4 min read

Inside an Individual’s Physiological Signature

Originally published in Medical Device & Technology


John Varaklis, chief strategy officer for physIQ, discusses the company's machine-learning based algorithms for the detection and early detection of physiological conditions to inform both clinical care and clinical drug development.

physIQ Inc. holds to the following theory: The human body, if its myriad biomarkers are interpreted correctly, can predict its health-related future. Such markers can portend hospitalization,1 systemic breakdowns2 and even the progress of an existing disease.3 John Varaklis, chief strategy officer for the Chicago-based data analytics firm, discusses physIQ’s theories and more. He joined the company about 18 months ago.

MDT: Please explain physIQ’s philosophy.

JV: People have always looked at parts of the human body in terms of analyzing change, be it one system or a particular part. But that one part might not represent how other parts of the body would react to the same change. When someone climbs steps, more than the person’s skeletal system is put into motion; respiratory, circulatory, nervous systems all respond as well. What physIQ’s thinking was, consider all systems when someone stands up and walks up the stairs. How does the heart respond to the number of steps taken, what is the compensation for using particular muscles? These are all connected. They can be seen in each person’s physiological signature. We each will respond differently, but the response is consistent.

MDT: So what is physIQ looking for?

JV: If you know what is normal for that person, then you know the body is experiencing something if that person’s systems change, for better or worse. Artificial intelligence (AI) and an analytics platform are excellent at detecting these subtle changes. This is how our approach differentiates from others.

MDT: But there are standards – 98.6 degrees, for example, for temperature. Why doesn’t physIQ use these standards?

JV: The temperature standard is based on general information and based on the middle of the curve. Some people have lower temperatures and are fine, while others run higher and are not ill in any way. If the standardized baseline is used, subtle indications of illness might be missed or incorrectly determine that a patient is in distress when they are perfectly fine.

MDT: Can you provide more details on your system?

JV: By combining proprietary AI and wearable biosensors, physIQ can continuously monitor patients…the platform ingests the raw waveform data from the sensors and uses deep neural net technology to derive personalized health insights and predictive analytics that can be used in a wide variety of healthcare and clinical trial applications.

For example, in a COVID vaccination study4 that just completed enrolling 130 persons, the point is to see if continuous monitoring of various biomarkers can pinpoint activation of the immune system and the ensuing systemic inflammation. The physIQ platform, via the biosensors on each patient, will collect ECG, skin temperature and other information during physical activity and sleep. A substudy will look at the connection between the recorded physiologic changes and the humoral and T-cell immunity responses fostered by the vaccine.

MDT: Has physIQ's theory and/or system’s efficacy been validated in other clinical trials?

JV: Yes. physIQ has conducted proof of concept studies in Sierra Leone with Ebola patients;6 in Rwanda2 to compare the platform’s collected biodata against nurse-obtained vitals from patients with sepsis; and in hospitalized patients with heart failure,1 the system demonstrated that it could predict re-admission. physIQ has numerous FDA clearances for its processes.5 These results have been published in peer-review journals, which we do because we appreciate the healthcare community’s scrutiny of our results.

MDT: Please explain the physIQ machine-learning based algorithm.

JV: physIQ has developed several machine-learning based algorithms for the detection and early detection of physiological conditions. We create a baseline of the patient’s biodata for 48 hours, per our FDA clearance. This baseline creates a digital print. We compare those actuals, and any changes in the analytics demonstrates a change from baseline. We are sensor agnostic to get a better understanding of how changes work. We keep growing our trove of physiological data. And we keep raw data, so we can re-run these data in the future.

MDT: What has physIQ learned?

JV: A generalized approach, as was applied in the past, is not the best way to effectively deliver health care. Having the ability to use personalized data makes the biggest difference. Think of how different each person’s response is to therapies. Some people respond well, others have unwanted reactions. We believe future treatments will be personalized, and technology like ours will be crucial to providing that care.

MDT: What does physIQ want to learn?

JV: My feeling is that ultimately, as we gather more data based on real-world use cases, this will help us understand how to move from treating a disease to preventing onset. We are very much looking at chronic diseases, such as asthma and diabetes that greatly impact lives and cost of care. This may be an ambitious view, but I feel that physIQ the potential to change medicine and we are excited to work with like-minded partners to achieve this goal.

Christine Bahls is a freelance writer for medical, clinical trials, and pharma information.


References

  1. Josef Stehlik, Carsten Schmalfuss, Biykem Bozkurt, Jose Nativi-Nicolau, Peter Wohlfahrt, Stephan Wegerich et al. Continuous Wearable Monitoring Analytics Predict Heart Failure Hospitalization. The LINK-HF Multicenter Study. Circulation: Heart Failure. 2020;13:e006513.
  2. Garbern SC, Mbanjumucyo G, Umuhoza C, et al. Validation of a wearable biosensor device for vital sign monitoring in septic emergency department patients in Rwanda. Digit Health. 2019 Sep 30;5:2055207619879349.
  3. Larimer K, Wegerich S, Splan J, Chestek D, Prendergast H, Vanden Hoek T. Personalized Analytics and a Wearable Biosensor Platform for Early Detection of COVID-19 Decompensation (DeCODe): Protocol for the Development of the COVID-19 Decompensation Index. JMIR Res Protoc 2021;10(5):e27271 (VIII): The Eight Study and Immunologic Response Sub-study. ClinicalTrials.gov Identifier: NCT05237024.
  4. Press release. physIQ Receives FDA Clearance of AI-Based Beat Detection Algorithm Building Off of Trove of Over 1.5 Million Hours of Physiological Data. August 20, 2020. https://www.physiq.com/press-media/physiq-receives-fda-clearance-of-ai-based-beat-detection-algorithm/
  5. Steinhubl SR, Feye D, Levine AC, Conkright C, Wegerich SW, Conkright G. Validation of a portable, deployable system for continuous vital sign monitoring using a multiparametric wearable sensor and personalised analytics in an Ebola treatment centre. BMJ Glob Health. 2016 Jul 5;1(1):e000070.Sierra Leone