Google’s AI team has developed deep-learning tools, which can help with accurate detection of breast cancer, which has metastasized or spread to lymph nodes. Two research papers published by the Google AI team have highlighted deep-learning methods to make this possible, and could help pathologists in detection of breast cancers, which spread.
Google had first showcased its deep learning–based approach called LYmph Node Assistant, or LYNA back in 2017. According to an official blog post, the AI research team at Google relied on “gigapixel-sized pathology slides of lymph nodes from breast cancer patients” in order to develop the algorithm. The latest two published papers have shown that the LYNA system was able to spot a metastatic cancer from a slide without cancer with 99 per cent accuracy in both datasets.
The blog post by the Google AI team also says that the system was able to”accurately pinpoint the location of both cancers and other suspicious regions within each slide.” In some scenarios, these regions are too small to be detected accurately even by pathologists.
“Pathologists in this study were able to more reliably detect micrometastases with LYNA”, said Google.
The LYNA system could help highlight areas of concern for doctors, pathologists, and would help improve the process of review and final diagnosis, according to Google. The idea with this technology is to help assist doctors and pathologists, rather than replace the humans. “Pathologists with LYNA assistance were more accurate than either unassisted pathologists or the LYNA algorithm itself,” notes the blog post.
Google says this proves that its algorithms will prove to be more helpful when used by people, rather than the algorithm working alone.
However, the blog post does caution that the studies have limitations, including “limited dataset sizes” and the fact that only a single lymph node slide was examined for every patient instead of the multiple slides, that are common for a complete clinical case. The post admits that further work will be needed to determine impact of LYNA in real-life patient scenarios.
The two papers published are titled “Artificial Intelligence Based Breast Cancer Nodal Metastasis Detection: Insights into the Black Box for Pathologists” and “Impact of Deep Learning Assistance on the Histopathologic Review of Lymph Nodes for Metastatic Breast Cancer.”