Generalization of prostate cancer classification for multiple sites using deep learning
(2018) 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018 2018-April. p.191-194- Abstract
Deep learning has the potential to drastically increase the accuracy and efficiency of prostate cancer diagnosis, which would be of uttermost use. Today the diagnosis is determined manually from H&E stained specimens using a light microscope. In this paper several different approaches based on convolutional neural networks for prostate cancer classification are presented and compared, using three different datasets with different origins. The issue that algorithms trained on a certain site might not generalize to other sites, due to for example inevitable stain variations, is highlighted. Two different techniques to overcome this complication are compared; by training the networks using color augmentation and by using digital stain... (More)
Deep learning has the potential to drastically increase the accuracy and efficiency of prostate cancer diagnosis, which would be of uttermost use. Today the diagnosis is determined manually from H&E stained specimens using a light microscope. In this paper several different approaches based on convolutional neural networks for prostate cancer classification are presented and compared, using three different datasets with different origins. The issue that algorithms trained on a certain site might not generalize to other sites, due to for example inevitable stain variations, is highlighted. Two different techniques to overcome this complication are compared; by training the networks using color augmentation and by using digital stain separation. Furthermore, the potential of using an autoencoder to get a more efficient downsampling is investigated, which turned out to be the method giving the best generalization. We achieve accuracies of 95% for classification of benign versus malignant tissue and 81% for Gleason grading for data from the same site as the training data. The corresponding accuracies for images from other sites are in average 88% and 52% respectively.
(Less)
- author
- Arvidsson, Ida LU ; Overgaard, Niels Christian LU ; Marginean, Felicia Elena LU ; Krzyzanowska, Agnieszka LU ; Bjartell, Anders LU ; Astrom, Kalle LU and Heyden, Anders LU
- organization
-
- eSSENCE: The e-Science Collaboration
- ELLIIT: the Linköping-Lund initiative on IT and mobile communication
- Mathematics (Faculty of Engineering)
- Urological cancer, Malmö (research group)
- Department of Translational Medicine
- BioCARE: Biomarkers in Cancer Medicine improving Health Care, Education and Innovation
- EpiHealth: Epidemiology for Health
- publishing date
- 2018-05-23
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- Autoencoder, Convolutional neural network, Digital stain separation, Gleason grade, Prostate cancer
- host publication
- 2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018
- volume
- 2018-April
- pages
- 4 pages
- publisher
- IEEE Computer Society
- conference name
- 15th IEEE International Symposium on Biomedical Imaging, ISBI 2018
- conference location
- Washington, United States
- conference dates
- 2018-04-04 - 2018-04-07
- external identifiers
-
- scopus:85048138355
- ISBN
- 9781538636367
- DOI
- 10.1109/ISBI.2018.8363552
- project
- Lund University AI Research
- Digital Optimisation of Gleason Scoring (DOGS)
- language
- English
- LU publication?
- yes
- id
- 3794a793-341a-4efe-adc1-e19d9d3130a6
- date added to LUP
- 2018-06-19 14:01:40
- date last changed
- 2024-08-19 19:42:24
@inproceedings{3794a793-341a-4efe-adc1-e19d9d3130a6, abstract = {{<p>Deep learning has the potential to drastically increase the accuracy and efficiency of prostate cancer diagnosis, which would be of uttermost use. Today the diagnosis is determined manually from H&E stained specimens using a light microscope. In this paper several different approaches based on convolutional neural networks for prostate cancer classification are presented and compared, using three different datasets with different origins. The issue that algorithms trained on a certain site might not generalize to other sites, due to for example inevitable stain variations, is highlighted. Two different techniques to overcome this complication are compared; by training the networks using color augmentation and by using digital stain separation. Furthermore, the potential of using an autoencoder to get a more efficient downsampling is investigated, which turned out to be the method giving the best generalization. We achieve accuracies of 95% for classification of benign versus malignant tissue and 81% for Gleason grading for data from the same site as the training data. The corresponding accuracies for images from other sites are in average 88% and 52% respectively.</p>}}, author = {{Arvidsson, Ida and Overgaard, Niels Christian and Marginean, Felicia Elena and Krzyzanowska, Agnieszka and Bjartell, Anders and Astrom, Kalle and Heyden, Anders}}, booktitle = {{2018 IEEE 15th International Symposium on Biomedical Imaging, ISBI 2018}}, isbn = {{9781538636367}}, keywords = {{Autoencoder; Convolutional neural network; Digital stain separation; Gleason grade; Prostate cancer}}, language = {{eng}}, month = {{05}}, pages = {{191--194}}, publisher = {{IEEE Computer Society}}, title = {{Generalization of prostate cancer classification for multiple sites using deep learning}}, url = {{http://dx.doi.org/10.1109/ISBI.2018.8363552}}, doi = {{10.1109/ISBI.2018.8363552}}, volume = {{2018-April}}, year = {{2018}}, }