Reimagining Oncology Education With Human-Guided AI

New clinical trials, guideline updates, therapeutic approaches, and diagnostic technologies are continually reshaping how oncology clinicians learn and practice. For trainees preparing for board examinations and practicing clinicians seeking to stay current, conventional tools are not always sufficient to keep pace with the speed and volume of information. In this episode of Exploring AI in Oncology, Dr. Roupen Odabashian, Medical Oncologist at Abbotsford Regional Hospital, shares his path from hematology/oncology fellowship to building MeducationAI, an AI-enabled learning platform for cancer care professionals. His conversation with Dr. Waqas Haque explores the accelerating pace of oncology knowledge, the need for human oversight in AI-supported education, and how clinicians can develop practical AI literacy. 

From Cardiology Aspirations to Oncology and AI

Dr. Odabashian did not initially plan to enter oncology. During residency, he had been interested in cardiology, but his experiences in the cardiac care unit led him to reconsider the specialty and, more broadly, the type of work he wanted to pursue. Rotations in hematology introduced him to oncology as a field rich in data, complexity, and opportunity for meaningful patient care. He ultimately found that oncology offered both intellectual depth and a perspective on medicine that he had not previously encountered in training. “I never looked back,” he said. “Oncology has opened so many doors in my life.” 

During fellowship, his work advising venture capital groups exposed him to a growing wave of AI startups. He heard frequent discussions about AI systems, large language models, and software products, but he recognized that he lacked the technical background to evaluate the underlying technology with confidence. 

Rather than remaining at a high-level understanding, he chose to learn the fundamentals himself. He began with a private Python tutor, then continued independently with programming, mathematics, calculus, linear algebra, and AI coursework. That early foundation, he explained, made the technical language of AI more accessible and helped him understand how models, APIs, development environments, and multi-agent systems fit together.

For clinicians interested in taking a similar path, he emphasized persistence over formal technical credentials. A short period of guided learning, he suggested, can give beginners enough familiarity with core concepts to continue developing skills independently.

Building AI Tools Around How People Learn

Dr. Odabashian’s experience studying for oncology boards informed the development of MeducationAI. He saw an opportunity to build a more adaptive, interactive learning environment for oncology trainees than traditional question banks alone could provide.

The platform began with a focus on hematology/oncology fellows, pairing a question bank with tools intended to support different learning preferences. Its features include options for creating quizzes, flashcards, notes, images, and visual learning materials, as well as access to selected educational podcasts. The guiding idea is that oncology learners do not all process complex material in the same way. Some may prefer written questions and explanations, while others may benefit from visual representations, repetition, or interactive self-assessment. “Some of us like images, some of us like quizzes, some of us like flow charts,” Dr. Odabashian said. “No matter what your learning process is, we have it there.”

The platform’s scope is also expanding beyond fellows. Dr. Odabashian described efforts to support practicing oncologists, physician assistants, nurse practitioners, and other professionals involved in cancer care. One planned feature would allow users to create personalized research updates based on the topics most relevant to their practice. 

Why Human Oversight Remains Essential 

A central theme of the discussion was the role of the human in the loop. Dr. Odabashian stressed that AI is not permitted to independently determine what educational content reaches learners. Instead, AI is used to assist with tasks such as literature retrieval, content generation, and question development, while clinicians and subject-matter experts review the output.

For board-focused educational content, for example, the team considers whether a new development is sufficiently established and relevant to likely examination material. A headline announcement may be important for practicing clinicians, but it may not yet be appropriate for a board-preparation question.

The platform’s question-development workflow is designed as a multi-step process. AI tools help identify relevant literature and support question drafting, but physician reviewers validate whether the question is appropriate, whether the answer reflects current standards, and whether the distractors are educationally sound.

This review is especially important in areas where the evidence base is evolving or clinical practice varies. Dr. Odabashian noted that difficult treatment decisions—such as later-line therapy in non–small cell lung cancer or management of acute leukemias—require expert judgment that cannot be reduced to simple pattern matching. “In these topics, there is so much variability in practice,” he explained. “You want to focus on the most common practice.” 

He also highlighted a potential weakness of AI systems: their tendency to perform less reliably in less-represented clinical populations or in situations where newer evidence conflicts with more established patterns in the training data. In one example from his research, an AI system that was asked about male breast cancer continued to frame its response around a female patient. In another, it defaulted to a targeted-therapy pathway for BRAF-mutant melanoma without adequately incorporating newer evidence supporting immunotherapy in a relevant clinical context. These examples illustrate why AI-generated content requires ongoing clinical review, particularly when evidence is incomplete, contradictory, or rapidly changing.

A Different Learning Need for Every Career Stage 

Transitioning from fellowship to an attending position also changed Dr. Odabashian’s perspective on educational technology. While fellowship learning often centers on broad knowledge acquisition and exam preparation, practicing oncologists face a different challenge: identifying the most relevant new information within limited time. “As an attending, the tools that I use in learning today are very different than the tools that I used two months ago,” he said.

That realization has informed MeducationAI’s broader direction. The goal is not only to help fellows pass examinations, but also to help oncology professionals remain informed, reinforce knowledge, and apply learning tools throughout their careers. Ultimately, Dr. Odabashian described the mission in patient-centered terms: helping those who treat cancer become more effective clinicians. “I want to teach people how to better take care of cancer so they can help more people,” he said.

A Practical Path Into AI for Oncology Professionals 

For oncologists and trainees who want to learn more about AI, Dr. Odabashian recommended first identifying the goal. The appropriate learning path differs for someone who wants to understand AI well enough to use it responsibly, someone who wants to build AI-enabled workflows, and someone who wants to develop technical models.

For clinicians who primarily want to stay informed, he recommended following AI-focused newsletters and educational resources. For those interested in building or integrating tools, he encouraged learning foundational concepts in programming and AI systems. He described three broad levels of engagement:

• AI literacy: Understanding terminology, capabilities, limitations, and practical use cases.

• AI workflow development: Connecting models and tools to create useful systems, sometimes described as the “plumbing” of AI.

• Model development: Building or training more technical AI systems directly. 

Courses can provide a useful foundation, but he emphasized that hands-on experimentation is often where the learning becomes most valuable. “You learn more by building things,” he said.

AI as a Tool, Not a Replacement for Oncologists 

The conversation concluded with a broader question: will AI replace physicians? Dr. Odabashian’s view is that AI will unquestionably change oncology, but that its role is more likely to resemble previous transformative technologies than a wholesale replacement of clinicians. He compared AI to advances such as polymerase chain reaction testing and DNA sequencing—tools that have expanded the capabilities of oncologists rather than eliminated the need for clinical expertise. “Now we have another tool,” he said. “It’s called AI, and it’s very, very powerful, but it’s just like any other test.” 

He acknowledged that patients are already using tools such as ChatGPT to interpret medical information and to better understand conversations with their clinicians. Rather than discouraging this, he encourages patients to use AI as a supplemental educational resource—for example, by asking a system to explain a visit transcript in clearer language.

However, he questioned whether patients facing a cancer diagnosis will fully replace the physician relationship with an AI system, particularly when treatment decisions involve uncertainty, risk, values, emotional support, and complex tradeoffs. The future of oncology may therefore be less about AI replacing clinicians than about clinicians who can thoughtfully use AI to improve education, communication, and care. As Dr. Odabashian summarized, AI may become an exceptionally powerful tool in oncology—but human judgment and human connection will continue to matter.

Key Takeaways

• AI can help oncology professionals keep pace with rapidly expanding medical knowledge. Tools that combine question banks, visual learning, flashcards, notes, quizzes, and personalized updates may support different learning styles and career stages.

• Human oversight remains essential in AI-powered medical education. Clinicians and subject-matter experts should review AI-generated questions, explanations, and literature summaries before they are shared with learners.

• AI may be less reliable when evidence is limited, contradictory, or underrepresented. Examples such as lesser-seen subtypes and evolving treatment strategies highlight the risks of relying on pattern recognition without expert clinical review.

• Clinicians can approach AI learning at different levels. AI literacy, building AI-enabled workflows, and developing models require progressively deeper technical knowledge; hands-on experimentation can help reinforce learning.

• AI is more likely to augment oncologists than replace them. Like previous advances such as PCR and DNA sequencing, AI may expand clinicians’ capabilities while preserving the importance of human judgment, communication, and trust in cancer care.

About the Speakers 

Roupen Odabashian, MD, is a Medical Oncologist at Abbotsford Regional Hospital in British Columbia, Canada. He specializes in the treatment of patients with solid tumor malignancies, and his research focuses on the intersection of artificial intelligence and oncology. Dr. Odabashian is the founder of MeducationAI, an AI-powered education platform for oncology fellows, attendings, nurse practitioners, and physician associates.

Waqas Haque, MD, MPH, is a third-year Hematology/Oncology Fellow at the University of Chicago. He recently completed his Internal Medicine Residency at New York University (NYU) in a Clinical Investigator Track. Dr. Haque has balanced his patient care work with a variety of research projects. During his fellowship training at the University of Chicago, he plans to further his work in innovative clinical trial design, value-based care delivery to cancer patients, and clinical investigation.