Computer vision could help pathologists more easily identify cancer cells.
JACKSONVILLE, FLORIDA (PRWEB) APRIL 13, 2016
Jacksonville-based advanced data analytics solutions company NLP Logix has been selected as one of the top five applicants in an international competition to develop a machine learning model to help pathologists more efficiently and accurately detect cancer metastasis in lymph node images.
NLP Logix will present the methodology of its submission at the CAMELYON16 Challenge to be held April 13 in Prague. The conference is sponsored by the 2016 IEEE International Symposium on Biomedical Imaging (ISBI).
Digital pathology is an emerging field in the medical industry but the process of evaluating whole-slide, large pixel images of tissue specimens can be labor intensive and time consuming for pathologists attempting to detect cancer or identify metastases. The CAMELYON16 Challenge asked participants to apply a scientific approach for developing algorithms that would assist pathologists in improved cancer diagnosis of patients as well as reduce the workload of pathologists.
Under NLP Logix Lead Scientist Matt Berseth, the Jacksonville team began work on the project in January, building upon its proprietal LogixStudio machine learning toolkit by adding computer vision capabilities. “While we built the capabilities to be able to read pathology images” said Berseth, “we can apply this technology to just about any computer vision challenge regardless of industry.”
To reduce the size of the images, slides were broken down into roughly 250,000 image patches, each of which was compressed into smaller pixel images. An image patch was labeled as a positive if more than 50 percent of the pixels overlapped with the masked region.
The team used Google’s TensorFlow software package to “train” the network to recognize patterns within the image slides most likely to contain cancer tumors. Berseth and his colleagues also applied predictive modeling methods to identify each of 20 local regions where the tumors could be found.
The results of the competition will be revealed at the ISBI conference which can be followed via live stream at http://camelyon16.grand-challenge.org/program/.
According to Ted Willich, CEO of NLP Logix, the technology developed by the company could be deployed in the pathology industry with virtually no lead time required. “We’ve built it out in such a way that it could be put into production and used right way,” said Willich. “We really focus on getting that end product into the hands of our customers.”
