A Google Tool Is Better Than Pathologists At Detecting Advanced Breast Cancer

The tool also makes it quicker and easier for doctors to accurately diagnose metastatic tumors.

Google has developed a new tool that proves AI could become instrumental in detecting cancer. The company has developed a new system, named Lymph Node Assistant (or LYNA), that can spot the features of metastasized breast cancer with  higher accuracy than practicing pathologists.

As the company outlined in its blog, the team trained the LYNA algorithm to recognize characteristics of metastatic tumors using pathology slides, meant to mimic the diversity of samples and articles seen in routine clinical practice. The resulting AI system was able to accurately distinguish the difference between cancer and non-cancer slides 99 percent of the time— even when looking for extremely small metastases that are notoriously hard for doctors to detect.

According to MIT Technology Review, studies show that human pathologists can miss these small metastases as much as 62 percent of the time, particularly when under time pressure. The algorithm has a much better rate at successfully identifying these types of tumors in a more efficient period of time.


While the technology does have a higher rate of accuracy than human pathologists, Google points out that its not meant to replace them, but complement them. LYNA will serve as a useful tool for pathologists, making it quicker and easier to diagnose metastatic tumors. The algorithm can greatly decrease the average time it takes to check a slide, cutting it to down to as low one minute per slide in one study.

Google is still determining how to best apply the technology to real-life health care. Still, the system, if implemented, could have important implications for the practice of cancer detection in clinics across the globe.

As Google's AI blog states of the project, "Further work will be needed to assess the impact of LYNA on real clinical workflows and patient outcomes. However, we remain optimistic that carefully validated deep learning technologies and well-designed clinical tools can help improve both the accuracy and availability of pathologic diagnosis around the world."

Cover image via  Tomas K / Shutterstock.


Subscribe to our newsletter and get the latest news and exclusive updates.