no sonographer required. ✅These shifts are especially powerful in under-resourced or preventive care settings. Call to action for Health Systems, Payers, MedTech and Innovators: 1️⃣ Advance interoperability—connect consumer and bedside data with clinician workflows. 2️⃣ Fund pragmatic RCTs to validate outcomes, not just signal accuracy. 3️⃣ Build reimbursement models that reward early detection and smarter triage. 4️⃣ Design inclusively—this must close gaps, not widen them. 💡 We’re past proof of concept and evolve the platform. Time to implement boldly, equitably, and at scale. #DigitalHealth #AIinHealthcare #CardiovascularCare #HealthEquity #Wearables

Wearable medical devices used across the globe are producing millions of data every second. By 2028, the Global Internet of Things (#IoT) Sensors in Healthcare market is anticipated to soar to $8.6 billion, growing at an impressive CAGR of 24.6%, up from $2.9 billion in 2023. However, managing and utilizing such large volumes of data poses a challenge. To address this challenge, the intersection of the Medical Internet of Things (#MIoT) and Federated Machine Learning (#FedML) comes into play, opening up a new paradigm in healthcare analytics. FedML enables secure collaboration among institutions for training ML models using large-scale medical datasets while maintaining patient privacy. By harnessing the power of MIoT devices and applying FedML algorithms, healthcare practitioners can extract meaningful insights, improve diagnosis accuracy, and provide more tailored treatments.  For treating chronic diseases such as diabetes and heart disease, healthcare providers can gain a comprehensive understanding of a patient's health status by collecting and analyzing data from various connected medical devices, such as glucose monitors, blood pressure monitors, and wearable fitness trackers. Imagine a world where medical devices can seamlessly communicate and collaborate, providing intelligent insights to patients and healthcare providers. With #MIoT and #FedML, this world is becoming a reality.  Ref: https://lnkd.in/eFMt6jp3 Pic credit - freepik

A refreshing and hopeful take on how AI can change the world - and us as humans - for the better from Arianna Huffington Arianna highlights the understanding that the outcome of our health is influenced by five foundational daily behaviors: sleep, food, movement, stress management and connection. That’s right: our own behaviors, on a daily basis, have the biggest impact on the health and vitality we experience. AI, combined with the continuous data from tech wearables, has the capacity to change the game when it comes to understanding each of us and what we need to do to live our healthiest lives. Oura uses AI in our product offering today, that’s why we call it a “personalized health companion.” The nature of how Oura learns you and establishes your baselines is rooted in predictive technology: the value of inference is visible every time Oura delivers an insight message or suggests that you might want to consider whether or not you are getting sick. Oura AAD (automatic activity detection) makes a personalized conclusion to reduce the work of tracking your movement and identifying your activity. Generative AI has opened up entirely new possibilities that are exciting for the future of health coaching and education. Of course, there are pitfalls like hallucinations or easily prompt-duped LLMs or trolls tricking Generative AI into delivering health advice in iambic pentameter. While I believe that AI care - properly safeguarded - is better than no care at all, we have a long way to go before these systems can give reliable health or medical advice. In the meantime, serving as a support mechanism for fostering healthy behavior is a low-risk way we can use AI today. As Arianna so eloquently shares, “Ultimately, AI is a tool. What it will be is what we will make of it. So as we are looking to architect our lives for the new year and beyond, let’s focus not just on what we want AI to be, but who we want to be.” I couldn’t agree more. #WithOura

Health devices and apps outside clinical environments are a really innovative area in healthcare right now.   From wearable fitness trackers to remote monitoring tools, patients are generating more health data than ever.   Tapping into these new streams alongside traditional EHR records holds great potential: ✅ Provide a more holistic view of a person's health over time. ✅ Enable early detection of issues based on changes. ✅ Allow physicians to intervene before acute events. ✅ Reduce utilization by addressing small declines before escalation. ✅ Personalize care plans based on individual behaviors and trends.   But it requires thoughtfully bridging siloed data sources into connected insights. Technology needs to integrate device and app data with clinical systems to paint a comprehensive portrait.   Approaching this judiciously based on use cases where it adds high value is key, along with governance to ensure proper data use. But expanding the definition of "patient data" beyond traditional settings offers a fuller understanding of each person.   I'm eager to see how organizations leverage this wealth of information to individualize care in impactful new ways. How could you apply these emerging streams in your healthcare system?   #datagovernance #digitalhealth #ehr

In today's age of digital health solutions, the concept of care is no longer confined within the four walls of a hospital room. The rise of remote patient monitoring devices is a testament to how technology is reshaping healthcare delivery, turning homes into extended healthcare spaces. Here's how it's gaining ground. 🌐 Expanding Care Boundaries - With an aging population, an increase in chronic diseases, there's an urgency to extend care beyond hospitals. Remote patient monitoring devices answer that call, providing real-time health data right from patients' homes. 📊 Personalized Care - These devices, ranging from wearable tech to blood pressure cuffs, offer insights into vital signs, medication adherence, and even movement. This real-time data ensures clinicians can craft personalized care plans, addressing health issues proactively. 🔐 Data Integration and Security - While capturing data remotely, the question arises: where does this data reside? Most of it, according to experts, is securely stored in HIPAA-compliant clouds, ensuring seamless integration with patient EHRs and facilitating informed clinical decisions. 🌍 A Wider Reach for Clinical Trials - Remote monitoring isn't just about real-time data; it's also expanding the demographic for clinical trials, making them accessible to a broader audience. 🩺 Empowering Patients - The continuous monitoring lets patients live confidently, maintaining their health awareness and staying connected to their healthcare providers. For patients with conditions like asthma, diabetes, or heart failure, this is transformative. However, as with all technology, there's a learning curve. ⚠️ Adoption & Maintenance - Proper maintenance and correct usage are pivotal for optimal data gathering. 🛡️ Privacy & Security - Data protection remains paramount, necessitating protocols for data collection, transmission, and storage. 🔄 Digital Equity - Solutions should be robust and accessible, performing effectively even in low-signal areas. The next few years will see a surge in the adoption of these tools, especially with AI paving the way for advanced patient data analytics. One thing is clear, the future of healthcare is not just in hospitals. It's in the very homes we live in. Share your thoughts! 🏥🏠🌐 #artificialintelligence #digitalhealth #medtech

80% of our health happens at home. 🏡 Telemedicine as a point solution and particularly “pay per service”, needs to do more to address the white spaces between appointments. Unless the appointment is solely for a prescription… We’ve all been there and very grateful for swift help. Data from wearables plus coaching can inform great aftercare and interventions that go beyond AI powered chat bots. It helps to detect patterns and correlates them to identify physiological changes, habits and trends that can then be discussed and improved through telehealth. This supportive loop with the right expert practitioners clinicians and coaches is what elite sport has been doing for a while. It’s a great blueprint to treat and support women as they age, in particular. #elitesport #coachingbasedcare #longevitylifestyle #womenshealth