TY - GEN
T1 - How to fool radiologists with generative adversarial networks? A visual turing test for lung cancer diagnosis
AU - Chuquicusma, Maria J.M.
AU - Hussein, Sarfaraz
AU - Burt, Jeremy
AU - Bagci, Ulas
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/5/23
Y1 - 2018/5/23
N2 - Discriminating lung nodules as malignant or benign is still an underlying challenge. To address this challenge, radiologists need computer aided diagnosis (CAD) systems which can assist in learning discriminative imaging features corresponding to malignant and benign nodules. However, learning highly discriminative imaging features is an open problem. In this paper, our aim is to learn the most discriminative features pertaining to lung nodules by using an adversarial learning methodology. Specifically, we propose to use un-supervised learning with Deep Convolutional-Generative Adversarial Networks (DC-GANs) to generate lung nodule samples realistically. We hypothesize that imaging features of lung nodules will be discriminative if it is hard to differentiate them (fake) from real (true) nodules. To test this hypothesis, we present Visual Turing tests to two radiologists in order to evaluate the quality of the generated (fake) nodules. Extensive comparisons are performed in discerning real, generated, benign, and malignant nodules. This experimental set up allows us to validate the overall quality of the generated nodules, which can then be used to (1) improve diagnostic decisions by mining highly discriminative imaging features, (2) train radiologists for educational purposes, and (3) generate realistic samples to train deep networks with big data.
AB - Discriminating lung nodules as malignant or benign is still an underlying challenge. To address this challenge, radiologists need computer aided diagnosis (CAD) systems which can assist in learning discriminative imaging features corresponding to malignant and benign nodules. However, learning highly discriminative imaging features is an open problem. In this paper, our aim is to learn the most discriminative features pertaining to lung nodules by using an adversarial learning methodology. Specifically, we propose to use un-supervised learning with Deep Convolutional-Generative Adversarial Networks (DC-GANs) to generate lung nodule samples realistically. We hypothesize that imaging features of lung nodules will be discriminative if it is hard to differentiate them (fake) from real (true) nodules. To test this hypothesis, we present Visual Turing tests to two radiologists in order to evaluate the quality of the generated (fake) nodules. Extensive comparisons are performed in discerning real, generated, benign, and malignant nodules. This experimental set up allows us to validate the overall quality of the generated nodules, which can then be used to (1) improve diagnostic decisions by mining highly discriminative imaging features, (2) train radiologists for educational purposes, and (3) generate realistic samples to train deep networks with big data.
KW - Computed Tomography (CT)
KW - Computer Aided Diagnosis (CAD) systems
KW - Deep learning
KW - Generated samples
KW - Generative Adversarial Networks (GANs)
KW - Lung nodules
KW - Visual Turing Test
UR - http://www.scopus.com/inward/record.url?scp=85048118763&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85048118763&partnerID=8YFLogxK
U2 - 10.1109/ISBI.2018.8363564
DO - 10.1109/ISBI.2018.8363564
M3 - Conference contribution
AN - SCOPUS:85048118763
T3 - Proceedings - International Symposium on Biomedical Imaging
SP - 240
EP - 244
BT - 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018
PB - IEEE Computer Society
T2 - 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018
Y2 - 4 April 2018 through 7 April 2018
ER -