Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images.

Autor: Yu G; Department of Biomedical Engineering, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China., Sun K; Department of Biomedical Engineering, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China., Xu C; Department of Biostatistics and Epidemiology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, 73104, USA., Shi XH; Department of Computer & Information Sciences, College of Science and Technology, Temple University, Philadelphia, PA, 19122, USA., Wu C; Department of Statistics, Florida State University, Tallahassee, FL, 32306, USA., Xie T; Department of Biomedical Engineering, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China., Meng RQ; Electronic Information Science and Technology, School of Physics and Electronics, Central South University, 410083, Changsha, Hunan, China., Meng XH; Center for System Biology, Data Sciences and Reproductive Health, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China., Wang KS; Department of Pathology, Xiangya Hospital, School of Basic Medical Science, Central South University, 410078, Changsha, Hunan, China. 375527162@qq.com., Xiao HM; Center for System Biology, Data Sciences and Reproductive Health, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China. hmxiao@csu.edu.cn., Deng HW; Center for System Biology, Data Sciences and Reproductive Health, School of Basic Medical Science, Central South University, 410013, Changsha, Hunan, China. hdeng2@tulane.edu.; Deming Department of Medicine, Tulane Center of Biomedical Informatics and Genomics, Tulane University School of Medicine, New Orleans, LA, 70112, USA. hdeng2@tulane.edu.
Jazyk: angličtina
Zdroj: Nature communications [Nat Commun] 2021 Nov 02; Vol. 12 (1), pp. 6311. Date of Electronic Publication: 2021 Nov 02.
DOI: 10.1038/s41467-021-26643-8
Abstrakt: Machine-assisted pathological recognition has been focused on supervised learning (SL) that suffers from a significant annotation bottleneck. We propose a semi-supervised learning (SSL) method based on the mean teacher architecture using 13,111 whole slide images of colorectal cancer from 8803 subjects from 13 independent centers. SSL (~3150 labeled, ~40,950 unlabeled; ~6300 labeled, ~37,800 unlabeled patches) performs significantly better than the SL. No significant difference is found between SSL (~6300 labeled, ~37,800 unlabeled) and SL (~44,100 labeled) at patch-level diagnoses (area under the curve (AUC): 0.980 ± 0.014 vs. 0.987 ± 0.008, P value = 0.134) and patient-level diagnoses (AUC: 0.974 ± 0.013 vs. 0.980 ± 0.010, P value = 0.117), which is close to human pathologists (average AUC: 0.969). The evaluation on 15,000 lung and 294,912 lymph node images also confirm SSL can achieve similar performance as that of SL with massive annotations. SSL dramatically reduces the annotations, which has great potential to effectively build expert-level pathological artificial intelligence platforms in practice.
(© 2021. The Author(s).)
Databáze: MEDLINE
Externí odkaz: