Imagine you’ve just been diagnosed with cancer and you’re enrolled in a clinical trial for a promising new drug. You notice that after chemotherapy you struggle with occasional fatigue. Yet, whenever you visit your doctor every couple of weeks, you feel fine. This is not only a frustrating experience for patients, but it can also compromise the safety data in a clinical trial. The good news is that we may be able to solve this problem with open-source artificial intelligence (AI) and old-fashioned scientific validation.
Fatigue is a symptom that is tough to measure. But we know that one manifestation of fatigue is visual. Rather than relying solely on subjective clinical assessments or patient self-reported data, we can computationally measure subtle changes in facial characteristics, vocal patterns, language usage, movement, and response time. Through these measurements called digital biomarkers, we can consistently and objectively quantify them and create a clearer and deeper assessment of a patient’s disease state and response to treatment. AI can analyze patient video and audio data to pinpoint these critical, subtle disease characteristics and behavioral trends to personalize care.
This future isn’t limited to cancer, either. Many diseases, from depression to Parkinson’s, need to be assessed by analyzing visual and auditory elements of their symptoms. These signs of how a person is feeling or responding to treatment may be too subtle for a clinician to catch in real-time, or not visible during a singular visit. That’s why digital biomarkers represent a paradigm shift. Unprecedented levels of consistency and automation in assessing the effectiveness of a patient’s treatment plan can help researchers drastically elevate the integrity of clinical trial data, and hopefully drive toward better outcomes for patients.
If we can address challenges to the wider adoption of digital biomarkers, we can create clinical trials that understand the nuances of how patients’ experience illness and guide us toward novel treatments. One of the key opportunities to drive adoption is creating an open-source platform to drive development of these digital measurements.
Similar to how internet security firms will challenge hackers to crack their code in an effort to make it stronger, open-source AI platforms allow the scientific community to validate and improve the accuracy of this software. We want outside researchers to use and interrogate our algorithms – as well as build and test their own – because we believe trust for novel measurements should be built in the public domain through peer review. Additionally, as we continue to embed diversity in everything we do, we value the unique perspectives and approaches to algorithm development that will best serve patients of all backgrounds and with any condition where these biomarkers could impact care.
Democratizing access to scientifically validated digital biomarker algorithms drives transparency, helping to validate digital biomarkers as a legitimate means to understanding disease and patient behavior.
Though it may seem counterintuitive for a company to reveal how their platform works, we believe that it is the best path toward the expansion and validation of these emerging approaches to measuring disease. Our hope is that the scientific community could revolutionize the detection of symptoms and how researchers draw conclusions about a drug’s impact.
