TY - JOUR
T1 - TOP-GAN
T2 - Stain-free cancer cell classification using deep learning with a small training set
AU - Rubin, Moran
AU - Stein, Omer
AU - Turko, Nir A.
AU - Nygate, Yoav
AU - Roitshtain, Darina
AU - Karako, Lidor
AU - Barnea, Itay
AU - Giryes, Raja
AU - Shaked, Natan T.
N1 - Publisher Copyright:
© 2019
PY - 2019/10
Y1 - 2019/10
N2 - We propose a new deep learning approach for medical imaging that copes with the problem of a small training set, the main bottleneck of deep learning, and apply it for classification of healthy and cancer cell lines acquired by quantitative phase imaging. The proposed method, called transferring of pre-trained generative adversarial network (TOP-GAN), is hybridization between transfer learning and generative adversarial networks (GANs). Healthy cells and cancer cells of different metastatic potential have been imaged by low-coherence off-axis holography. After the acquisition, the optical path delay maps of the cells are extracted and directly used as inputs to the networks. In order to cope with the small number of classified images, we use GANs to train a large number of unclassified images from another cell type (sperm cells). After this preliminary training, we change the last layers of the network and design automatic classifiers for the correct cell type (healthy/primary cancer/metastatic cancer) with 90–99% accuracies, although small training sets of down to several images are used. These results are better in comparison to other classic methods that aim at coping with the same problem of a small training set. We believe that our approach makes the combination of holographic microscopy and deep learning networks more accessible to the medical field by enabling a rapid, automatic and accurate classification in stain-free imaging flow cytometry. Furthermore, our approach is expected to be applicable to many other medical image classification tasks, suffering from a small training set.
AB - We propose a new deep learning approach for medical imaging that copes with the problem of a small training set, the main bottleneck of deep learning, and apply it for classification of healthy and cancer cell lines acquired by quantitative phase imaging. The proposed method, called transferring of pre-trained generative adversarial network (TOP-GAN), is hybridization between transfer learning and generative adversarial networks (GANs). Healthy cells and cancer cells of different metastatic potential have been imaged by low-coherence off-axis holography. After the acquisition, the optical path delay maps of the cells are extracted and directly used as inputs to the networks. In order to cope with the small number of classified images, we use GANs to train a large number of unclassified images from another cell type (sperm cells). After this preliminary training, we change the last layers of the network and design automatic classifiers for the correct cell type (healthy/primary cancer/metastatic cancer) with 90–99% accuracies, although small training sets of down to several images are used. These results are better in comparison to other classic methods that aim at coping with the same problem of a small training set. We believe that our approach makes the combination of holographic microscopy and deep learning networks more accessible to the medical field by enabling a rapid, automatic and accurate classification in stain-free imaging flow cytometry. Furthermore, our approach is expected to be applicable to many other medical image classification tasks, suffering from a small training set.
KW - Biological cells
KW - Deep learning
KW - Holography
KW - Image classification
KW - Machine learning algorithms
KW - Quantitative phase imaging
UR - http://www.scopus.com/inward/record.url?scp=85068968504&partnerID=8YFLogxK
U2 - 10.1016/j.media.2019.06.014
DO - 10.1016/j.media.2019.06.014
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:85068968504
SN - 1361-8415
VL - 57
SP - 176
EP - 185
JO - Medical Image Analysis
JF - Medical Image Analysis
ER -