How AI Is Turning Medical Intuition Into Evidence

Now, AI provides the potential to illuminate what was once trapped in intuition, experience and institutional memory alone.

Cedars-Sinai investigators and clinicians are turning to AI as a powerful tool for identifying the patterns locked in enormous volumes of data. AI analysis is revealing the gravity behind split-second surgical gestures, a tumor’s molecular signatures and nuances in crucial communications.

This is not a story of machines mastering medicine but rather a story of the invisible becoming visible as technology begins to keep pace with the best ideas of medicine’s brightest minds.  

“What felt like a dream only a few years ago is now within reach,” said Eytan Ruppin, MD, PhD, deputy director of the Translational Research Institute and director of Integrative Data Sciences at Cedars-Sinai. “Advances in AI make it possible to extract clinically meaningful insights from pathology slides or molecular sequencing in an hour or a day—work that once took weeks, as well as significant expense.”

From Intuition to Evidence

During a single operation, surgeons make thousands of decisions. Until recently, medicine had no practical way of knowing which of these decisions mattered most to patient outcomes.

“For all the decades we’ve been doing surgeries, we’ve had no way to link what happens on the day of the surgery directly to the outcome,” said Andrew Hung, MD, a professor of Urology and Computational Biomedicine at Cedars-Sinai. “Now we can use AI to help us analyze a surgery in tiny increments to determine which decisions affect the outcomes.”

Hung’s work focuses on a persistent challenge in prostate cancer surgery: Why does one patient recover quality of life after surgery while another whose cancer was removed with equal success does not?

While prostate surgery often succeeds in removing a cancer, long-term quality-of-life outcomes—particularly sexual function—have remained a vexing challenge. Surgeons have long understood that subtle differences in technique matter, but until recently there was no practical way to isolate and analyze those differences at scale.

Using AI and computer vision, Hung’s team analyzes surgical video at the level of individual gestures, reviewing the split-second movements and instrument decisions that unfold throughout a procedure and identifying the gestures associated with better long-term outcomes.  

Earlier work required painstaking manual review, with investigators viewing footage second by second and annotating tens of thousands of individual gestures. Those analyses revealed that gentler tissue-separating motions were more likely to be associated with recovery of erectile function one year after surgery, while techniques using heat were more likely to correlate with poorer outcomes.  

What once took weeks of manual analysis can now be accomplished in minutes, across much larger datasets, accelerating not only research, but the ability to refine how surgery is taught and practiced.

The implications are profound: Surgery, long taught as tradition and passed from mentor to trainee, can now increasingly be understood as a measurable science—one that can be studied, improved and taught with new precision.

Data to Discovery: Mining Knowledge From Across a Healthcare System

Biobanks teeming with samples, patient charts rich with clinician notes, pathology slides, radiology images, patient surveys, monitoring and device data, clinical trial datasets, registries—a healthcare system is a treasure trove of information. Somewhere in this vast cache of details, numbers, values, results, symptoms, prescription combinations and observations are vital clues that can lead to new discoveries and transformational outcomes.

Graciela Gonzalez-Hernandez, PhD, vice chair for research and education in the Department of Computational Biomedicine, leads initiatives to transform disparate data into actionable intelligence.

“We have to stop and think about what these facts mean,” she said. “Data can be anything. But what we actually have is knowledge—the knowledge of physicians and the knowledge of a patient’s experience and recovery over years reflected in medical charts.”  

Gonzalez-Hernandez leads a series of National Institutes of Health-funded pilot studies examining patient perspectives on adherence and treatment tolerability across five disease areas: bladder cancer, cardiovascular disease, inflammatory bowel disease, HIV and pediatric obesity. These projects reflect a central challenge in medicine: cultivating understanding beyond whether a treatment works to grasp how patients actually experience it.

Symptoms, side effects, fears, misconceptions and other adherence barriers can be buried in clinical notes, patient surveys, online forums and conversations, making them tricky to analyze comprehensively and coherently. Increasingly, Gonzalez-Hernandez’s team is using large language models to analyze patient perspectives from less traditional sources, such as subreddits and other digital communities where people are often more candid than they are in the clinic.

Her collaborations span neurodegenerative disease, inflammatory bowel syndrome, cancer, emergency medicine, heart disease and more—underscoring how AI increasingly serves as a connective tool across the healthcare system’s most crucial clinical and research work.  

In partnership with the Cedars-Sinai Board of Governors Regenerative Medicine Institute, Gonzalez-Hernandez is applying AI tools to better understand heterogeneity in ALS. Drawing on clinical records and other data sources, investigators are working to identify distinct subtypes of the disease to better understand how it progresses over time.  

While researchers have long known ALS presents and evolves differently in each patient, answers explaining why the disease develops remain elusive. A small number of cases have known genetic origins.

“For 85% of patients with ALS, we have no clue why they get it,” said Clive Svendsen, PhD, executive director of the Regenerative Medicine Institute and the Kerry and Simone Vickar Family Foundation Distinguished Chair in Regenerative Medicine.  

AI tools can assist in detecting patterns across large datasets that might otherwise take years to unravel through traditional study methods.

Another study, with Carl Berdahl, MD, MPH, associate professor of Medicine and Emergency Medicine, examines how AI can evaluate missed opportunities for diagnosis, especially in identifying subtle clues that present earlier in a patient’s journey.  

Diagnosis is a complex process often shaped by time, pressure, incomplete information and overlapping symptoms. By learning from previous cases, AI tools can assist clinicians in identifying patterns that support earlier recognition of conditions and more effective decision-making in urgent settings.

A Future Within Reach

The astonishing power of AI is defined by how rapidly it is expanding the questions researchers can meaningfully ask and answer, Ruppin said.

Ruppin’s work uses AI to extract clinically relevant insights from pathology slides and molecular sequencing, compressing the path from observation to intervention. When AI models can identify molecular signatures and clinically relevant information directly from images, clinicians can more quickly develop precision treatment strategies.

In a study published in Cell, Ruppin and his team demonstrated that their novel AI tool, Path2Space, can predict gene expression across a tumor based on digital pathology slides alone. Using Path2Space reduced weeks of sequencing to minutes, enabling researchers to analyze tumors at a scale previously out of reach.

By identifying spatial patterns of gene activity within tumors, the tool could help uncover biomarkers tied to treatment response and patient outcomes. Researchers hope the technology will eventually expand access to precision cancer care by making advanced tumor analysis faster, more affordable and more accessible.  

In oncology and translational medicine, that compression carries profound implications, saving time, money and lives.  

“Testing that was once prohibitively expensive and time consuming could potentially be done for minimal expense in one day for each patient,” Ruppin said. “The narrowing gap between discovery and what we can bring to our patients is extraordinary.”

For him, the pace of that change is striking.  

“For the first time, I feel that decades of meandering exploration may be converging toward real clinical impact,” he said. “These are times of unprecedented possibility.”