โก Quick Summary
This study developed an interpretable deep learning model to predict the molecular subtypes of endometrial cancer from H&E-stained whole slide images. The model achieved a macro-average AUROC of 0.867, demonstrating its potential for enhancing personalized treatment strategies in endometrial cancer.
๐ Key Details
- ๐ Dataset: Fudan cohort (n=364), TCGA (n=296), Suzhou (n=36)
- ๐งฉ Features used: H&E-stained whole slide images (WSIs)
- โ๏ธ Technology: Deep learning model for end-to-end prediction
- ๐ Performance: Macro-average AUROC of 0.867 (95% CI: 0.823-0.911)
๐ Key Takeaways
- ๐ฌ Molecular subtypes of endometrial cancer are crucial for prognosis and treatment.
- ๐ค Deep learning offers a novel approach to predict these subtypes accurately.
- ๐ Class-wise AUROCs: MSI-H: 0.846, NSMP: 0.876, p53abn: 0.910, POLEmut: 0.835.
- ๐ Morphological features correlate with molecular subtypes, aiding in interpretation.
- ๐ก This model provides a theoretical basis for individualized treatment strategies.
- ๐ External validation enhances the model’s generalizability and clinical applicability.
๐ Background
Endometrial cancer is a significant health concern, with its molecular subtypes playing a vital role in determining patient prognosis and treatment effectiveness. Traditional methods of subtype classification can be time-consuming and subjective. The integration of deep learning into pathology offers a promising avenue for improving diagnostic accuracy and treatment personalization.
๐๏ธ Study
The study utilized data from the Fudan cohort, comprising 364 patients, to train a deep learning model aimed at predicting four distinct molecular subtypes of endometrial cancer. The model’s performance was validated using two external cohorts, TCGA and Suzhou, ensuring its robustness and applicability in clinical settings.
๐ Results
The deep learning model demonstrated impressive performance, achieving a macro-average AUROC of 0.867 during 5-fold cross-validation. The class-wise AUROCs indicated strong predictive capabilities across all subtypes, with the p53-abnormal subtype showing the highest AUROC of 0.910. These results underscore the model’s potential as a reliable tool for molecular subtype prediction.
๐ Impact and Implications
The implications of this study are profound. By providing an accurate and interpretable tool for predicting molecular subtypes, this model paves the way for more tailored treatment strategies in endometrial cancer. The ability to correlate histological features with molecular characteristics enhances our understanding of tumor biology and could lead to improved patient outcomes.
๐ฎ Conclusion
This research highlights the transformative potential of deep learning in the field of oncology, particularly in the context of endometrial cancer. The development of an interpretable model for predicting molecular subtypes not only enhances diagnostic accuracy but also supports the move towards personalized medicine. Continued exploration in this area is essential for advancing cancer treatment and improving patient care.
๐ฌ Your comments
What are your thoughts on the use of deep learning in cancer diagnostics? We invite you to share your insights and engage in a discussion! ๐ฌ Leave your comments below or connect with us on social media:
An interpretable deep learning model for predicting endometrial cancer molecular subtypes from H&E-stained slides.
Abstract
The molecular subtype of endometrial cancer is important for predicting prognosis and treatment effectiveness. This study aimed to develop an interpretable deep learning model based on H&E-stained whole slide images (WSIs) to predict the molecular subtype of endometrial cancer. Data from the Fudan cohort (nโ=โ364) were used to train an end-to-end prediction network for identifying four molecular subtypes. Two external cohorts-the TCGA (nโ=โ296) and Suzhou (nโ=โ36)-were used to validate model generalizability and potential clinical applicability. We further assessed the correlation between histological and molecular features at both the macro- (WSI) and micro- (patch) levels. The network achieved a macro-average area under the receiver operating characteristic curve (AUROC) of 0.867 (95% CI: 0.823-0.911) in 5-fold cross-validation. The class-wise AUROCs were 0.846 (95% CI: 0.798-0.894) for the microsatellite instability-high (MSI-H) subtype, 0.876 (95% CI: 0.831-0.921) for the nonspecific molecular profile (NSMP) subtype, 0.910 (95% CI: 0.818-1.000) for the p53-abnormal (p53abn) subtype, and 0.835 (95% CI: 0.784-0.886) for the POLE-mutated (POLEmut) subtype. Morphological analysis revealed that MSI-H-subtype tumors exhibited increased stromal lymphocytic infiltration; POLEmut tumors showed higher heterogeneity, solid growth patterns, and elevated tumor grade; p53abn tumors were characterized by papillary growth and serous-like features; while NSMP tumors demonstrated high stromal cellularity. This method provides an accurate and interpretable tool for molecular subtype prediction, offering a theoretical basis for future individualized treatment strategies in endometrial cancer.
Author: [‘Guo Q’, ‘Cui H’, ‘Zhang Y’, ‘Tang S’, ‘Yan W’, ‘Zhou X’, ‘Ding H’, ‘Zhou J’, ‘Ju X’, ‘Feng Z’, ‘Zhu J’, ‘Bai F’, ‘Zhong Y’, ‘Li H’, ‘Xu J’, ‘Wu X’, ‘Wang X’, ‘Wen H’]
Journal: NPJ Precis Oncol
Citation: Guo Q, et al. An interpretable deep learning model for predicting endometrial cancer molecular subtypes from H&E-stained slides. An interpretable deep learning model for predicting endometrial cancer molecular subtypes from H&E-stained slides. 2026; (unknown volume):(unknown pages). doi: 10.1038/s41698-026-01280-w