Wearables and Personalized Medicine
Adam Martin

Wearables have been on the market for years now, gradually improving and providing increasingly insightful data on user health metrics. Most wearables contain an array of sensors allowing the user to track aspects of their physical health. This includes heart rate, motion, calories burned, and some devices now support ECG and BMI measurements. This vast trove of data is valuable to consumers, as it allows for the measurement and gamification of key health metrics. But can this data also be useful for health professionals in determining a patient’s activity levels and tracing important events in their health history?

Many wearable devices, predominantly smartwatches, provide high-granularity data to the various apps that consume it. The Apple Watch Core Motion API provides accelerometer, gyroscope, pedometer, magnetometer, altitude, and other measurements at a rate of 50hz. This is in addition to the heart rate data that is sampled throughout the day. Apple also provides a Movement Disorder Manager interface for the analysis of Parkinson’s disease symptoms. FitBit and Pebble devices provide similar tracking capabilities. Beyond existing consumer smartwatches, there is hope for smart tattoos, VR displays, footwear, and fabrics. These wearables could measure a user’s electrolyte and metabolite levels in their perspiration. They could measure abnormal gaits or detect bacteria (Yetisen, 2018).

This high-fidelity data describing a wide variety of user activities could be invaluable to a healthcare professional hoping to find some insight in a patient’s condition. However, the process for extraction, transformation, and transfer of this data is unclear. With different device protocols and APIs providing information of varying quality and quantity, there is a need for a centralized, structured database for collection and analysis. Along with this, there is a potential for the application of AI on the analysis of wearable data. Raw sensor values will likely be incomprehensible to most analysts, so clustering of movement types and fuzzy logic on various parameters can allow a healthcare professional to better understand the meaning behind the data. Furthermore, this data can be used to feed into a system of “predictive preventative diagnosis”. Patients suffering from a variety of psychological or physical ailments can provide valuable data that highlights periods of symptom expression and also predicts prognosis (Piwek, 2016). When something is measured, it is easier to begin to act towards fixing it.

The artificial intelligence algorithms employed in the processing of collected data can be as diverse and complex as the systems they attempt to understand. Time series analysis for oscillating signals involving Fourier transforms. Feature extraction analysis through PCA. Noise reduction and motion clustering. These applications ignore the extra layer of abstraction, which involves the diagnosis and prediction aspects of wearable data. The field of wearables devices is growing, along with the promise of better digital representations, or ‘digital twins’, of patients. While there are still are matters to consider, including patient well-being and data privacy, the prognosis of wearables changing the healthcare industry looks good.

  Works Cited
Piwek, L. (2016). The Rise of Consumer Health Wearables: Promises and Barriers. PLOS MEDICINE.
Yetisen, A. K. (2018). Wearables in Medicine. Wiley Online Library.