How Is AI Shaping Patient-Centered, Multidisciplinary Oncology Care?

Artificial intelligence (AI) holds transformative potential for oncology, ranging from streamlining tumor board workflows and automating treatment planning to enabling real-time adaptive radiotherapy and expanding equitable access to clinical trials for diverse and underserved patient populations. As AI tools continue to mature, clinical validation, algorithmic bias, and data interoperability have become central concerns for health systems and multidisciplinary care teams alike.

CancerNetwork® spoke with Nevine Hanna, MD, MPH, FACRO, DABR, lead radiation oncologist at the Thompson Proton Therapy Center and director of the Radiation Oncology Division, about how AI can be meaningfully integrated into tumor board decision-making and precision radiation therapy planning.

Hanna began by discussing how AI can aggregate clinical data to support multidisciplinary decision-making in tumor boards without replacing human oversight. She then described how auto-contouring and adaptive radiotherapy algorithms are already transforming radiation oncology practice. Additionally, she outlined the rigorous validation standards AI tools should meet before clinical deployment.

She addressed the need to train AI on racially, ethnically, and socioeconomically diverse patient populations to prevent algorithmic bias, and described strategies for overcoming electronic health record (EHR) data silos to enable real-time clinical trial matching. Hanna also explored how AI can allow clinicians to spend more time delivering compassionate, personalized care, concluding with a forward-looking vision of the most impactful AI innovations ahead, including multimodal AI models and adaptive radiotherapy automation to predictive toxicity analytics. She ultimately affirmed that the future of oncology will remain fundamentally patient centered.

CancerNetwork: How can AI-driven clinical tools be integrated into tumor board workflows to streamline cross-specialty decision-making across medical, surgical, and radiation oncology?

Hanna: AI can serve as a decision-support tool for tumor boards, and it can be done effectively, but it cannot be the decision maker, per se. AI can automate and aggregate pathology reports, imaging findings, genomic data, prior treatments, and relevant clinical guidelines, consolidating all of that into a summary before the tumor board convenes. This will ultimately reduce the time that staff and clinicians spend collecting and gathering information, allowing specialists to focus on the clinical discussion.

From a radiation oncology perspective, AI could also help identify patients who may benefit from standard photon radiation vs proton radiation and [identify] relevant ongoing clinical trials through RTOG [Radiation Therapy Oncology Group] or other cooperative oncology groups focused on treatment modalities and their combination.

The most effective implementation is one that enhances multidisciplinary collaboration, facilitating good discussion among medical oncologists, surgical oncologists, radiation oncologists, pathologists, radiologists, and support staff, including dietitians, social workers, and palliative care teams, while maintaining physician oversight so that all final decisions are made with the clinician and the patient in mind.

From a radiation oncology perspective, how are AI applications enhancing precision in treatment planning, auto-contouring, and organ-at-risk sparing for advanced therapies?

Radiation oncology is wonderful in that we have been using AI and automation for quite some time, and these tools have been continually improving. When I attended the first United Nations governance meeting in Geneva to discuss guardrails for AI, one of the first observations made was that radiation oncologists must already know a great deal about AI in medicine, which underscores how deeply embedded AI has become in our specialty.

One of the most significant advances is auto-contouring. It rapidly segments tumors and organs at risk, reducing contouring time from hours to minutes. This improves consistency and efficiency, allowing clinicians to focus precisely on where radiation should go and what it should spare. AI also improves treatment planning by generating predictable, achievable dose distributions and optimizing beam arrangements rather [than relying on iterative manual adjustments] by physics staff.

In proton therapy, AI supports adaptive radiotherapy by accounting for anatomic changes and range uncertainties. In photon therapy, it assists in conformal planning and normal tissue sparing. Overall, AI improves efficiency and consistency in ways that help narrow the variation between community and academic practice. At the end of the day, physician review remains critical; the radiation oncologist is ultimately responsible for validating contours, confirming treatment intent, and ensuring the overall quality of the plan.

What rigorous benchmarks and clinical validation processes should care teams apply before deploying AI algorithms into daily practice for cancer management?

Before deployment, AI tools should undergo the same scientific rigor and scrutiny we expect from any other clinical technology. That means robust external validation across multiple institutions and diverse patient populations, and a requirement to demonstrate clinically meaningful improvements, not merely statistical ones. How did this tool improve the clinical endpoint and outcome, as opposed to serving solely as a data collection point?

The way I would evaluate an AI tool in oncology encompasses its accuracy, reliability, and reproducibility; workflow efficiency; patient outcomes; and, first and foremost, safety. Prospective clinical evaluation is ideal, and continuous monitoring after implementation is essential to detect performance drift.

How can oncology leadership ensure that AI tools are trained on diverse patient populations to prevent algorithmic bias and expand equitable access to high-quality care?

These algorithms can be made more equitable by training and validating them using diverse populations in terms of race, ethnicity, socioeconomic status, geographic location, age, and comorbidities. Leadership should regularly audit AI performance across these patient subgroups to identify disparities early and ensure transparency in reporting outcomes, data sets, and expected limitations.

AI has potential to expand access to expertise, particularly for underserved communities. However, that potential can only be realized if equity and inclusion are treated as core design principles from the outset rather than as secondary goals.

How can health systems overcome EHR data silos so AI tools can synthesize comprehensive patient histories for enhanced clinical trial matching and care planning?

Data fragmentation is one of the most significant barriers to effective AI in oncology. Successful systems will require interoperability; the ability for EHR systems to communicate with imaging platforms, pathology databases, genomic repositories, and clinical trial registries. What AI can then do is synthesize structured and unstructured information into a comprehensive longitudinal patient profile. Imagine a patient arriving for care and being automatically matched to a relevant clinical trial, with care coordination and treatment planning pathways already identified.

Achieving this will require strong data governance, standardized data formats, and secure information-sharing frameworks. We must address uncertainty about data security and the Health Insurance Portability and Accountability Act (HIPAA) compliance by embedding those safeguards into the infrastructure from the beginning, protecting patient privacy while enabling meaningful clinical insights.

As AI tools automate routine analytical tasks, how can multidisciplinary care teams leverage these technologies to spend more time delivering compassionate, personalized patient care without sacrificing clinical autonomy?

The greatest value of AI is not replacing the clinician but giving them more time. In a perfect system, AI reduces administrative burden, automates documentation, assists with treatment planning, and organizes clinical information, freeing physicians to spend more time discussing goals of care, educating patients, and addressing their concerns. Patients in oncology remember how we communicated with them and how we supported them during the most difficult moments. AI should enhance that human connection, not replace it.

The ideal model is AI assisted and physician led. Technology handles the repetitive work; clinicians focus on empathy, judgment, and shared decision-making. Moreover, as AI reduces delays, from pathology results to clinical documentation, patients will move through the care continuum more efficiently. That continuity means the conversation a clinician had with a patient a few days prior can be resumed at the next visit, with all relevant data synthesized and ready. AI will do nothing but assist a good, flowing conversation between clinician and patient.

Looking ahead over the next 3 to 5 years, which emerging AI innovations do you anticipate will have the most meaningful impact on collaborative, patient-centered oncology care?

First, multimodal AI models that integrate imaging, pathology, genomics, and laboratory values into a unified clinical picture. From a radiation oncology standpoint, adaptive radiotherapy will become increasingly automated; right now, a radiation oncologist must go to the machine, approve an adaptive plan, and then allow treatment to proceed. As trust in these systems is established through repeated validation, real-time adaptive radiotherapy could proceed without a manual stop-and-review step.

Second, AI-driven clinical trial matching would be amazing [with] a system that identifies the optimal trial for a given patient in real time, surfacing that recommendation automatically rather than requiring clinicians to manually search through extensive eligibility criteria.

Third, predictive analytics [to] identify toxicity risks and personalize treatment intensity before therapy even begins based on the patient’s imaging, laboratory values, history, genetics, family history, and socioeconomic status. Despite all these advances, the future of oncology will remain fundamentally patient-centered. With data that are readily available, multidisciplinary care can become more personalized, more equitable, and more effective; and the clinician-patient conversation can be continuous and real time, building from visit to visit, supported by AI rather than interrupted by waiting for information.