LSU Joins Meta’s Open Science Effort, Putting AI to Work on Healthcare Challenges 

Five projects across four LSU campuses aim to give patients clearer answers about what to expect after surgery, alert care providers about falls, improve cancer care, train doctors using virtual reality and protect babies who can’t advocate for themselves.

A baby’s breathing tube slips loose in the night, or a young doctor faces a rare, frightening procedure for the first time, or a cancer patient waits too long for treatment to begin. These are just some examples of healthcare challenges five new AI-powered projects across four LSU campuses will work to solve with $300,000 in support from Meta.

Meta invited LSU researchers to present their research and take part in its open science work. A select group of them recently joined nearly 100 scientists at the Meta Open Source AI for Science Summit at Meta headquarters in Menlo Park, California.

One System, Working Together

The funded projects bring together LSU in Baton Rouge, LSU Shreveport, LSU Health New Orleans and LSU Health Shreveport. Doctors, computer scientists and engineers are working side by side.

“No single campus can do this alone, which is the point of leveraging a university system tied to a unified research enterprise with two health sciences centers,” LSU President Wade Rousse said. “Our doctors know what their patients need. Our computer scientists know what technology can do. When they work together, Louisiana wins, and so do families across the nation and the world. This partnership with Meta recognizes what LSU can do when we leverage all the components of our brand. It’s why we’re proud to be one LSU.”

“At Meta, we believe that AI has the power to drive the next wave of medical advancements to ensure the future is for everyone, and we are proud to support LSU researchers working across Louisiana on such an impactful range of projects,” Meta Director of AI and Development Partnerships Laura McGorman said.

Learn more about the five projects:

Giving Surgeons, Patients Clearer Answers (LSU Shreveport, LSU Health Shreveport) 

Patients facing brain, spine or nerve surgery often ask: “What should I expect afterward?” Today, that answer often depends on one doctor’s experience or one hospital’s records.

Subhajit Chakrabarty, associate professor of computer science at LSU Shreveport, and co-investigator Dr. Caleb Stewart, a research surgery specialist at Dartmouth College and former neurosurgery resident at LSU Health Shreveport, are building a shared, secure registry for neurosurgery patients. It covers conditions such as trigeminal neuralgia, a severe facial pain disorder, as well as spine problems, nerve injuries, brain tumors and cavernous malformations. Open-weight AI models from Meta will be used to read doctors’ notes and study about 10,000 patient images within LSU’s secure institutional infrastructure. The team will then predict how patients are expected to recover over time.

“Behind every patient are thousands of details: scans, test results and doctors’ notes,” Chakrabarty said. “No one person can see all the patterns hiding in that information. By bringing our institutions, experts and AI together, we hope to give surgeons clearer answers based on other patients like the one in front of them.”

Watching Over People, Without Cameras (LSU Shreveport)

Falls are among the greatest dangers for older adults and hospital patients. Yet many people do not want a camera watching them in the bedroom or bathroom.

A second project led by Chakrabarty uses radio signals, which can sense a person’s movement through walls and in the dark without taking pictures. The challenge is teaching these systems what a human body looks like. That normally requires expensive motion-capture hardware while Chakrabarty’s team will use an ordinary camera, along with Meta Segment Anything Model 3D Body, to create training examples. After that, cameras are no longer needed.

“People should be able to get help after a fall without being watched on camera in their own homes,” Chakrabarty said. “Radio signals can do that, but they must be interpreted.  Open tools from Meta let us achieve affordable skeletal ground truth at a fraction of the usual cost, so privacy-protecting monitoring can reach more homes and care facilities.”

Finding What Doctors Can Fix (LSU, LSU Health New Orleans)

Head and neck cancer surgery comes with serious, life-altering risks. Patients can develop complications after reconstruction, stay in the hospital longer than they need or wait too long to start radiation. In some, the cancer returns.

Supratik Mukhopadhyay, professor of environmental sciences at LSU, is working with Dr. Sagar Kansara of LSU Health New Orleans to build AI that predicts these problems early. The models boil down to simple formulas a surgeon can work out and immediately act on. One uses just four measurements and flags 91 percent of the patients who end up staying in the hospital longer than necessary. Muse Glimmer, an open-weight model from Meta, will help digest patient notes, so medical records can be kept secure and never have to be transmitted.

"A variable can be highly predictive without being what drives the outcome, and separating those two things takes more than data.” Mukhopadhyay said. “Meta’s open-weight AI models encode what the literature knows as constraints on the causal structure, so the same reasoning can run over records that arrive as scans.”

Practice Makes for Safer Practice (LSU, LSU Health New Orleans)

Doctors used to learn on real patients. Today, simulation allows them to practice first. Dr. Peter DeBlieux, assistant dean for advanced learning and simulation at LSU Health New Orleans, and Felipe Fronchetti, assistant professor of computer science and engineering at LSU, are taking that idea further.

Their project is called COPES, which stands for communication, optimal pathway, procedural training, environment and specific coaching. It combines virtual reality, touch-based simulation and an AI coach. In the first prototype, a learner practices inserting a chest tube in a simulated combat field hospital. The coach explains the patient’s injuries, assesses the learner’s understanding and offers feedback. The learner also practices talking calmly with an awake, worried patient. Kevin Lord, associate professor of pharmacology, biochemistry and experimental therapeutics at LSU Health New Orleans, is helping to design the lessons.

“In medicine, we used to learn by practicing on patients, but now we can practice before we ever touch one,” Dr. DeBlieux said. “An AI coach in virtual reality lets a young doctor rehearse a rare, high-stress procedure again and again, hear honest feedback and learn to speak with patients. This means better care when it matters most.”

Protecting Babies Who Cannot Cry (LSU, LSU Health New Orleans)

Premature and critically ill babies often depend on breathing tubes. If a tube slips out, oxygen levels can fall rapidly. Infants who have a tracheostomy (a surgically made hole in the neck into the windpipe) but are in otherwise stable health are often sent home, where it can be hard for caregivers to watch them every second of every day. A tube in the airway also silences a baby’s cry, suppressing the most natural alarm.

Eyes-On is a camera-based monitor for cribs. Dr. Michael Dunham, professor of medicine and director of the Medical Innovation Design Studio at LSU Health New Orleans, and Mukhopadhyay are developing the technology. A nurse or parent marks the baby’s tube on a screen, and the system learns what to follow and what to ignore. It then alerts caregivers if the tube moves beyond the expected range. The system runs locally on a small processor beside the crib, so video never leaves the room. The technology draws on three open-weight AI models from Meta—Distillation with No Labels v3 (DINOv3), Segment Anything Model 3 and Muse Glimmer—to help recognize and track the tube and review potential signs of dislodgement. The team is working with Manning Children’s Hospital in New Orleans and Our Lady of the Lake Children’s Hospital in Baton Rouge on the development.

“Losing an airway is one of the events we fear most in the neonatal intensive care unit,” Dr. Dunham said. “These infants cannot tell us when something is wrong, and our goal is to recognize the problem as early as possible so caregivers can respond right away.”

“Most medical AI needs thousands of annotated patient images before it can do anything useful—we wanted to skip that step,” Mukhopadhyay said. “A clinician traces the tube once, on one baby, and the models take it from there. This is only possible because of Meta’s computer vision and open-weight models.”

Learn more about Meta’s academic partnerships here.