Get started
Work

MSCAnet: A multi-scale cross-attention network for chromosomal instability prediction in breast cancer whole-slide images

Year: 2026
Type: article
Abstract: Chromosomal instability (CIN) is an important biomarker for breast cancer (BC) diagnosis, prognosis, and treatment stratification. However, conventional laboratory-based methods for detecting CIN are time-consuming and costly, while deep learning methods often struggle to capture both local cellular details and global tissue context from gigapixel whole-slide images (WSIs). To address these challenges, we propose a multi-scale cross-attention neural network ( MSCAnet) for CIN status prediction from hematoxylin and eosin-stained breast cancer WSIs. MSCAnet employs dual patch-level branches to capture complementary nuclear and tissue structural patterns, with a CIN-aware cross-attention module dynamically integrating multi-scale features. On the The Cancer Genome Atlas Breast Invasive Carcinoma (TCGA-BRCA) dataset, MSCAnet achieved an accuracy of 77.46% and an area under the receiver operating characteristic curve of 0.809, representing the highest numerical performance among the evaluated methods. However, the 95% confidence intervals overlapped with those of the top-performing baselines, indicating that the observed differences were not statistically conclusive. Visualization further revealed intra-tumor heterogeneity in CIN patterns. In an additional task, MSCAnet achieved an accuracy of 82.72% for human epidermal growth factor receptor 2 status prediction, providing preliminary evidence of task transferability. Overall, MSCAnet provides a preliminary computational pathology framework for exploring molecular biomarker prediction from routine hematoxylin and eosin-stained WSIs, highlighting the potential of multi-scale attention mechanisms in computational oncology.
DOI: 10.1016/j.intonc.2026.100091
Language: English
Open Access status: diamond
OpenAlex ID: W7214618751
FWCI:
Cites: 27
Cited by:
Related to: 10