Transforming Cancer Care: The Role of Tumor Boards and AI Innovation
Cancer treatment has witnessed remarkable advancements in recent years. However, there are instances where conventional therapies do not suffice, necessitating a more tailored approach. This is where a "tumor board" comes into play—a collaborative team of specialized experts including radiologists, pathologists, oncologists, surgeons, and other healthcare professionals. They amalgamate their expertise to devise the most effective treatment strategies for complex cancer cases.
Tumor boards are critical, high-stakes meetings designed to address high-risk patient scenarios. Timothy Keyes, a data scientist at Stanford Health Care and a combined MD and PhD candidate in cancer biology and biomedical informatics at Stanford University School of Medicine, highlights the significance of these meetings. He describes them as "high-stakes, high-cost meetings for high-risk patients."
During his medical training, Keyes assisted oncologists in preparing cases for presentation to the tumor board. This preparation process is exhaustive, involving the collection and synthesis of information from a multitude of sources such as electronic health records, imaging scans, and medical literature. The data must then be condensed into a comprehensive summary for the tumor board’s review.
In a groundbreaking development, a new tool from Microsoft is revolutionizing this process by alleviating the administrative burden and accelerating the workflow involved in tumor board preparation. This tool, known as the healthcare agent orchestrator, allows Stanford’s data scientists and developers to build and evaluate AI agents that streamline these tasks. This innovative solution is now accessible to others through the Agent Catalog in Azure AI Foundry.
The healthcare agent orchestrator empowers autonomous AI agents to consult various data sources and collaborate on tasks that traditionally require extensive time investment. These tasks include constructing a chronological timeline of the patient’s medical history, synthesizing current literature, consulting treatment guidelines, sourcing clinical trials, and generating reports. By leveraging clinically grounded knowledge, the AI agents deliver accurate and reliable results. Although Stanford Health Care continues to test this application in a research setting, it has yet to be deployed in real-time clinical use.
All AI agents are integrated with Microsoft 365 Copilot, enabling busy clinicians to utilize them without extensive onboarding. Clinicians can simply express their needs in natural language within familiar applications like Teams or Word. This seamless integration eliminates the need for additional applications in their workflow. Stanford Health Care is one of the institutions rigorously testing Microsoft’s healthcare agent orchestrator within the Foundry platform.
These agents overcome data fragmentation from various sources, including clinician notes, insurance staff records, nurse documentation, and diverse imaging formats like CT scans and pathology slides. Keyes acknowledges the challenges of deploying a chat model for such tasks, but emphasizes the specialized focus of the agents, which are directed by the healthcare agent orchestrator to the appropriate agent for each request. Setting up the initial agents from Azure AI Foundry Agent Catalog and deploying them into Microsoft Teams took approximately ten minutes, according to Keyes.
The data organizer component aggregates clinical notes, lab results, medications, and genomic data from diverse formats, structuring the information into a concise abstract. This abstract includes citations, allowing clinicians to verify the information or explore relevant sections in detail.
Keyes recounts an experience with fellow medical trainees when his attending physician requested a radiology report from the electronic health record. The traditional process involved numerous clicks, whereas the AI agent provided the needed information promptly. The agent’s citations were verified against the actual notes and found to be accurate.
The radiology agent utilizes advanced AI models on Azure AI Foundry to interpret radiology images, while the pathology agent analyzes whole-slide images to provide relevant pathology findings. Another agent identifies clinical trials for which the patient may be eligible.
The medical research agent employs reasoning models to search scientific papers on cancer, providing links for quick retrieval of full documents. At the end of the process, a report creation agent compiles a summary of the key components of the patient’s case for discussion at the tumor board, generating a Word document or PowerPoint presentation.
Traditionally, preparing a single patient’s case for a tumor board could take Keyes several hours. In testing, AI agents have demonstrated the potential to expedite this work tenfold. Given that Stanford Health Care operates more than a dozen tumor boards serving approximately 4,000 patients, the potential time savings are substantial.
"The agents facilitate easier, faster, and more efficient work," Keyes explains, highlighting the importance of conserving time during meetings with multiple clinicians. Time savings also benefit patients, allowing clinicians to dedicate more attention to patient care.
While many industries envision autonomous AI systems making independent decisions, Keyes emphasizes the continued importance of clinician oversight. "We want clinicians to remain in charge of patient care, ensuring they can verify the AI’s output."
In conclusion, the integration of AI in cancer care is a promising development that aims to enhance the capabilities of healthcare professionals. By reducing administrative burdens and streamlining processes, AI enables clinicians to focus more on patient care, ultimately benefiting both healthcare providers and patients. This collaboration between human expertise and technological innovation holds great promise for the future of cancer treatment.
For more Information, Refer to this article.
