The 2025 death of 23-year-old Patrick O'Donnell from an aggressive brain cancer is pushing healthcare leaders to examine how artificial intelligence can build more individualized treatment plans for complex diseases. At the U.S. News Healthcare of Tomorrow conference in Washington June 17, his mother and the medical team that treated him described how his three-year fight with glioblastoma multiforme is shaping decisions about where AI can supplement clinical judgment - not just for paperwork, but for the hard moments when standard options fall short.
A case that reframes the AI conversation
Anne O'Donnell remembered the shock of the diagnosis in August 2022. Her son, then a 20-year-old University of Utah hockey goaltender, had a seizure after practice. A biopsy at UC San Diego revealed glioblastoma multiforme. "My 20-year-old was sitting there with every aspiration going out the window," she said. "I had Googled glioblastoma. I had no idea what it was, and so we left stunned."
Patrick O'Donnell died on Aug. 23, 2025, at age 23. Before his death, he agreed to genetic sequencing and research, hoping others might benefit. His openness created the foundation for a panel discussion at the conference that included his neurosurgeon, Dr. Alexander Khalessi, chair of neurosurgery at UC San Diego Health and the system's chief innovation officer; Philip Rackliffe, president and CEO of GE HealthCare Advanced Imaging Solutions; and Sumita Singh, executive vice president and general manager of health at U.S. News.
Singh described the panel's purpose: "There is another person on this panel who you can't see, and that's Paddy. This is a story of Paddy and how that drove Dr. Khalessi to think about … what standard of care is limiting."
AI's role in personalizing treatment decisions
Much of AI's near-term use in hospitals has focused on easing clinical documentation. At the conference, panelists said the larger payoff may be in unifying fragmented medical data to create care plans that adapt to an individual's disease biology.
Khalessi described how O'Donnell's care moved through surgery, chemotherapy, and immunotherapy guided by the tumor's specific molecular markers. "That involves bringing together and reconciling a lot of different types of information," he said - advanced imaging, molecular biology, genetic sequencing, and point-of-care testing. The challenge, he added, is repeating that precision consistently, especially with a disease as fluid as glioblastoma: "Once you know the right thing to do in the care of an individual patient, how you hit that mark every time."
UC San Diego Health has been building AI into its clinical processes to make care more reliable, Khalessi said. For healthcare professionals watching these shifts, understanding the technology behind such platforms is becoming a core skill. Resources that focus on AI for Healthcare can help teams move beyond theory and into the practical integration of data-driven decision support.
Smarter imaging, faster clinical decisions
Medical imaging is another area where AI is changing the speed and precision of clinical decisions. Rackliffe pointed to multimodality image fusion as a way to give physicians a more complete view of an individual patient. By feeding data across different scanning technologies, AI tools can reduce ambiguity in interpreting results.
"The opportunity we have is actually being super prescriptive on the image that you're getting and in real time looking at that patient and knowing exactly what margin to get to in the easiest, most minimally invasive way possible so that you can potentially have the best outcome," Rackliffe said.
Such tools don't replace the clinician's judgment; they narrow the time needed to arrive at a precise treatment path. That distinction surfaced repeatedly across the discussion - technology should inform decisions, not make them.
Transparency, consent, and patient trust
Panelists stressed that AI must be deployed with clear boundaries. Patients need to know when AI is being used in their care, and physicians must retain final authority over treatment choices. Consent and openness were presented as necessary, not optional, to maintain trust.
Anne O'Donnell said her son's willingness to share his data reflected that mindset. "Our son was really into science and supporting research," she said. "And he's like, 'Let's go - if somebody else can learn something about this disease, I'm all for it.'" He knew he might not see the breakthroughs. His decision points to a fundamental expectation the panel echoed: patients will accept AI's role if it clearly belongs to a larger effort to improve outcomes, not to cut corners.
Why this matters for healthcare professionals
The discussion at the Healthcare of Tomorrow conference signals that AI in medicine is moving well beyond dictation software and into the mechanics of how care is planned, delivered, and imaged. For clinical leaders, imaging specialists, and health IT teams, the message is pragmatic: the tools arriving now demand fluency in data integration, multimodal imaging, and patient communication about AI. The case of Patrick O'Donnell underscores that behind every algorithm is a real biological problem that resists quick answers - and a family watching to see whether technology can make the next case different.
Your membership also unlocks:
