How AI pre-preparation and real-time data synthesis are changing what tumor boards can accomplish in the same meeting time.
The multidisciplinary tumor board - a regularly scheduled meeting where oncologists, radiologists, pathologists, surgeons, radiation oncologists, and specialized nurses review complex cancer cases together - is the most comprehensive model for oncology decision-making in routine clinical practice. Cases reviewed by multidisciplinary teams have measurably better outcomes than cases managed by individual physicians: a 2019 meta-analysis in the Journal of Clinical Oncology found that multidisciplinary team review was associated with a 15% reduction in all-cause mortality for breast and colorectal cancer patients compared to non-MDT-reviewed cases.
The problem is access and bandwidth. Large academic medical centers run tumor boards multiple times per week, covering major cancer types. Community hospitals with smaller oncology programs may run a single general oncology tumor board monthly. The median case gets approximately 8 minutes of tumor board discussion time. And most health systems have no formalized mechanism to identify which patients should be presented to the tumor board - the selection process is often informal, driven by which treating physician thinks to submit a case rather than by systematic identification of the patients who would benefit most from multidisciplinary review.
These limitations mean that the tumor board, for all its clinical value, is an intermittent, unevenly distributed resource rather than the systematic quality safeguard it could be. AI-assisted tumor board preparation and execution is the pathway toward a model where every patient who should receive multidisciplinary review actually does.
The majority of tumor board discussion time should be spent on analysis and debate - generating insights that no single specialist would reach independently. In practice, a significant proportion of meeting time is spent on data presentation: showing imaging, reviewing the pathology report, reciting the clinical history. This is necessary but not high-value work. It is information transfer that happens in the meeting because it has not happened before the meeting.
Pegasi addresses this through automated tumor board case summaries. For each case submitted for tumor board review, the platform generates a structured one-page pre-meeting document that includes: a timeline of key clinical events (imaging findings, biopsy results, laboratory trends, prior treatments) formatted as a visual timeline; quantitative imaging features extracted from the most recent relevant scans; genomic panel highlights with actionable variants flagged; relevant clinical trial eligibility based on the patient's diagnosis and molecular profile; and analogous cases from the Pegasi network where similar presentations resulted in specific treatment outcomes.
This summary is distributed to all tumor board participants 24-48 hours before the meeting. The goal is for participants to arrive having already absorbed the factual context of the case, so that the meeting time can begin at the level of clinical debate rather than data presentation. In Pegasi-supported tumor board sessions at Houston Methodist, case presentation times have decreased from an average of 11 minutes to 4 minutes - freeing a mean of 7 minutes per case for substantive discussion within the same overall meeting duration.
Even well-prepared tumor boards frequently encounter the experience of needing data that is not in the prepared summary. A physician asks: "Has this patient had any prior hepatic imaging? The current CT is ambiguous about whether this hepatic lesion is new." Or: "What does our experience show for KRAS G12C patients with this histology - how many did we treat, and what were the outcomes?" These questions require either interrupting the meeting to search the EHR manually or deferring the question and reconvening.
Pegasi's real-time query capability allows tumor board coordinators to answer these questions during the meeting without disrupting the flow. The platform maintains a full structured clinical history for each presented patient, searchable by clinical event type, date range, or data element. A coordinator with a laptop at the meeting can retrieve the hepatic imaging history, display the most recent MRI alongside the current CT, and answer the question within 60 seconds. For institutional outcome queries, Pegasi's network data - de-identified case outcomes from all partner institutions - supports real-time cohort analysis: "How many KRAS G12C patients with stage III colorectal cancer and liver metastases have we treated in the past three years, and what was the median progression-free survival on first-line chemotherapy?" Answering this type of question during a meeting, rather than scheduling a separate informatics request that returns results two weeks later, qualitatively changes the quality of the decisions the tumor board can make.
Clinical trials are consistently underpresented in tumor board discussions. Survey data from the American Society of Clinical Oncology consistently shows that only 5-7% of adult cancer patients enroll in clinical trials, despite significantly better outcomes in treated populations - the "trial effect," which likely reflects both the more intensive monitoring and the experimental treatment benefit. A substantial proportion of patients who would benefit from trial participation and would qualify for available studies are never offered the option, primarily because identifying relevant trials requires a time-intensive manual search that does not fit naturally into the tumor board discussion rhythm.
Pegasi's Trial Match module, launched in mid-2025, integrates with ClinicalTrials.gov to automatically screen each tumor board case against the full registry of active, enrolling oncology studies. The screening algorithm evaluates enrollment criteria systematically - diagnosis, stage, prior treatment lines, molecular profile, performance status, and organ function parameters - and returns a ranked list of eligible studies with site proximity and enrollment status. For each eligible study, the tumor board summary includes the trial name, phase, sponsor, primary endpoint, and a contact for the principal investigator at the nearest enrolling site.
In the first six months of Trial Match deployment at participating institutions, the proportion of tumor board cases with at least one documented eligible clinical trial discussion increased from 18% to 61%. This does not mean 61% of patients enrolled in trials - trial enrollment involves patient preferences, logistical factors, and eligibility verification that go well beyond a tumor board discussion. But surfacing trial options in the meeting where treatment decisions are being made is the necessary prerequisite for any of that downstream enrollment to occur.
One of the more promising applications of AI-assisted tumor board support is enabling high-quality multidisciplinary review at institutions that lack the subspecialty depth to run comprehensive tumor boards independently. A 40-bed community hospital in rural Texas may have one general oncologist, a radiologist who covers multiple services, and no in-house pathologist. Convening a melanoma tumor board with surgical oncology, medical oncology, radiation oncology, and dermatopathology expertise is simply not possible with that staffing configuration.
Virtual tumor board networks - where cases from community sites are presented to subspecialty teams at regional academic centers via video conference - partially address this access gap, but they create new coordination burdens for both sites. Pegasi's pre-meeting summary and real-time data access capabilities reduce those coordination burdens substantially. The community oncologist submits the case through Pegasi. The platform generates the summary, assembles the imaging and pathology data in a standardized format, and distributes it to the virtual tumor board participants at the receiving academic center. The meeting itself focuses on clinical decision-making because the data assembly has already been done. The time cost of virtual tumor board participation decreases when the preparation burden decreases, making it more sustainable for subspecialists to participate in regional referral networks on a consistent basis.
Pegasi's tumor board support tools are explicitly designed as preparation and information management infrastructure, not as decision-makers. The clinical judgments made in a tumor board - whether to recommend surgery, how to sequence chemotherapy and radiation, whether a patient's functional status supports aggressive treatment, how to weigh treatment benefit against quality of life in a patient with limited life expectancy - require the synthesis of medical evidence, patient values, institutional experience, and specialist expertise that no current AI system is equipped to perform. As we discuss in our article on building trust with oncologists, the appropriate role for AI in clinical workflows is to direct physician expertise toward the decisions that genuinely require it - not to substitute for it.
The value of AI tumor board support is measured in the quality of decisions reached by the human clinicians in the room, not in any autonomous recommendation the platform generates. Better-prepared physicians with more complete data and systematic identification of clinical trial options make better decisions. The AI is the infrastructure that makes the human decision-making better. That division of labor - AI handling information assembly and synthesis, physicians handling judgment and patient care - is the version of AI-assisted oncology that we believe is both achievable today and worth building toward.