DeepSeek R1, an open-source AI model developed by the Chinese company DeepSeek, has demonstrated superior performance in summarizing complex cancer pathology reports compared to physician-written summaries, according to a new Northwestern Medicine study published in JCO Clinical Cancer Informatics. The findings suggest that AI tools could help oncologists manage increasingly detailed patient records while reducing the risk of missing critical genetic information that influences treatment decisions .

Why Are Cancer Pathology Reports Becoming So Complex?

Modern cancer care has fundamentally changed how much information clinicians must synthesize. As biomarker testing expands and patients live longer, pathology reports have grown exponentially in length and complexity. These documents now span multiple institutions, include detailed molecular and genetic findings, and often require physicians to extract actionable insights under significant time pressure .

The challenge is real. Patients undergoing repeated biopsies and genetic sequencing can accumulate pathology reports spanning dozens of pages. Even a single missed detail, such as an overlooked genetic marker or molecular finding, can change treatment decisions and patient outcomes .

How Did Researchers Test AI Models Against Physician Summaries?

Northwestern investigators analyzed 94 de-identified pathology reports from lung cancer patients. Each report contained detailed information about histopathological findings (microscopic tumor characteristics), immunohistochemical results (protein expression testing), and molecular and genetic data relevant to treatment decisions .

The research team evaluated six open-source language models, which are AI systems that researchers can download and run locally rather than cloud-based chatbots like ChatGPT. The models tested included:

• Meta's Llama 3.0, 3.1, and 3.2 models: Open-source models developed by Meta that can be deployed on local servers
• Google's Gemma 9B: A smaller, efficient model designed for local deployment
• Mistral 7.2B: A compact open-source model optimized for performance on standard hardware
• DeepSeek R1: An open-source reasoning model from DeepSeek that emphasizes step-by-step problem-solving

A panel of oncologists assessed each AI-generated summary for accuracy, completeness, conciseness, and potential clinical risk. The results were striking: across all models tested, AI-generated summaries were consistently rated as more complete than physician-written versions. The largest differences appeared in the inclusion of molecular and genomic findings, which are often critical for determining which targeted therapies a patient should receive .

DeepSeek R1 and Meta's Llama 3.1 emerged as the strongest performers in the study, demonstrating that open-source models can match or exceed the capabilities of proprietary systems in specialized medical applications .

What Are the Practical Implications for Oncology Clinics?

The potential benefits extend beyond simply generating better summaries. If AI can reliably synthesize complex pathology reports, clinicians can review key findings more efficiently, important genetic details are less likely to be overlooked, and documentation becomes more standardized across institutions .

"As cancer care becomes increasingly complex, the burden of synthesizing complex reports is growing rapidly. What we're seeing is that AI can help ensure critical pathological and genomic details are consistently captured, not as a replacement for physicians, but as a tool to augment clinical decision-making," said Dr. Mohamed Abazeed, chair and professor of radiation oncology at Northwestern University Feinberg School of Medicine.

Dr. Mohamed Abazeed, Chair and Professor of Radiation Oncology, Northwestern University Feinberg School of Medicine

This shift could allow physicians to focus more on patient care rather than spending hours manually reviewing and synthesizing reports. For patients with complex cancers, the stakes are particularly high. Missing a key pathological finding or an actionable genetic marker could fundamentally alter the treatment strategy .

Steps to Implement AI-Assisted Report Summarization in Your Clinic

• Start with pilot testing: Begin with a small group of pathology reports and compare AI-generated summaries to physician-written ones to validate accuracy in your specific clinical context
• Choose the right model: Consider deploying open-source models like Llama 3.1 or DeepSeek R1 locally rather than relying on cloud-based services, which offers better data privacy and control
• Establish validation protocols: Before full deployment, conduct internal testing and validation studies to ensure the AI tool meets your institution's standards for accuracy and completeness
• Train your team: Educate oncologists and pathologists on how to review AI-generated summaries effectively and identify any missing information before clinical use

The Northwestern team is currently developing an app using Llama 3.1 that will allow physicians to upload pathology reports and receive AI-generated summaries for their review. However, the study authors emphasize that before deploying such tools clinically, institutions need more testing and validation studies to ensure safety and reliability .

"If AI can reliably synthesize these reports, clinicians can review key findings more efficiently, important genetic details are less likely to be overlooked and documentation becomes more standardized," explained Troy Teo, instructor of radiation oncology at Feinberg.

Troy Teo, Instructor of Radiation Oncology, Northwestern University Feinberg School of Medicine

The study, titled "Toward Automating the Summarization of Cancer Pathology Reports Using Large Language Models to Improve Clinical Usability," represents a significant step toward integrating AI into oncology workflows. Rather than replacing physician expertise, these tools are designed to enhance clinical decision-making by ensuring that critical information is consistently captured and easily accessible .

As cancer care continues to evolve and patients accumulate more complex medical records, AI-assisted summarization could become an essential part of modern oncology practice. The key takeaway is that open-source models like DeepSeek R1 are proving competitive with proprietary alternatives in specialized medical applications, offering institutions more flexibility and control over how they implement these technologies.