AI Outperforms Doctors at Summarizing Cancer Pathology Reports in Northwestern Study

A Northwestern Medicine study found that AI models from Meta, Google, DeepMind, DeepSeek, and Mistral AI generate more complete summaries of complex cancer pathology reports than physicians do. The research tested six different AI models on their ability to summarize cancer pathology reports, comparing their performance against summaries created by doctors. The findings suggest AI could serve as a valuable tool for improving clinical communication between healthcare providers and enhancing patient understanding of complex medical diagnoses. This represents a potential application of large language models in healthcare settings where clear communication of technical pathology findings is critical.

AI Models Outperform Physicians in Pathology Report Summarization

• ▪AI models can generate more complete summaries of complex cancer pathology reports than physicians, according to a Northwestern Medicine study

Study Methodology and Models Tested

• ▪The Northwestern Medicine study tested six AI models for summarizing cancer pathology reports
• ▪The AI models tested in the Northwestern Medicine study were developed by Meta, Google, DeepSeek and Mistral AI

Perspective of Northwestern Medicine researchers

• ▪AI-generated summaries of cancer pathology reports could improve patient understanding of their diagnoses
• ▪AI-generated summaries of cancer pathology reports could improve clinical communication between healthcare providers