Development and Validation of a Pathomics Model for Prognosis Prediction in Neoadjuvant Therapy-Treated Breast Cancer: A Retrospective, Multicenter Study

Abstract

Accurate outcome prediction is essential for personalized therapy in patients with breast cancer receiving neoadjuvant therapy (NAT). In this study, we developed and validated a multimodal artificial intelligence model, the ClinicHistomics Integrated Outcome Prediction Model (CIOPM), to predict prognosis in NAT-treated breast cancer patients by integrating clinicopathological data and hematoxylin and eosin (H&E)-stained surgical specimen whole-slide images (WSIs). A total of 847 WSIs from 835 patients across four multicenter cohorts collected between January 2008 and May 2020 were enrolled in model development and external validation. The CIOPM demonstrated robust predictive performance for overall survival (OS) and disease-free survival (DFS) in the two validation cohorts (VCs), yielding C-indices of 0.933 (95% CI: 0.878–0.977) and 0.915 (95% CI: 0.850–0.960) for OS in VC 1, and 0.947 (95% CI: 0.895–0.983) and 0.937 (95% CI: 0.905–0.965) for DFS in VC 2, respectively. The CIOPM enables accurate stratification of patients with NAT-treated breast cancer into high- and low-risk groups, achieving an AUC of 0.957 for both OS and DFS prediction. Moreover, it demonstrated superior performance in ablation studies, subgroup analyses, and clinical risk model comparisons. This model provides reliable risk stratification and personalized treatment decisions.

Keywords

breast cancer / deep learning / neoadjuvant therapy / prognosis / whole-slide image

References

[1] W. J. Gradishar, M. S. Moran, J. Abraham, et al., “Breast Cancer, Version 3.2022, NCCN Clinical Practice Guidelines in Oncology,” Journal of the National Comprehensive Cancer Network 20, no. 6 (2022): 691–722.

[2] L. M. Spring, G. Fell, A. Arfe, et al., “Pathologic Complete Response After Neoadjuvant Chemotherapy and Impact on Breast Cancer Recurrence and Survival: A Comprehensive Meta-Analysis,” Clinical Cancer Research 26, no. 12 (2020): 2838–2848.

[3] G. von Minckwitz, J. U. Blohmer, S. D. Costa, et al., “Response-Guided Neoadjuvant Chemotherapy for Breast Cancer,” Journal of Clinical Oncology 31, no. 29 (2013): 3623–3630.

[4] B. Asselain, W. Barlow, J. Bartlett, et al., “Long-Term Outcomes for Neoadjuvant Versus Adjuvant Chemotherapy in Early Breast Cancer: Meta-Analysis of Individual Patient Data From Ten Randomised Trials,” Lancet Oncology 19, no. 1 (2018): 27–39.

[5] J. Chen, L. Jin, and L. Chen, “Patients Achieved pCR During Neoadjuvant Chemotherapy Had Better Outcome Than Adjuvant Chemotherapy Setting in Breast Cancer: A Comparative Study,” Cancer Treatment and Research Communications 36 (2023): 100719.

[6] C. Liedtke, C. Mazouni, K. R. Hess, et al., “Response to Neoadjuvant Therapy and Long-Term Survival in Patients With Triple-Negative Breast Cancer,” Journal of Clinical Oncology 41, no. 10 (2023): 1809–1815.

[7] F. Conforti, L. Pala, I. Sala, et al., “Evaluation of Pathological Complete Response as Surrogate Endpoint in Neoadjuvant Randomised Clinical Trials of Early Stage Breast Cancer: Systematic Review and Meta-Analysis,” BMJ (Clinical Research Ed) 375 (2021): e066381.

[8] F. Li, Y. Yang, Y. Wei, et al., “Deep Learning-Based Predictive Biomarker of Pathological Complete Response to Neoadjuvant Chemotherapy From Histological Images in Breast Cancer,” Journal of Translational Medicine 19, no. 1 (2021): 348.

[9] S. J. Sammut, M. Crispin-Ortuzar, S. F. Chin, et al., “Multi-Omic Machine Learning Predictor of Breast Cancer Therapy Response,” Nature 601, no. 7894 (2022): 623–629.

[10] H. Xin, Y. Zhang, Q. Lai, et al., “Automatic Origin Prediction of Liver Metastases via Hierarchical Artificial-Intelligence System Trained on Multiphasic CT Data: A Retrospective, Multicentre Study,” EClinicalMedicine 69 (2024): 102464.

[11] S. Park, C. Y. Ock, H. Kim, et al., “Artificial Intelligence-Powered Spatial Analysis of Tumor-Infiltrating Lymphocytes as Complementary Biomarker for Immune Checkpoint Inhibition in Non-Small-Cell Lung Cancer,” Journal of Clinical Oncology 40, no. 17 (2022): 1916–1928.

[12] X. Chen, X. Wang, K. Zhang, et al., “Recent Advances and Clinical Applications of Deep Learning in Medical Image Analysis,” Medical Image Analysis 79 (2022): 102444.

[13] H. Shang, J. Li, T. Jiao, et al., “Differentiation of Lung Metastases Originated From Different Primary Tumors Using Radiomics Features Based on CT Imaging,” Academic Radiology 30, no. 1 (2023): 40–46.

[14] Z. Wang, L. Yu, X. Ding, X. Liao, and L. Wang, “Shared-Specific Feature Learning With Bottleneck Fusion Transformer for Multi-Modal Whole Slide Image Analysis,” IEEE Transactions on Medical Imaging 42, no. 11 (2023): 3374–3383.

[15] J. Lipkova, R. J. Chen, B. Chen, et al., “Artificial Intelligence for Multimodal Data Integration in Oncology,” Cancer Cell 40, no. 10 (2022): 1095–1110.

[16] H. Li, F. Yang, and X. Xing, “Multi-Modal Multi-Instance Learning Using Weakly Correlated Histopathological Images and Tabular Clinical Information,” in 24th International Conference on Medical Image Computing and Computer Assisted Intervention (Springer International Publishing, 2021), 529–539.

[17] F. Li, Y. Yang, Y. Wei, et al., “Predicting Neoadjuvant Chemotherapy Benefit Using Deep Learning From Stromal Histology in Breast Cancer,” NPJ Breast Cancer 8, no. 1 (2022): 124.

[18] X. Wang, J. Zhao, E. Marostica, et al., “A Pathology Foundation Model for Cancer Diagnosis and Prognosis Prediction,” Nature 634, no. 8035 (2024): 970–978.

[19] J. Xiang, X. Wang, X. Zhang, et al., “A Vision-Language Foundation Model for Precision Oncology,” Nature 638, no. 8051 (2025): 769–778.

[20] Z. Gu, R. Eils, and M. Schlesner, “Complex Heatmaps Reveal Patterns and Correlations in Multidimensional Genomic Data,” Bioinformatics 32, no. 18 (2016): 2847–2849.

[21] Z. Liu, X. Meng, H. Zhang, et al., “Predicting Distant Metastasis and Chemotherapy Benefit in Locally Advanced Rectal Cancer,” Nature Communications 11, no. 1 (2020): 4308.

[22] A. E. Giuliano, S. B. Edge, and G. N. Hortobagyi, “Eighth Edition of the AJCC Cancer Staging Manual: Breast Cancer,” Annals of Surgical Oncology 25, no. 7 (2018): 1783–1785.

[23] W. F. Symmans, F. Peintinger, C. Hatzis, et al., “Measurement of Residual Breast Cancer Burden to Predict Survival After Neoadjuvant Chemotherapy,” Journal of Clinical Oncology 25, no. 28 (2007): 4414–4422.

[24] E. A. Mittendorf, J. Vila, S. L. Tucker, et al., “The Neo-Bioscore Update for Staging Breast Cancer Treated With Neoadjuvant Chemotherapy: Incorporation of Prognostic Biologic Factors Into Staging After Treatment,” JAMA Oncology 2, no. 7 (2016): 929–936.

[25] M. Ilse, J. Tomczak, and M. Welling, “Attention-Based Deep Multiple Instance Learning,” in International Conference on Machine Learning (PMLR, 2018), 2127–2136.

[26] T. Ching, X. Zhu, and L. X. Garmire, “Cox-nnet: An Artificial Neural Network Method for Prognosis Prediction of High-Throughput Omics Data,” PLoS Computational Biology 14, no. 4 (2018): e1006076.

[27] I. Loshchilov and F. Hutter, “Decoupled Weight Decay Regularization,” preprint, arXiv, November 14, 2017.

[28] H. Ishwaran, U. B. Kogalur, E. H. Blackstone, and M. S. Lauer, “Random Survival Forests,” Annals of Applied Statistics 2, no. 3 (2008): 841–860.

[29] H. Ishwaran and U. B. Kogalur, “Random Survival Forests for R,” R News 7, no. 2 (2007): 25–31.

[30] S. Pölsterl, “scikit-survival: A Library for Time-to-Event Analysis Built on Top of scikit-learn,” Journal of Machine Learning Research 21, no. 212 (2020): 1–6.

[31] M. Tan and Q. Le, “Efficientnet: Rethinking Model Scaling for Convolutional Neural Networks,” in International Conference on Machine Learning (PMLR, 2019), 6105–6114.

[32] T. Hothorn and B. Lausen, “Maximally Selected Rank Statistics in R,” R News 2, no. 1 (2002): 3–5.

Rights & permissions

2026 The Author(s). MedComm published by Sichuan International Medical Exchange & Promotion Association (SCIMEA) and John Wiley & Sons Australia, Ltd.