Turning everyday labs into early warnings
The idea for HepaticMD began with a simple question. Could a computer learn to spot signs of liver disease using only the tests that most patients already have on file?
Osta focused on metabolic dysfunction-associated fatty liver disease (MASLD), formerly known as nonalcoholic fatty liver disease and now one of the most common causes of liver scarring. MASLD is often associated with diabetes and obesity and can progress silently over many years before symptoms appear.
What sets Oster’s models apart is not only their accuracy, but also their transparency.
“Most computers and AI models just give you an output of yes or no, fibrosis or no fibrosis,” Osta says. “We provide details and detail the data that is input into the predictive models that drive the predictions.”
Using publicly available data from the National Health and Nutrition Examination Survey, Osta built and trained a model that predicts fibrosis risk from routine laboratory values such as liver enzymes, platelets, hemoglobin A1C, and creatinine, as well as demographics such as age and gender. To test how well his model performed beyond the initial data set, he validated his results using an independent cohort that included mortality outcomes. The results held up and HepaticMD performed comparable to recommended non-invasive tools to predict fibrosis. A second independent external validation using data from the National Institute of Diabetes and Digestive and Kidney Diseases, one of the world’s largest data repositories of MASLD, is currently underway.
This clarity helps clinicians understand what is influencing a patient’s risk and, in turn, helps patients understand their outcomes. For example, someone with normal liver enzymes but a high A1C and low platelet count may still be flagged as high risk. HepaticMD reveals why.
“Behind AI models there are usually black box algorithms,” Osta says. “As clinicians, it’s important that the tools we use are transparent and explainable. This helps us make better decisions about the types of tools we integrate into our workflow.”
From code to clinic
Now, Osta is moving HepaticMD from research to real-world use. With funding and guidance from Texas Health Catalyst, Dell Med’s translational research accelerator, he is building an application that integrates with electronic medical records to automatically analyze test results and flag patients at risk for liver fibrosis.
“HepaticMD is an example of the innovation that Texas Health Catalyst seeks to support by leveraging cutting-edge AI to deliver non-invasive, cost-saving diagnostic solutions with great potential to improve health outcomes,” said Nicole Clark, MBA, Director of Programs and Partnerships. “Its ability to predict long-term mortality and reduce unnecessary liver biopsies is consistent with the program’s mission to accelerate impactful early-stage ideas into practical health products.”
Dr. Jack Virostko, associate professor of diagnostic medicine and principal investigator at Osta, says HepaticMD reflects the creativity and interdisciplinary thinking that defines clinician-driven innovation.
“As science becomes more fragmented, we need people who can bridge disciplines,” he says. “Eli has the knowledge and skills to identify clinical problems and solve them with a data-driven approach.”
For Virostko, that blend of curiosity and technical fluency represents more than a single project, it shows where medicine is headed.
“AI will change every aspect of life,” he says. “We need future leaders like Eli who can use their power to advance and improve medicine.”
Innovation rooted in Austin
HepaticMD’s evolution reflects Austin’s growing role as a hub for medical innovation. Anchored by Dell Medical School and the University of Texas Medical Center, the region is becoming a place where engineers, physicians, and data scientists come together to define the future of health.
This convergence was part of Osta’s motivation to continue training at the University of Texas. The university’s strengths in computational science and engineering, from research leadership to infrastructure like the Texas Advanced Computing Center, provided the foundation he was looking for to continue building. With dedicated research rotations within the training program, mentorship from faculty and entrepreneurs, and access to cutting-edge tools, he is currently expanding his portfolio of AI projects, including a tool that uses large-scale language models to summarize radiology reports over time.
“Diagnostic radiology stands to benefit the most from early adoption of AI in healthcare,” he says. “What I hope will be a new trend is more clinician input, clinicians themselves, driving development. We can be more than just a key stakeholder.”
For Osta, MASLD is exactly the kind of condition that requires that approach. “MASLD is important to me because it impacts a lot of underserved populations,” he says. As HepaticMD moves closer to real-world medicine, it reflects not only new tools for early detection but also a broader commitment to delivering better care to patients who need it most, articulating the University of Texas Medical Center’s mission to revolutionize the way people gain and maintain health, one question, one algorithm, and one patient at a time.
