Bingyan Qin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Bingyan Qin has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Changbiao Cao
Department of Pathology, First Affiliated Hospital of Guangxi Medical University, Nanning, China
1 shared publication - Gang Chen
Department of Pathology, First Affiliated Hospital of Guangxi Medical University, Nanning, People’s Republic of China
1 shared publication - Jian-Di Li
Department of Pathology, The First Affiliated Hospital of Guangxi Medical University, Nanning,China
1 shared publication - Jing-Wen Ling
School of Information and Management, Guangxi Medical University , No. 336 Wuhua Avenue, Wuming District, Nanning, Guangxi Zhuang Autonomous Region 530199
1 shared publication - Ke-Jun Wu
Department of Pathology, The First Affiliated Hospital of Guangxi Medical University, Nanning, P.R. China
1 shared publication - Rui Lin
Department of Pathology, First Affiliated Hospital of Guangxi Medical University, Nanning, China
1 shared publication
Latest Publications
- 19 May 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (7) , 2017-2031
Multimodal integration of radiomics, pathomics, and clinical data enhances grading of adult diffuse gliomas using machine learning
Adult diffuse gliomas exhibit marked heterogeneity, making comprehensive evaluation of their biological behavior difficult. This study aimed to develop and validate a multimodal machine learning framework that fuses MRI-based radiomic features, whole-slide image (WSI)-derived pathomic signatures, and clinical variables for the comprehensive assessment of adult diffuse gliomas. Radiomic features were extracted from multiparametric MRI, and pathomic features from hematoxylin-eosin (H&E) stained WSIs; both were combined with clinical variables (age, sex, tumor dimensions, anatomical location). Unimodal radiomic/pathomic models and multimodal integrated models were built using three machine learning algorithms: random forest (RF), support vector machine (SVM), and eXtreme Gradient Boosting (XGBoost). A retrospective cohort of 373 pathologically confirmed patients was used for model construction and internal validation, with an independent external cohort of 49 patients for external validation. 40 radiomic and 20 pathomic features were retained via statistical testing and RF-based feature ranking. The optimal unimodal radiomic and pathomic models achieved internal validation AUCs of 0.89 and 0.83, respectively. The multimodal model showed superior performance (internal AUC = 0.90) and stable generalizability in external validation (AUC = 0.93). The multimodal model achieved AUC 0.93 with higher balanced accuracy and sensitivity. DeLong tests showed no statistically significant difference between multimodal and radiomics (P = 0.377) or pathomics (P = 0.085) in internal test; however, the multimodal model showed clinically meaningful improvements in sensitivity and balanced accuracy. The developed and validated multimodal machine learning model integrating radiomics, pathomics, and clinical information exhibits stable and reliable performance for the grading assessment of adult diffuse gliomas.
- 19 May 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (7) , 2017-2031
Multimodal integration of radiomics, pathomics, and clinical data enhances grading of adult diffuse gliomas using machine learning
Adult diffuse gliomas exhibit marked heterogeneity, making comprehensive evaluation of their biological behavior difficult. This study aimed to develop and validate a multimodal machine learning framework that fuses MRI-based radiomic features, whole-slide image (WSI)-derived pathomic signatures, and clinical variables for the comprehensive assessment of adult diffuse gliomas. Radiomic features were extracted from multiparametric MRI, and pathomic features from hematoxylin-eosin (H&E) stained WSIs; both were combined with clinical variables (age, sex, tumor dimensions, anatomical location). Unimodal radiomic/pathomic models and multimodal integrated models were built using three machine learning algorithms: random forest (RF), support vector machine (SVM), and eXtreme Gradient Boosting (XGBoost). A retrospective cohort of 373 pathologically confirmed patients was used for model construction and internal validation, with an independent external cohort of 49 patients for external validation. 40 radiomic and 20 pathomic features were retained via statistical testing and RF-based feature ranking. The optimal unimodal radiomic and pathomic models achieved internal validation AUCs of 0.89 and 0.83, respectively. The multimodal model showed superior performance (internal AUC = 0.90) and stable generalizability in external validation (AUC = 0.93). The multimodal model achieved AUC 0.93 with higher balanced accuracy and sensitivity. DeLong tests showed no statistically significant difference between multimodal and radiomics (P = 0.377) or pathomics (P = 0.085) in internal test; however, the multimodal model showed clinically meaningful improvements in sensitivity and balanced accuracy. The developed and validated multimodal machine learning model integrating radiomics, pathomics, and clinical information exhibits stable and reliable performance for the grading assessment of adult diffuse gliomas.