Annotation-free glioma grading from pathological images using ensemble deep learning
Glioma grading is critical for treatment selection, and the fine classification between glioma grades II and III is still a pathological challenge. Traditional systems based on a single deep learning (DL) model can only show relatively low accuracy in distinguishing glioma grades II and III. Introdu...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-03-01
|
| Series: | Heliyon |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2405844023018613 |
| _version_ | 1797851718114869248 |
|---|---|
| author | Feng Su Ye Cheng Liang Chang Leiming Wang Gengdi Huang Peijiang Yuan Chen Zhang Yongjie Ma |
| author_facet | Feng Su Ye Cheng Liang Chang Leiming Wang Gengdi Huang Peijiang Yuan Chen Zhang Yongjie Ma |
| author_sort | Feng Su |
| collection | DOAJ |
| description | Glioma grading is critical for treatment selection, and the fine classification between glioma grades II and III is still a pathological challenge. Traditional systems based on a single deep learning (DL) model can only show relatively low accuracy in distinguishing glioma grades II and III. Introducing ensemble DL models by combining DL and ensemble learning techniques, we achieved annotation-free glioma grading (grade II or III) from pathological images. We established multiple tile-level DL models using residual network ResNet-18 architecture and then used DL models as component classifiers to develop ensemble DL models to achieve patient-level glioma grading. Whole-slide images of 507 subjects with low-grade glioma (LGG) from the Cancer Genome Atlas (TCGA) were included. The 30 DL models exhibited an average area under the curve (AUC) of 0.7991 in patient-level glioma grading. Single DL models showed large variation, and the median between-model cosine similarity was 0.9524, significantly smaller than the threshold of 1.0. The ensemble model based on logistic regression (LR) methods with a 14-component DL classifier (LR-14) demonstrated a mean patient-level accuracy and AUC of 0.8011 and 0.8945, respectively. Our proposed LR-14 ensemble DL model achieved state-of-the-art performance in glioma grade II and III classifications based on unannotated pathological images. |
| first_indexed | 2024-04-09T19:22:23Z |
| format | Article |
| id | doaj.art-735d48dfcecf4ea8a4e39dc870ca405b |
| institution | Directory Open Access Journal |
| issn | 2405-8440 |
| language | English |
| last_indexed | 2024-04-09T19:22:23Z |
| publishDate | 2023-03-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Heliyon |
| spelling | doaj.art-735d48dfcecf4ea8a4e39dc870ca405b2023-04-05T08:27:37ZengElsevierHeliyon2405-84402023-03-0193e14654Annotation-free glioma grading from pathological images using ensemble deep learningFeng Su0Ye Cheng1Liang Chang2Leiming Wang3Gengdi Huang4Peijiang Yuan5Chen Zhang6Yongjie Ma7Department of Neurobiology, School of Basic Medical Sciences, Beijing Key Laboratory of Neural Regeneration and Repair, Capital Medical University, Beijing 100069, China; Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing 100871, ChinaDepartment of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing 100053, China; Cell and Molecular Biology Lab of Neurosurgical Department, Xuanwu Hospital, Capital Medical University, Beijing 100053, China; CHINA-INI Scientific and Technological Innovation Lab, Beijing 100053, China; National Clinical Research Center for Geriatric Diseases, Beijing 100053, ChinaHenan Key Laboratory of Medical Tissue Regeneration, School of Basic Medical Sciences, Xinxiang Medical University, Henan, Xinxiang 453003, ChinaDepartment of Pathology, Xuanwu Hospital, Capital Medical University, Beijing 100053, ChinaPeking University Shenzhen Graduate School, Shenzhen 518055, ChinaSchool of Mechanical Engineering and Automation, Beihang University, Beijing 100191, China; Corresponding author.Department of Neurobiology, School of Basic Medical Sciences, Beijing Key Laboratory of Neural Regeneration and Repair, Capital Medical University, Beijing 100069, China; Corresponding author.Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing 100053, China; Cell and Molecular Biology Lab of Neurosurgical Department, Xuanwu Hospital, Capital Medical University, Beijing 100053, China; CHINA-INI Scientific and Technological Innovation Lab, Beijing 100053, China; Corresponding author. Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing 100053, China.Glioma grading is critical for treatment selection, and the fine classification between glioma grades II and III is still a pathological challenge. Traditional systems based on a single deep learning (DL) model can only show relatively low accuracy in distinguishing glioma grades II and III. Introducing ensemble DL models by combining DL and ensemble learning techniques, we achieved annotation-free glioma grading (grade II or III) from pathological images. We established multiple tile-level DL models using residual network ResNet-18 architecture and then used DL models as component classifiers to develop ensemble DL models to achieve patient-level glioma grading. Whole-slide images of 507 subjects with low-grade glioma (LGG) from the Cancer Genome Atlas (TCGA) were included. The 30 DL models exhibited an average area under the curve (AUC) of 0.7991 in patient-level glioma grading. Single DL models showed large variation, and the median between-model cosine similarity was 0.9524, significantly smaller than the threshold of 1.0. The ensemble model based on logistic regression (LR) methods with a 14-component DL classifier (LR-14) demonstrated a mean patient-level accuracy and AUC of 0.8011 and 0.8945, respectively. Our proposed LR-14 ensemble DL model achieved state-of-the-art performance in glioma grade II and III classifications based on unannotated pathological images.http://www.sciencedirect.com/science/article/pii/S2405844023018613Glioma gradingDeep learningEnsemble learningPathology |
| spellingShingle | Feng Su Ye Cheng Liang Chang Leiming Wang Gengdi Huang Peijiang Yuan Chen Zhang Yongjie Ma Annotation-free glioma grading from pathological images using ensemble deep learning Heliyon Glioma grading Deep learning Ensemble learning Pathology |
| title | Annotation-free glioma grading from pathological images using ensemble deep learning |
| title_full | Annotation-free glioma grading from pathological images using ensemble deep learning |
| title_fullStr | Annotation-free glioma grading from pathological images using ensemble deep learning |
| title_full_unstemmed | Annotation-free glioma grading from pathological images using ensemble deep learning |
| title_short | Annotation-free glioma grading from pathological images using ensemble deep learning |
| title_sort | annotation free glioma grading from pathological images using ensemble deep learning |
| topic | Glioma grading Deep learning Ensemble learning Pathology |
| url | http://www.sciencedirect.com/science/article/pii/S2405844023018613 |
| work_keys_str_mv | AT fengsu annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT yecheng annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT liangchang annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT leimingwang annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT gengdihuang annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT peijiangyuan annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT chenzhang annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning AT yongjiema annotationfreegliomagradingfrompathologicalimagesusingensembledeeplearning |