Radiopathomic Graph Deep Learning for Multiscale Spatial-Contextual Modeling of Intratumoral Heterogeneity to Predict Breast Cancer Response to Neoadjuvant Therapy - PubMed
5 hours ago
- #Radiopathomics
- #Graph Deep Learning
- #Breast Cancer
- The study developed an explainable Radiopathomic Graph Deep Learning (RPGDL) system for predicting pathological complete response to neoadjuvant therapy in breast cancer.
- The system uses graph neural networks on radiomic and pathomic graphs from MRI and biopsy images, trained on 582 patients and externally tested on 468 patients.
- The radio-pathomic GNN outperformed single-modality GNNs, achieving an AUC of 0.95 in training and 0.91 in external testing, with explainable insights into predictions.