From Barbarism to Digital Twins: Is Medicine Finally Evolving?

Medicine is grounded in empirical studies and experiments, but diagnosis for a variety of ailments through history (and even today) have been based on what can be observed.

Consider the Transorbital Lobotomies (https://www.britannica.com/science/psychosurgery) (where an ice pick was used to access and cut part of the frontal lobe of the brain via the eye socket to treat schizophrenia). While we all look back on this practice and can (appropriately) deem this as barbaric, these operations were practiced because the approach was to fix what they could see was wrong.

A physician practicing a transorbital lobotomy

Looking back, the advances are undeniable. But even today while we assume clinical decisions are grounded in precision and laws of nature, they are still based on observation and retrospective learning. Let’s retrace the early steps that laid the foundation for modern medicine, and where Radiel Health fits into its future.

The Birth of Quantitative Medicine

The first quantitative methods began reshaping medicine in Britain between 1750 and 1830, when a group of “arithmetic observationalists” pioneered the use of mathematical analysis and systematic clinical measurements to reduce uncertainty in diagnosis. This had cascading effects in health. It led to the implementation of objective clinical tools, and aided in the creation of novel therapeutics such as the smallpox vaccine which lowered the disease’s mortality rate from 14% to 2% (pmc.ncbi.nlm.nih.gov).

Number of countries reporting the occurrence of endemic smallpox between 1923 and 1978, grouped by continents. (1) marks the WHO resolution on global smallpox eradication in 1959; (2) shows the start of the intensified eradication program in 1967; (3) marks the last case of smallpox in Somalia

This statistical groundwork set the stage for Edward Jenner’s landmark 1796 cowpox trials, in which none of his vaccinated subjects died, firmly establishing the power of quantitative evidence to save lives pmc.ncbi.nlm.nih.gov.

In 1991 Dr. Gordon Guyatt of McMaster University coined the term “Evidence-Based Medicine” (EBM), formally establishing the practice of using published research to guide every clinical decision.

Pillars of Evidence-Based Medicine

The Era of Precision Medicine

Today, these principles are widely adopted to improve the standard of care evolving into personalized medicine. In this reality, AI-driven digital twins and rich patient data guide treatment down to the individual level. In 2025, it holds strong potential in building in a healthy future is personalized as researchers, corporations and startups capture more data about patients and make smarter, personalized and optimized decisions on each case.

There are multiple examples of brilliant startups in the space:

• Unlearn
  • Uses digital twins—AI-powered virtual models of patients—to accelerate clinical trials by simulating control groups, reducing trial size, time, and cost.
• Tempus
  • A precision medicine company that uses AI and real-world clinical and molecular data to personalize cancer care and other disease treatments by enabling data-driven clinical decisions.
• Deep Genomics
  • Uses AI to predict how genetic mutations affect disease and drug response.

It is important to note that these fields have grown in response to a core fundamental truth:

<aside> 💡

Our current modes of treatment are based on evidence garnered through observation.

</aside>

Blind spots in Precision healthcare

Most startups obsess over precision drugs and digital tools for medication management; however, surgery and interventional medicine has been largely overlooked. Within the precision medicine market, over 640 startups raised over $12.4 billion. Surgery and interventional medicine is absent from the top funded catagories. This isn’t a case of low relevance — it’s a case of missed recognition:

• In industrialized countries, nearly half of all adverse events in hospitalized patients are related to surgical care, and at least half of the harm from surgery is considered preventable.
• While adverse drug reactions (ADRs) are estimated to cause about 100,000 deaths annually in the U.S, the mortality rate from surgical complications is 1,573 deaths per 1 million surgeries. With over 92 million surgeries performed annually in the U.S., this translates to approximately 145,000 deaths per year
• Surgery accounts for 1 of every 3 hospital stays in high-income countries

Economically, this has a huge toll on the health systems:

• Surgical complications increase hospital costs by $19,626 per patient on average while Adverse drug reactions cost approximately $2,262 per inpatient case.
• Surgical errors alone (a specific cause of surgical complications) cost U.S. hospitals $1.5 billion yearly.

Despite its immense economic cost — and more importantly, its toll on patients — surgery continues to operate with a 'one-size-fits-all' approach, especially when compared to other fields that have embraced advanced technological tools.

Clinical Areas Impacted

The lack of personalization within surgery is evident across multiple specialties where the opportunity to optimise patient outcomes has not been realized yet.

Consider the process of hemodialysis, a key procedure needed at the end stage of kidney disease. Hemodialysis is a lifesaving procedure that filters your blood a patient’s kidneys cannot, using a machine to clear out toxins and excess fluid my.clevelandclinic.org.

Process of kidney dialysis

In this process, the most reliable access is an arteriovenous (AV) fistula. Due to repetitive use of the veins in recycling filtered blood into a patient resulting in their frequent puncture, there is high risk of vein collapse. The prompts surgeons to stitch an artery to a vein, allowing it to get the artery’s strong pulsating blood flow and prevent its collapse.

An AV Fistula helps maintain patency

Following the procedure, over the next 4–6 weeks the connection between the artery and vein would thicken ontariorenalnetwork.ca.

Even though AV fistulas are the gold standard, up to 70% never mature properly, and only about 67% remain open after one year revistanefrologia.compmc.ncbi.nlm.nih.gov. This is largely due to a limited personalized approach, that cannot account for each patient’s unique vessel geometry and flow during the preoperative planning stage of surgery. Overtime, this triggers neointimal hyperplasia where an overgrowth of cells in the vessel wall blocks the access pubmed.ncbi.nlm.nih.gov. Over a span of 5 years, only 60% of vascular access points stay open and usable for dialysis (or other procedures) without needing any further procedures to fix or improve it. The rest of the patients have to go through repeat surgeries to try new artery-vein combinations, expensive interventions and may even face failure after multiple surgeries.

We believe there’s a possibility to do better. Considering these failures are a matter of mismatched fluid (blood) flow to some areas, we thought to ourselves:

What if we can personalize medicine by giving each patient their own physics engine?

Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) has long been a staple in aeronautical, mechanical and civil engineering. From Ansys (https://www.ansys.com/simulation-topics/what-is-computational-fluid-dynamics#:~:text=Computational fluid dynamics (CFD) is,our lives in endless ways.), CFD is the science of using computers to predict liquid and gas flows based on the governing equations of conversation of mass, momentum and energy. In most CFD experiments, the fluid in question is air or water through a man-made structure. However, as our blood is a fluid, we can also perform CFD through the structures present in our own bodies, our arteries.

In doing so, CFD has emerged as a transformative tool across a range of surgical and interventional disciplines. In the past, top medical labs have used CFD to:

• Optimize congenital heart surgery (https://pubmed.ncbi.nlm.nih.gov/32878458/)
• Personalize stent interventions (https://arxiv.org/abs/2205.04518)
• Enable patient-specific virtual treatment planning (https://pubmed.ncbi.nlm.nih.gov/26512019/)
• Guide dialysis access planning for AV Fistulas (https://openscholarship.wustl.edu/eng_etds/769/)

But let’s zoom in on that last point, where CFD was able to guide dialysis access planning for AV Fistulas. As previously mentioned, standard AV Fistula planning leaves up to 70% of fistulas failing to mature (just ~30% maturity) and only about 67% still open at one year. In contrast, CFD-informed optimization, like choosing ideal anastomotic angles and configurations, can cut primary failure down to 10–20%, lifting one-year primary patency into the 80–90% range!

The main barriers to modern CFD technology comes at these three main costs:

1. Too expensive. Patient-specific CFD typically requires high-performance computing resources and extensive model validation against experimental or imaging data, which drives up cost quadco.engineeringarxiv.org.
2. Too time-consuming. Even streamlined hemodynamic simulations can take hours to days to run, far longer than the pace of a typical clinical workflow pmc.ncbi.nlm.nih.govarxiv.org.
3. Requires specialized expertise. Setting up, tuning, and interpreting CFD models generally demands training in fluid mechanics and numerical methods—effectively asking surgeons to become engineers quadco.engineeringarxiv.org.

How can we make CFD more accessible to surgeons and clinicians alike, so they can harness the power of physics when making patient specific diagnosis?

That’s where we come in.

Introducing AI-Powered Clinical CFD

Radiel Health is a platform that provides AI-powered CFD simulations to test surgical and interventional strategies. Our mission? To empower every clinician with instant, patient-specific 3D modeling and hemodynamic insights. We want surgeries to be safer, smarter, and truly tailored to each individual.

But how?

We are constructing a pipeline that allows us to implement an autosegmentation pipeline across various imaging methods, such as MRI, CT and Ultrasound imaging. This would allow us to feed a patient’s anatomy into our Deep Learning models trained off previous CFD data, to predict a new patient’s CFD values. All in real-time.

Our first model, AV flow, focuses on optimizing AV fistula patency which provides metrics such as:

• Time-averaged Wall Shear Stress (WSS)
• Oscillatory Shear Index (OSI)

This would empower clinicians to monitor these metrics both post-operatively and pre-operatively and monitor:

• Stenosis
• Thrombosis
• Patency

This paves way to implement novel models across a variety of specialties, such as:

• Interventional Radiology
• Pediatric Cardiac Surgery
• Neurosurgery

Medicine has always advanced by observing, iterating, and learning—but it no longer has to stop there. With AI-powered clinical CFD, we’re no longer confined to hindsight. We can now simulate, personalize, and optimize before a single incision is made. It’s time we evolve from a system that waits for complications to one that anticipates them—and prevents them.

Connect with us by emailing hello@radielhealth.com to demo our initial models to help us build this future.