San Francisco, CA -- (ReleaseWire) -- 09/12/2019 --Neuromation Researchers Demonstrate New AI Technique that Dramatically Improves Reliability of Breast Cancer Diagnosis
Neuromation, the San Francisco-based developer of applied artificial intelligence solutions and cutting edge tools for AI developers, is proud to announce that a paper co-authored by two senior members of its research team, Alexander Rakhlin and Sergey Nikolenko, was accepted for publication in the proceedings of I CCV19-VRMI in Seoul, Korea in October.
ICCV, the International Conference on Computer Vision, is the premier international computer vision event globally. VRMI is a workshop at ICCV dedicated to addressing the challenges of visual recognition model development in the medical image domain.
The paper, Breast Tumor Cellularity Assessment using Deep Neural Networks, demonstrates a new technique that dramatically improves the rating reliability for diagnosis of breast cancer over existing techniques known in the literature.
According to the World Cancer Research Fund, Breast cancer is the most commonly occurring cancer in women and the second most common cancer overall. There were over 2 million new cases in 2018. In the United States, about 1 in 8 women will develop invasive breast cancer over the course of her lifetime.
According to author Sergey Nikolenko, Chief Research Officer of Neuromation:
"The only definitive way of diagnosing breast cancer is by removing a sample of breast cells for testing, known as a biopsy. In current clinical practice, tumor cellularity, that is the relative proportion of tumor and normal cells of this sample, is manually estimated by pathologists. Currently, this process may take several days before results are available and is prone to errors and low agreement rates between assessors."
According to the paper's lead author, Neuromation researcher Alexander Rakhlin:
"In this work, we evaluated three strong novel Deep Learning-based approaches for automatic assessment of tumor cellularity from post-treated breast surgical specimens. We validated the proposed methods on the BreastPathQ SPIE challenge dataset that consisted of 2395 image patches selected from whole slide images acquired from 64 patients. Compared to expert pathologist scoring, our best performing method yielded a Cohen's kappa coefficient of 0.70 (vs. 0.42 previously known in literature) and an intra-class correlation coefficient of 0.89 (vs. 0.83). Our results suggest that Deep Learning-based methods have a significant potential to alleviate the burden on pathologists, enhance the diagnostic workflow, and, thereby, facilitate better clinical outcomes in breast cancer treatment.
Neuromation is extremely proud that this paper will be presented in October, as October is Breast Cancer Awareness Month, which is an annual campaign to increase awareness of the disease and to help those affected by breast cancer through early detection, education and support. We hope that this important work will eventually contribute to improved treatment and life outcomes for millions of women worldwide.
Media Contact:
Daria Posrednikova
E: dasha@neuromation.io