Each year, the treatment decisions for more than 230,000 breast cancer
patients in the U.S. hinge on whether the cancer has metastasized away from the
breast. Metastasis detection is currently performed by pathologists reviewing
large expanses of biological tissues. This process is labor intensive and
error-prone. We present a framework to automatically detect and localize tumors
as small as 100 x 100 pixels in gigapixel microscopy images sized 100,000 x
100,000 pixels. Our method leverages a convolutional neural network (CNN)
architecture and obtains state-of-the-art results on the Camelyon16 dataset in
the challenging lesion-level tumor detection task. At 8 false positives per
image, we detect 92.4% of the tumors, relative to 82.7% by the previous best
automated approach. For comparison, a human pathologist attempting exhaustive
search achieved 73.2% sensitivity. We achieve image-level AUC scores above 97%
on both the Camelyon16 test set and an independent set of 110 slides. In
addition, we discover that two slides in the Camelyon16 training set were
erroneously labeled normal. Our approach could considerably reduce false
negative rates in metastasis detection.
Description
Detecting Cancer Metastases on Gigapixel Pathology Images
cite arxiv:1703.02442Comment: Fig 1: normal and tumor patches were accidentally reversed - now fixed. Minor grammatical corrections in appendix, section "Image Color Normalization"
%0 Generic
%1 liu2017detecting
%A Liu, Yun
%A Gadepalli, Krishna
%A Norouzi, Mohammad
%A Dahl, George E.
%A Kohlberger, Timo
%A Boyko, Aleksey
%A Venugopalan, Subhashini
%A Timofeev, Aleksei
%A Nelson, Philip Q.
%A Corrado, Greg S.
%A Hipp, Jason D.
%A Peng, Lily
%A Stumpe, Martin C.
%D 2017
%K ai cnn deep example heathcare learning
%T Detecting Cancer Metastases on Gigapixel Pathology Images
%U http://arxiv.org/abs/1703.02442
%X Each year, the treatment decisions for more than 230,000 breast cancer
patients in the U.S. hinge on whether the cancer has metastasized away from the
breast. Metastasis detection is currently performed by pathologists reviewing
large expanses of biological tissues. This process is labor intensive and
error-prone. We present a framework to automatically detect and localize tumors
as small as 100 x 100 pixels in gigapixel microscopy images sized 100,000 x
100,000 pixels. Our method leverages a convolutional neural network (CNN)
architecture and obtains state-of-the-art results on the Camelyon16 dataset in
the challenging lesion-level tumor detection task. At 8 false positives per
image, we detect 92.4% of the tumors, relative to 82.7% by the previous best
automated approach. For comparison, a human pathologist attempting exhaustive
search achieved 73.2% sensitivity. We achieve image-level AUC scores above 97%
on both the Camelyon16 test set and an independent set of 110 slides. In
addition, we discover that two slides in the Camelyon16 training set were
erroneously labeled normal. Our approach could considerably reduce false
negative rates in metastasis detection.
@misc{liu2017detecting,
abstract = {Each year, the treatment decisions for more than 230,000 breast cancer
patients in the U.S. hinge on whether the cancer has metastasized away from the
breast. Metastasis detection is currently performed by pathologists reviewing
large expanses of biological tissues. This process is labor intensive and
error-prone. We present a framework to automatically detect and localize tumors
as small as 100 x 100 pixels in gigapixel microscopy images sized 100,000 x
100,000 pixels. Our method leverages a convolutional neural network (CNN)
architecture and obtains state-of-the-art results on the Camelyon16 dataset in
the challenging lesion-level tumor detection task. At 8 false positives per
image, we detect 92.4% of the tumors, relative to 82.7% by the previous best
automated approach. For comparison, a human pathologist attempting exhaustive
search achieved 73.2% sensitivity. We achieve image-level AUC scores above 97%
on both the Camelyon16 test set and an independent set of 110 slides. In
addition, we discover that two slides in the Camelyon16 training set were
erroneously labeled normal. Our approach could considerably reduce false
negative rates in metastasis detection.},
added-at = {2019-11-06T21:52:09.000+0100},
author = {Liu, Yun and Gadepalli, Krishna and Norouzi, Mohammad and Dahl, George E. and Kohlberger, Timo and Boyko, Aleksey and Venugopalan, Subhashini and Timofeev, Aleksei and Nelson, Philip Q. and Corrado, Greg S. and Hipp, Jason D. and Peng, Lily and Stumpe, Martin C.},
biburl = {https://www.bibsonomy.org/bibtex/25df9a5884dee2b576f8d10b7cf5d072c/nosebrain},
description = {Detecting Cancer Metastases on Gigapixel Pathology Images},
interhash = {b72bdf35f538ce5e1c2547b3cec39272},
intrahash = {5df9a5884dee2b576f8d10b7cf5d072c},
keywords = {ai cnn deep example heathcare learning},
note = {cite arxiv:1703.02442Comment: Fig 1: normal and tumor patches were accidentally reversed - now fixed. Minor grammatical corrections in appendix, section "Image Color Normalization"},
timestamp = {2021-04-29T15:23:18.000+0200},
title = {Detecting Cancer Metastases on Gigapixel Pathology Images},
url = {http://arxiv.org/abs/1703.02442},
year = 2017
}