A recent study conducted by researchers at Memorial Sloan Kettering Cancer Center demonstrated the role artificial intelligence can play in improving the accuracy and efficiency of pathologists when detecting lymph node metastases in breast cancer.
The research, published in The American Journal of Surgical Pathology, involved training an artificial intelligence (AI) algorithm on over 32,000 whole slide images (WSIs) matched with pathology reports from more than 8,000 patients.
The average reading time for pathologists dropped from 129 seconds per slide during the unassisted phase to 58 seconds with AI assistance, demonstrating a 55% gain in efficiency.
Additionally, two of the three pathologists also experienced improvements in sensitivity when detecting metastases, with sensitivity rising from an average of 74.5% when unassisted to 93.5% when employing the AI platform.
The study involved 3 experienced pathologists who each reviewed a dataset of 167 breast sentinel lymph node WSIs—69 of which contained varying sizes of cancer metastases and 98 of which were benign. Each pathologist assessed the slides twice: once without AI assistance and a second time with the assistance of a novel AI system known as Paige Breast Lymph Node (Paige BLN) following a 3-week break.
The Paige BLN system was designed to aid pathologists in identifying regions within the slides that are suspicious for harboring metastases. It operates on a deep learning algorithm that analyzes tissue images, providing a binary classification of slides and highlighting areas with the highest probability of cancer presence.
Previous studies have indicated that pathologists' performance could be enhanced with AI, particularly in distinguishing between isolated tumor cells and micrometastases.1 The current study adds to this body of evidence, indicating that AI may hold the potential to expedite the diagnostic process and improve diagnostic accuracy—especially critical given that incorrect staging can lead to inappropriate treatment plans.
The study's authors suggested that further research is needed to explore how such tools can be effectively implemented in clinical settings, ultimately improving treatment selection and patient care.
The study was funded by Paige.AI and several of the authors have business connections to the company.
References
- Steiner DF, MacDonald R, Liu Y, et al. Impact of deep learning assistance on the histopathologic review of lymph nodes for metastatic breast cancer. Am J Surg Pathol. 2018;42:1636.