Peter Chang, MD, Translates AI to Advance Healthcare

As Peter Chang, MD, steps into his new role as director of the UC Irvine Institute for Precision Health (IPH), he leans into his experience as both a radiologist and software engineer.

IPH takes a multidisciplinary approach to providing more personalized care by leveraging advances in engineering and computer science. “Success at the intersection of medicine and technology requires more than just having accomplished doctors and brilliant engineers in the same room,” says Chang, an associate professor of radiological sciences in the UC Irvine School of Medicine and a radiologist for UCI Health. “In my experience, having at least one individual who speaks both languages and can help bridge the chasm between these two widely divergent domains is critical to accelerating discovery and innovation.”

As co-founder of a medical AI startup, Chang is one such individual. His academic background and understanding of clinical needs, combined with his experience as a tech entrepreneur, allow him to translate AI research into real-world care.

The Intersection of Medicine and Technology

Chang always had an interest in software development and programming, but he never envisioned becoming a tech entrepreneur. “I’m trained as a physician, and I went to school with the idea of simply becoming a doctor,” he says. “For many years, I was just a radiology resident who enjoyed computer science, math and AI.”

Then, during the last year of his residency, he decided to enter the Brain Tumor Segmentation (BraTS) challenge, a competition that involved analyzing brain tumors. “I ended up winning, and as part of the experience, I realized that I didn’t need to be an expert radiologist or developer to be successful,” says Chang. “Simply knowing a little bit of both goes a long way.”

After winning the contest in 2016, Chang went on to launch Avicenna.AI in 2018 with Cyril Di Grandi, a software and medical device expert. The startup offers radiology tools to improve diagnostic accuracy and workflow efficiency. “At that point, I was in Silicon Valley and about to leave medicine altogether,” says Chang, “but I had some contacts here at UC Irvine, and the stars aligned.”

Chang joined the Department of Radiological Sciences that same year. He immediately helped launch the Center for AI and Diagnostic Medicine, the precursor to IPH, as well as the Center for Applied Artificial Intelligence Research (A²IR). The latter, where Chang has served as director since its inception, is a cross-disciplinary initiative focusing on deploying AI healthcare applications.

“We’ve been a little bit ahead of the curve all along, because we’ve had this vision that if we invest early, we can stay at the cutting edge,” says Chang, who is excited to now serve as director of both A2IR and IPH. “We have a lot of great researchers on campus building cool and interesting AI tools, and we have a lot of expert clinicians who want these tools, so a large part of the IPH moving forward will be bridging the gap to actually get those tools into practice.”

Scaling Success

Chang explains that a useful but niche tool — such as an algorithm that helps identify a rare brain tumor — usually lacks the financial incentives and resources needed for practical deployment.

“It is easy to stop after your research is published in the scientific literature, but it is bringing a tool to market that, at the end of the day, is what will help our patients — that’s the piece that is sometimes missing,” says Chang. “Quite often, we are stuck in a place where we all clearly see the need for a tool but don’t know how to get it into practice.” That’s where he has had success in the Department of Radiological Sciences.

“In the past few years, we’ve worked hard to develop a translational pipeline on a smaller scale, which has resulted in many successful deployments within our department,” he says. His knowledge of the governance structure and how to overcome resource restraints is key. “Uniting all the key department decision-makers and resources under one entity, we’ve been able to efficiently address administrative hurdles to take a tool from conception to practice in a very short period of time.”

His vision for IPH is to emulate this success on a larger scale in a hospital setting. For example, in recent months, he and his team have built an in-house large language model (LLM) platform approved for safe and secure analysis of protected health information (PHI) data in the hospital.

Since October 2025, a custom LLM that Chang developed (based on the open-source GPT-OSS-120B OpenAI model) has been used to systematically analyze ICU patient notes to facilitate clinical workflow and optimize care delivery. “UCI Health has been very open and transparent with security and regulatory requirements,” says Chang. “Working within those constraints, I am able to leverage my knowledge of clinical informatics, IT and data science to identify an AI solution that checks off all the boxes.”

Building AI Literacy

Chang also leads the healthcare AI curriculum at UC Irvine for both medical students as well as for computer science undergraduate and graduate students. “Education really is needed across the board, from simply understanding the security risks of PHI data in commercial platforms like ChatGPT, to more sophisticated things, such as evaluating for model bias or drift, which may manifest as AI tools behaving in unusual or unexpected ways over time,” he says.

He is also hosting AI literacy workshops for faculty, students, residents and clinicians, building off the didactic series offered within the Department of Radiological Sciences.

“The field moves so quickly that we’re now seeing applications in AI that touch almost every department, from data analysis to note-taking,” says Chang. “Within five or 10 years, AI will end up touching almost every aspect of the medical decision-making process for physicians.”

His biggest concern is widespread use of AI by the public for health-related issues without direct medical supervision, where people might not fully understand the limitations of the technology. “OpenAI just released a health module, and patients can now upload a lot of their own data,” says Chang. “As I explained in a recent Scientific American article, if patients just instinctively take everything at face value, you could certainly run into problems. Ultimately, all important decisions should be made in consultation with a qualified expert physician.”

Pushing the Envelope

As Orange County’s premier academic health system, UC Irvine has an advantage when it comes to exploring AI. “You can’t really push the envelope unless you’re in an academic medical setting,” says Chang, stressing how UC Irvine offers the perfect blend of state-of-the-art research and AI expertise with access to clinicians and real-life patient data.

“It is such a luxury to be able to tinker with an algorithm and then literally the next day, test to see if that new code works and get immediate feedback from my colleagues,” he says. “If you’re at a startup or other commercial entity, that iterative cycle requires months, if not years. Academic medical centers like UC Irvine are truly unique institutions where you have the intersection of all the expertise and resources needed to really advance healthcare.”

— Shani Murray