- Cardiovascular disease is the leading cause of death in the U.S., with the CDC attributing 796,494 deaths in 2014, with associated costs in the tens of billions of dollars. But recent developments in monitoring, combined with novel algorithms to assess the nonlinear structure of the cardiac electrical signal, may bode fundamental change in disease detection, monitoring, and treatment, leading to reduced mortality and cost.
- Observers have long noted the irregular temporal structure of the heart rate (HR), but typically treated it as extraneous noise. In the early 1980s researchers began to decompose the HR signal into its respective frequency components, corresponding notably to autonomic nervous control (sympathetic vis-à-vis parasympathetic), modulated by humoral and other mechanisms. Each branch has its own proprietary wavelength, with responses at different spots on the frequency spectrum.
- Clinical interest grew as research showed overall signal variability associated with physiological health; conversely ECG signal structure became less variable, or more periodic, as the generating system became senescent, sluggish, or sick. Initially it was thought that heart rate variability (HRV) aligned to classic “chaos theory” methods derived from physics, in which a single scale-invariant aperiodic structure could be quantified and generalized.
- But researchers realized that due to nonstationarity (baseline wander), signal noise (ectopic spasms and/or external [i.e. electrical] interference), and finite signal length, claims of finding a universal “chaotic structure” in the ECG signal must be tempered. Nonetheless, the field was established, with currently 1,500 peer-reviewed HRV papers published annually, with academic centers for the study of non-linear properties of the physiologic signal (e.g., The Margret and H.A. Rey Institute for Nonlinear Dynamics in Medicine at Harvard Med School).
- A challenge is that continued success has created a “balkanization” of sorts – with the abandonment of the idea of one aperiodicity measure, competing ones have emerged. Currently, published literature using HRV analysis might present any number of variability metrics, often with minimal justification other than ease of calculation, showing how “variability'’ might associate with a clinical condition like obesity, sleep apnea, diabetes, or hypertension. But there’s no guarantee that any HRV measure relates to another. This has paralyzed the introduction of HRV–associated tools into the clinical and/or perioperative setting, and discouraged the active collection of fine-grained waveform data needed to validate them. Perioperative ECG data is currently smoothed into a 30-second moving average and entered into patient records at 1 minute intervals.
- Institutional stasis is countered by widespread activity in the personal electronics market. But there, the “balkanization” continues – various consumer-level HRV monitoring products use variability metrics that are idiosyncratic, arbitrary, and unvalidated.
- How to advance science, and integrate signal variability into patient care, from personalized telemedicine all the way to acute-care settings? What’s the most comprehensive, objective, clinically relevant assessment of variability in cardiac electrical activity? How to leverage it to recruit a critical mass of “early adopters” with an agreed-upon, reproducible metric of variability? And what machine-learning program might entrain sophistication and specificity, as the physiological data set grows in breadth and longitudinal depth? We believe that geometric, or dynamical analysis, may yield further insights into the structural properties of ECG signals, and pave the way for new applications. The goal is to provide low-dimensional extraction, via principal component analysis, to create real-time feedback loops assisting each participant, thus creating a large, open, scalable database for more accurate risk stratification.
How do we score?
Average Judge Score
13 judges have scored this entry. The highest possible score is 5 .3.4
We divide the average judge score with the maximum score possible and add 1 to create a multiplier.1.7
We apply the judge multiplier to the total number of crowd votes.352.2
Easy one-click social registrationIs this safe?
We only receive the minimum information necessary to verify your account. We never get access to your friends/contacts or your profile, and we never post on your behalf. Your social account is used for logging in only.ORRegister via email
Send me updates on this challenge
In order to ensure a fair voting process and to make sure that no one votes more than once, we ask that you register either with a social networking account (easiest, only requires one click) or by registering with your email address (this will require you to click on a verification email that we will send you).
You only need to register once.