Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region.
Journal
Scientific reports
ISSN: 2045-2322
Titre abrégé: Sci Rep
Pays: England
ID NLM: 101563288
Informations de publication
Date de publication:
05 10 2020
05 10 2020
Historique:
received:
22
01
2020
accepted:
15
09
2020
entrez:
6
10
2020
pubmed:
7
10
2020
medline:
5
1
2021
Statut:
epublish
Résumé
Manual count of mitotic figures, which is determined in the tumor region with the highest mitotic activity, is a key parameter of most tumor grading schemes. It can be, however, strongly dependent on the area selection due to uneven mitotic figure distribution in the tumor section. We aimed to assess the question, how significantly the area selection could impact the mitotic count, which has a known high inter-rater disagreement. On a data set of 32 whole slide images of H&E-stained canine cutaneous mast cell tumor, fully annotated for mitotic figures, we asked eight veterinary pathologists (five board-certified, three in training) to select a field of interest for the mitotic count. To assess the potential difference on the mitotic count, we compared the mitotic count of the selected regions to the overall distribution on the slide. Additionally, we evaluated three deep learning-based methods for the assessment of highest mitotic density: In one approach, the model would directly try to predict the mitotic count for the presented image patches as a regression task. The second method aims at deriving a segmentation mask for mitotic figures, which is then used to obtain a mitotic density. Finally, we evaluated a two-stage object-detection pipeline based on state-of-the-art architectures to identify individual mitotic figures. We found that the predictions by all models were, on average, better than those of the experts. The two-stage object detector performed best and outperformed most of the human pathologists on the majority of tumor cases. The correlation between the predicted and the ground truth mitotic count was also best for this approach (0.963-0.979). Further, we found considerable differences in position selection between pathologists, which could partially explain the high variance that has been reported for the manual mitotic count. To achieve better inter-rater agreement, we propose to use a computer-based area selection for support of the pathologist in the manual mitotic count.
Identifiants
pubmed: 33020510
doi: 10.1038/s41598-020-73246-2
pii: 10.1038/s41598-020-73246-2
pmc: PMC7536430
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
16447Références
Veta, M. et al. Assessment of algorithms for mitosis detection in breast cancer histopathology images. Med. Image Anal. 20, 237–248 (2015).
doi: 10.1016/j.media.2014.11.010
Baak, J. P. A. et al. Proliferation is the strongest prognosticator in node-negative breast cancer: significance, error sources, alternatives and comparison with molecular prognostic markers. Breast Cancer Res. Treat. 115, 241–254 (2008).
doi: 10.1007/s10549-008-0126-y
Elston, C. W. & Ellis, I. O. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology 19, 403–410 (1991).
doi: 10.1111/j.1365-2559.1991.tb00229.x
Sledge, D. G., Webster, J. & Kiupel, M. Canine cutaneous mast cell tumors: a combined clinical and pathologic approach to diagnosis, prognosis, and treatment selection. Vet. J. 215, 43–54 (2016).
doi: 10.1016/j.tvjl.2016.06.003
Kiupel, M. et al. Proposal of a 2-tier histologic grading system for canine cutaneous mast cell tumors to more accurately predict biological behavior. Vet. Pathol. 48, 147–155 (2011).
doi: 10.1177/0300985810386469
Azzola, M. F. et al. Tumor mitotic rate is a more powerful prognostic indicator than ulceration in patients with primary cutaneous melanoma. Cancer 97, 1488–1498 (2003).
doi: 10.1002/cncr.11196
Meuten, D. J. Appendix: diagnostic schemes and algorithms. In Tumors in Domestic Animals (ed. Meuten, D. J.) 942–978 (Wiley, Hoboken, 2016).
doi: 10.1002/9781119181200.app
Bertram, C. A. et al. Computerized calculation of mitotic distribution in canine cutaneous mast cell tumor sections: mitotic count is area dependent. Vet. Pathol. 57, 214–226 (2020).
doi: 10.1177/0300985819890686
Jannink, I., Risberg, B., Van Diest, P. J. & Baak, J. P. Heterogeneity of mitotic activity in breast cancer. Histopathology 29, 421–428 (1996).
doi: 10.1046/j.1365-2559.1996.d01-509.x
Tsuda, H. et al. Evaluation of the interobserver agreement in the number of mitotic figures breast carcinoma as simulation of quality monitoring in the Japan national surgical adjuvant study of breast cancer (NSAS-BC) protocol. Jpn. J. Cancer Res. 91, 451–457 (2000).
doi: 10.1111/j.1349-7006.2000.tb00966.x
Focke, C. M., Decker, T. & van Diest, P. J. Intratumoral heterogeneity of Ki67 expression in early breast cancers exceeds variability between individual tumours. Histopathology 69, 849–861 (2016).
doi: 10.1111/his.13007
Stålhammar, G. et al. Digital image analysis outperforms manual biomarker assessment in breast cancer. Mod. Pathol. 29, 318–329 (2016).
doi: 10.1038/modpathol.2016.34
Meyer, J. S. et al. Breast carcinoma malignancy grading by Bloom–Richardson system vs proliferation index: reproducibility of grade and advantages of proliferation index. Mod. Pathol. 18, 1067–1078 (2005).
doi: 10.1038/modpathol.3800388
Meyer, J. S., Cosatto, E. & Graf, H. P. Mitotic index of invasive breast carcinoma. Achieving clinically meaningful precision and evaluating tertial cutoffs. Arch. Pathol. Lab. Med. 133, 1826–1833 (2009).
pubmed: 19886719
Fauzi, M. F. A. et al. Classification of follicular lymphoma: the effect of computer aid on pathologists grading. BMC Med. Inform. Decis. 15, 115 (2015).
doi: 10.1186/s12911-015-0235-6
Bonert, M. & Tate, A. J. Mitotic counts in breast cancer should be standardized with a uniform sample area. BioMed. Eng. OnLine 16, 28 (2017).
doi: 10.1186/s12938-016-0301-z
Stålhammar, G. et al. Digital image analysis of Ki67 in hot spots is superior to both manual Ki67 and mitotic counts in breast cancer. Histopathology 72, 974–989 (2018).
doi: 10.1111/his.13452
Kaman, E. J., Smeulders, A. W. M., Verbeek, P. W., Young, I. T. & Baak, J. P. A. Image processing for mitoses in sections of breast cancer: a feasibility study. Cytometry 5, 244–249 (1984).
doi: 10.1002/cyto.990050305
Maier, A., Syben, C., Lasser, T. & Riess, C. A gentle introduction to deep learning in medical image processing. Z. Med. Phys. 29, 86–101 (2019).
doi: 10.1016/j.zemedi.2018.12.003
Cireşan, D. C., Giusti, A., Gambardella, L. M. & Schmidhuber, J. Mitosis detection in breast cancer histology images with deep neural networks. MICCAI 16, 411–418 (2013).
pubmed: 24579167
Roux, L. et al. Mitosis detection in breast cancer histological images an ICPR 2012 contest. J. Pathol. Inf. 4, 8 (2013).
doi: 10.4103/2153-3539.112693
Veta, M. et al. Predicting breast tumor proliferation from whole-slide images: the TUPAC16 challenge. Med. Image Anal. 54, 111–121 (2019).
doi: 10.1016/j.media.2019.02.012
He, K., Zhang, X., Ren, S. & Sun, J. Deep Residual Learning for Image Recognition. In CVPR, 770–778 (IEEE, 2016).
Li, C., Wang, X., Liu, W. & Latecki, L. J. DeepMitosis: Mitosis detection via deep detection, verification and segmentation networks. Med. Image Anal. 45, 121–133 (2018).
doi: 10.1016/j.media.2017.12.002
Pati, P., Catena, R., Goksel, O. & Gabrani, M. A deep learning framework for context-aware mitotic activity estimation in whole slide images. In Digital Pathology (eds Tomaszewski, J. E. & Ward, A. D.) 7–9 (SPIE, Bellingham, 2019).
Bertram, C. A., Aubreville, M., Marzahl, C., Maier, A. & Klopfleisch, R. A large-scale dataset for mitotic figure assessment on whole slide images of canine cutaneous mast cell tumor. Sci. Data 6, 1–9 (2019).
doi: 10.1038/s41597-019-0290-4
Aubreville, M., Bertram, C. A., Klopfleisch, R. & Maier, A. SlideRunner - A Tool for Massive Cell Annotations in Whole Slide Images. In Bildverarb. für die Med. 2018, 309–314 (Springer, 2018).
Aubreville, M., Bertram, C. A., Klopfleisch, R. & Maier, A. Augmented mitotic cell count using field of interest proposal. In Bildverarbeitung für die Medizin 2019, 321–326 (Springer, 2019).
Ren, S., He, K., Girshick, R. & Sun, J. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Trans. Pat. Anal. Mach. Intel. 39(6), 1137–1149 (2017).
doi: 10.1109/TPAMI.2016.2577031
Marzahl, C. et al. Deep learning-based quantification of pulmonary hemosiderophages in cytology slides. Sci. Rep. 10, 1–10 (2020).
doi: 10.1038/s41598-020-65958-2
Lin, T.-Y., Goyal, P., Girshick, R., He, K. & Dollar, P. Focal loss for dense object detection. In 2017 IEEE International Conference on Computer Vision (ICCV), 2999–3007 (IEEE, 2017).
Ronneberger, O., Fischer, P. & Brox, T. U-Net—convolutional networks for biomedical image segmentation. In MICCAI, 234–241 (Springer, 2015).
Rahman, M. A. & Wang, Y. Optimizing intersection-over-union in deep neural networks for image segmentation. In Advances in Visual Computing, 234–244 (Springer, Cham, Cham, 2016).
Veta, M., Pluim, J. P. W., van Diest, P. J. & Viergever, M. A. Breast cancer histopathology image analysis: a review. IEEE Trans. Biomed. Eng. 61, 1400–1411 (2014).
doi: 10.1109/TBME.2014.2303852
Roux, L. et al. MITOS & ATYPIA Detection of Mitosis and Evaluation of Nuclear Atypia Score in Breast Cencer Histological Images. In Image Pervasive Access Lab (IPAL), Agency Sci., Technol. & Res. Inst. Infocom Res., Singapore, Tech. Rep 1, 1–8 (2014).
Kirpensteijn, J., Kik, M., Rutteman, G. R. & Teske, E. Prognostic significance of a new histologic grading system for canine osteosarcoma. Vet. Pathol. 39, 240–246 (2002).
doi: 10.1354/vp.39-2-240
Loukopoulos, P. & Robinson, W. F. Clinicopathological relevance of tumour grading in canine osteosarcoma. J. Comp. Pathol. 136, 65–73 (2007).
doi: 10.1016/j.jcpa.2006.11.005
Meuten, D., Munday, J. S. & Hauck, M. Time to standardize? Time to validate?. Vet. Pathol. 55, 195–199 (2018).
doi: 10.1177/0300985817753869