Medicine has always advanced when we learned how to detect something that was previously invisible.
The stethoscope gave physicians access to internal sound. The electrocardiogram transformed electrical activity into a clinical signal. X-rays revealed anatomy without surgery. Laboratory chemistry allowed disease to be inferred from molecules circulating in blood.
Each breakthrough changed medicine not simply because a new instrument had been invented, but because an observable signal became clinically interpretable information.
We may now be approaching another such transition.
Healthcare possesses extraordinary amounts of data: genomic sequences, medical images, laboratory tests, wearable-device streams, electronic health records and increasingly sophisticated molecular measurements.
At the same time, artificial intelligence is becoming capable of finding relationships within biological information that human observers could never identify unaided. NVIDIA, for example, is building AI infrastructure for healthcare and biological discovery, including platforms for AI-driven biology and drug development.
But AI creates an interesting question.
What if some of the most valuable biological data have not yet entered the data set?
AI can only discover patterns in variables that are measured. That makes the future of medical AI partly a computational problem – but equally a measurement problem.
The next revolution in medical AI may depend as much on new sensors as on new algorithms. The question is no longer merely how much healthcare data we possess. It is whether we are measuring the right things.
The Search for New Biological Signals
Living organisms generate numerous physical signals. Electrical activity drives the heart and nervous system. Cells maintain membrane potentials. Metabolic reactions generate heat. Molecular reactions produce optical emissions. Ionic gradients, redox reactions and mitochondrial respiration all participate in the energetic organization of living systems.
Some of these phenomena have mature medical instrumentation around them. Others do not. Before machine learning can reveal whether a physiological signal predicts disease, recovery, resilience, drug response or biological aging, the signal must first be captured reproducibly.
That is where platforms such as Aurorus become interesting.
Aurorus is being developed around a different proposition from the conventional single-biomarker model: capture a reproducible electrophysiological/photonic response from the body, extract quantitative features from those measurements, and then investigate how those patterns correspond to physiology, disease states and longitudinal change.
The important scientific question is not whether a theory about that signal sounds compelling. It is whether the measurement contains useful information. That question can be tested.
From Measurement to Machine Intelligence
This is where companies such as NVIDIA and investors such as Foresite Capital become especially relevant. Foresite describes its investment focus around the intersection of biology, artificial intelligence and engineering.
Those three disciplines describe precisely the infrastructure required to evaluate a novel biological signal: biology establishes what the signal may represent; engineering determines whether it can be captured reliably; and artificial intelligence determines whether patterns within the data have predictive meaning.
Instead of beginning with a predetermined biological interpretation and attempting to confirm it, researchers can begin with large quantities of longitudinal measurement data and ask: Can algorithms identify physiological states? Can patterns distinguish populations? Can changes precede conventional clinical findings? Can interventions produce measurable changes? Can measurements predict who will respond differently to treatment?
Those are empirical questions.
AI May Change What Counts as a Biomarker
Historically, medicine has preferred biomarkers that can be individually named and mechanistically explained: glucose, troponin, PSA, hemoglobin.
But machine intelligence allows a different possibility: high-dimensional physiological signatures whose predictive value arises from relationships among many variables rather than from one isolated number.
Medical imaging already operates this way. A radiologist rarely diagnoses disease from one pixel. Meaning emerges from spatial relationships among thousands or millions of measurements.
The same conceptual change may eventually occur with other biological signals. Instead of asking, ‘What single biomarker does this device measure?’ we may increasingly ask, ‘What physiological information exists within the multidimensional signal?’
The Opportunity for Aurorus
Aurorus could therefore become more than another diagnostic device. Its greatest long-term value may lie in creating a new biological data layer.
Imagine combining Aurorus measurements with blood chemistry, genomics, imaging, wearable data, clinical outcomes and treatment histories. AI could then investigate relationships no researcher would need to specify beforehand.
Some correlations would disappear. Some would prove meaningless. But others could reveal biological information that conventional measurement systems have simply never collected.
Artificial intelligence will not only make existing medical data smarter. It may make previously invisible biological signals useful.
Selected references
• NVIDIA Healthcare & Life Sciences – nvidia.com/industries/healthcare-life-sciences
• Foresite Capital – foresitecapital.com
Regulatory note: Aurorus is a non-invasive electrophysiological measurement platform. It was granted FDA De Novo classification (DEN150004) as a Class I non-invasive electrophysiological measurement device. It is not cleared for diagnostic purposes. Nothing in this article claims that any therapy, modality, or practice diagnoses, treats, cures, or prevents any disease.
End · 4 min
Nancy Rizzo Prior to developing the science and technology for Aurorus, Nancy spent 13 years at Mayo Clinic Scottsdale as a board certified radiological therapeutic physicist and cochaired the Complementary and Alternative Medicine program. Nancy is board certified as a Radiological Therapeutic Physicist by the American Board of Radiology (DABR) and holds a Ph D. in Energy Medicine with an emphasis on Biophysics from Holos University.
Follow →