More than 45 years ago, a young physician sat beside a 38-year-old woman in the intensive care unit, watching as she struggled to breathe before being placed on a ventilator. The physician delivered a diagnosis that would shape the rest of her career.
“I was new, a 27-year-old resident. I was not so eloquent in telling people things,” recalls Terry Heiman-Patterson. “And I had to tell her she had ALS, a fatal neurodegenerative disease.”
The interaction left a lasting impression on Heiman-Patterson, professor of neurology at the Lewis Katz School of Medicine.
Today, after decades spent caring for patients with the devastating disease, she is helping lead a new frontier in medicine—one that combines artificial intelligence and massive datasets to revolutionize ALS treatment.
“I still remember it. That was the moment I knew ALS would be my life’s work,” says Heiman-Patterson, who describes herself as a mission-driven clinician. “I just envision ALS as this rock, and I’m going to break it apart and just hammer at it until I make a difference.”
Currently, ALS—also known as Lou Gehrig’s disease in honor of the Hall of Fame New York Yankees baseball player whose diagnosis brought widespread public attention to the disease—affects more than 30,000 people in the United States. With no known cure, the average life expectancy after onset is two to five years. As ALS destroys neurons, patients lose muscle control and the ability to walk, talk, swallow and, eventually, breathe.
In her continued search for better ways to understand and treat ALS, Heiman-Patterson has turned to AI for help. In a multidisciplinary collaboration with Huanmei Wu, chair and professor in the Department of Health Services Administration and Policy and assistant dean for global engagement at the College of Public Health, Heiman-Patterson is using AI to help uncover patterns, make predictions and personalize care for ALS patients through the creation of a digital twin.
A virtual you
A digital twin in healthcare is an interactive, virtual replica of a patient. It’s created using large amounts of real-world health data such as medical history, genetics, lab results, symptoms and even information from wearable medical devices to better understand how a disease may progress or how a patient might respond to different treatments.
“The digital twin itself is not a new concept, but what we are focusing on—the digital twin for health—is still in its development stage,” Wu says. “Digital twins for health are especially valuable for complex and heterogeneous conditions, including cancer, diabetes, cardiovascular diseases like heart failure, neurodegenerative diseases such as Parkinson’s and Alzheimer’s, and other rare diseases.”
Together, Heiman-Patterson and Wu are building an AI-enabled system, with the goal of creating individualized digital representations of people living with ALS. The effort is known as the Digital Twin for Personalized Medicine (DT4PM) project.
The DT4PM project integrates open-access data from various sources, including the ALS Knowledge Portal. Advanced analytics can then simulate disease progression on a patient’s digital twin, predicting important milestones like the need for breathing or feeding support as muscle control worsens.
“ALS is rapidly progressive, and we still only have disease-directed drugs that are modestly effective, and maybe add three to four months to a patient’s lifespan,” says Heiman-Patterson. “What will digital twins provide? If we’re able to accomplish it, digital twins will enable me to model disease and improve my ability to predict needs ahead of time in a preemptive way, improving care. In addition, it will help people living with ALS to plan. All of this will lead to improved quality of life through timely interventions.”
In the case of ALS, Heiman-Patterson emphasizes the heterogeneous nature of the illness. She compares the multiple implicated causes of motor neuron damage in ALS to how we think about stomachaches and hopes digital twin technology could help treat the disease in the future.
“I always say, think about a stomachache,” she says. “You might have eaten something bad, somebody might have punched you in the stomach, you could have a stomach tumor, you could have an ulcer, you could have a gastrointestinal problem. But it all manifests as, ‘I have a stomachache.’ Similarly, with ALS, many different things could be causing damage to the motor neurons in the brain and spinal cord to damage the nervous system. It is a horrific disease.”
Because it often takes more than a year to diagnose, and the disease progresses rapidly, patients can typically only participate in one clinical trial, if any. Further, since most trials are placebo-controlled, the person living with ALS may end up in the one trial on placebo and never receive medicine. Digital twins can provide two benefits. First, they can help researchers better match patients to therapies and design more personalized clinical trials. Secondly, they can provide digital twins to replace and/or reduce the size of placebo groups, enabling more patients to receive medicines.
“Until there’s a cure, there’s technology,” says Heiman-Patterson. “The key to our ability to care for people with ALS and improve outcomes is to determine the characteristics of people that respond to a drug and then be able to select them for drug treatments. Will this work at some point? Perhaps.”
Roughly 10% of cases are hereditary, and digital twinning could help predict if a person will develop the disease long before symptoms even appear and simulate its progression.
“Can we predict when somebody is going to develop disease? Or if they’re going to develop disease? This is another way that digital twinning may be used,” she says. “The technology will also help me know how often I should see my patients. Do I need to see them every two months, every three months?”
Planning for what matters most
Being able to make predictions would make a world of difference for patients and their families, explains Heiman-Patterson.
“This disease is a family disease. It isn’t just the person living with ALS. It affects their kids, their spouses. ALS is so disruptive and horrible for families. I tell patients to do their bucket list now before they are too fatigued as the disease progresses.”
Heiman-Patterson says she always starts with asking patients about their goals.
“I can’t even tell you how often we talk with patients about things like family weddings,” she says. “A patient who was just diagnosed wants to know if he’ll be able to walk his daughter down the aisle in six months when he’s losing the ability to walk.”
Because large datasets are needed to make accurate predictions, she notes that current ALS patients are helping to pave the way for those who follow.
“These people are the most generous people you ever want to meet,” she says. “They’ll say, ‘I want to participate to help the next generation’ despite knowing they may not personally benefit. They’ll give biospecimens, let me draw their blood, take their spinal fluid, anything to help.”
About 10% of ALS cases are hereditary. Digital twinning could help predict if a person will develop the disease before symptoms appear and simulate its progression.
Powered by AI
Heiman-Patterson describes the data as being deposited into a large box that needs to be analyzed. Wu then uses her expertise.
“Using knowledge and models learned from large, population-level health data, digital twin models can simulate different care options and lifestyle changes,” says Wu. “These simulations aim to help patients and clinicians better understand possible options and have a better idea of how to improve their health condition.”
The DT4PM project integrates open-access data from multiple sources, including the ALS Knowledge Portal, a comprehensive resource for ALS researchers that centralizes thousands of clinical and molecular datasets. Advanced analytics can then simulate disease progression on a patient’s digital twin, predicting important milestones, such as the need for breathing or feeding support as a patient’s muscle control declines.
“The stars are aligned now because we have these repositories and a data portal that everybody’s putting their data in, and it’s open to anyone who wants to use it,” says Heiman-Patterson. “The whole idea is that it will be shared by scientists.” Before digital twins can be applied to real-life situations, Heiman-Patterson says more work needs to be done to test the models and validate them against real people—and the hope is that the technology is getting close.
“The vision is there,” she says. “But we’ll need to look at what we predicted to happen and ask, is that what really happened? We also need to ensure we have the computing and server space for large datasets and the funding to keep building these large models.”
More than four decades later, that first ALS diagnosis still drives Heiman-Patterson’s work.
“I’m always really positive with patients. I talk to them about hope, and what we can do,” she says. “I know they want a cure. It’s tough because there is this unspoken reality that the immense progress we’re making now isn’t going to help the person sitting in front of me at the moment. My whole goal is to make a difference, to help patients have independence along this journey and maintain their quality of life for as long as I can.”
High-tech, human care
Imagine if physicians could identify a disease years before symptoms appear or predict which treatment is most likely to work for an individual patient. Thanks to advances in artificial intelligence and advanced computing, that future is beginning to take shape.
AI and large-scale computer infrastructures are increasingly being woven into the healthcare landscape, helping clinicians analyze complex data, identify patterns and streamline administrative processes. Across Temple, researchers and clinicians are exploring how these emerging technologies can enhance patient care and safety, and they’re already seeing a measurable impact on improving outcomes.
But there’s no need to worry that you’ll be talking to a machine instead of a doctor at your next appointment. The goal is not to replace physicians, but to support them. Ultimately, these tools may be a game changer in giving healthcare providers something increasingly difficult to find in medicine: more time to spend face-to-face with patients.
Personalized diabetes dosing
For patients hospitalized with diabetes, an AI-powered tool is producing measurable outcomes—a substantial reduction in dangerously low blood sugar.
Temple Health was the first and largest health system on Epic’s EHR in the country to implement EndoTool Sub-Q, a predictive insulin-dosing platform for hospitalized patients. The technology continuously monitors blood sugar and uses advanced algorithms to personalize insulin recommendations based on each patient’s unique health data.
At Temple University Hospital’s main campus, nearly half of all inpatients have diabetes—a rate substantially higher than the national average of 12%—making effective diabetes management especially important.
Temple Health began a phased rollout of EndoTool in 2022, introducing it campus by campus before completing implementation across the health system in the fall of 2025.
Hypoglycemia, or dangerously low blood sugar, is one of the greatest risks for hospitalized diabetes patients. Since introducing EndoTool, Temple Health has documented a two to threefold reduction in hypoglycemia systemwide for those patients on the algorithm.
“This was one of our top priorities in introducing EndoTool, because hypoglycemia can be so dangerous and potentially deadly for patients,” says Benjamin Slovis, chief medical information officer at Temple Health. “This is an encouraging result, and a sign that we’re making a real difference.”
The software analyzes patient data in real time and adjusts insulin recommendations, helping clinicians make more personalized treatment decisions while supporting them with predictive insights.
“Integrating EndoTool into our systems is really about innovation translating into measurable patient outcomes,” says Slovis.
“We didn’t simply adopt a new technology. It is built into the clinical processes, training infrastructure and analytics systems necessary to achieve a tangible reduction in a dangerous complication of diabetes care.”
Researchers and clinicians across Temple are exploring how AI and large-scale computer infrastructures can enhance patient care and safety. They're already seeing a measurable impact on improving outcomes, including for conditions such as diabetes, skin cancer and dementia.
More accurate skin cancer detection
Researchers at Fox Chase Cancer Center, the College of Engineering and the Katz School have developed a new method that enhances the ability of AI models to detect and diagnose skin cancer in individuals with darker skin tones.
“The biggest issue with current AI cancer detection models is that they are more effective at detecting melanoma in lighter skin tones and often have difficulty detecting it in darker skin tones. As a result, when melanoma is detected in patients with darker skin, those patients tend to be diagnosed at later stages,” says Hayan Lee, corresponding author on the study, assistant professor in the nuclear dynamics and cancer research program, and a member of the Cancer Epigenetics Institute at Fox Chase.
The study, “MST-AI: Skin Color Estimation in Skin Cancer Datasets,” was published in the Journal of Imaging, a leading peer-reviewed academic research journal.
According to the researchers, existing AI models are not as effective at detecting melanoma in dark skin because of the kinds of data used to train them. This data often comes from a limited number of locations and time periods and is frequently concentrated in a single country, failing to represent the full diversity of patients. As a result, detection methods can become biased, causing the AI tool to diagnose skin cancer more accurately in people with lighter skin tones than in people of color.
“There’s this desire to have one big model, hoping that it can work for every skin type. I think this approach may be too general,” says Lee. “It’s important to understand and lessen the errors related to detection and skin types to create fair and accurate detection tools for everyone.”
By making sure AI has data on a wider range of skin tones, this research aims to close the gap in skin cancer detection and provide earlier, more accurate diagnoses for everyone beyond one-size-fits-all solutions
Advancing dementia treatment
What if we could better anticipate how Alzheimer’s evolves and guide more personalized care decisions before conditions worsen?
Xinghua Mindy Shi, associate professor of computer and information sciences at the Institute for Genomics and
Evolutionary Medicine in the College of Science and Technology, is tackling this challenge. The bottleneck for her research? Computing power.
That’s about to change, thanks to Pennsylvania’s new Keystone AI + Quantum Factory. The statewide consortium—which Temple helped launch as a founding member—brings together universities, industry partners and public-sector leaders to expand access to advanced computing infrastructure for large-scale AI and quantum research.
“Some of the large-scale AI training and simulations that previously have taken us weeks can now potentially be scaled down to a couple of days or even several hours,” says Shi.
Shi’s research, “Trustworthy Agentic AI for Personalized Dementia Management,” focuses on a central challenge in dementia and Alzheimer’s disease care: The disease develops gradually, producing incomplete, inconsistent and changing signals that make it difficult to anticipate a patient’s condition. Her work aims to develop an agentic AI system capable of continuously modeling a patient’s condition over time, accounting for uncertainty and helping guide personalized care decisions.
“The Keystone AI + Quantum Factory is creating the kind of shared infrastructure that no single institution could build on its own,” says Josh Gladden, Temple’s vice president for research. “Mindy’s research demonstrates how access to advanced computing can help accelerate work on urgent challenges by supporting the scale of computing that today’s research demands.”
Terry Heiman-Patterson and Huanmei Wu have embarked on the Digital Twin for Personalized Medicine (DT4PM) project, building an AI-enabled system to create individualized digital representations of people living with ALS.
Explore More in This Issue
Related News
Next-generation computing powers dementia-focused AI research
As a founding member of Pennsylvania’s Keystone AI + Quantum Factory, Temple is helping advance AI and quantum research while preparing the workforce of the future.