Accurately matching serial whole slide images for tumor heterogeneity analysis

• Published in: Biomedical signal processing and control Vol. 100; p. 106825
• Main Authors: Li, Xiaoxiao, Mei, Liye, Long, Mengping, Huang, Jin, Yang, Wei, Liu, Yan, Hao, Xin, Liu, Yiqiang, Shen, Hui, Hou, Jinxuan, Xu, Yu, Zhou, Fuling, Wang, Du, Wu, Jianghua, Hu, Taobo, Lei, Cheng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.02.2025
• Subjects: Breast cancer; Image matching; Multiscale ring encoder; Pathology image; Serial slice; Tumor heterogeneity; Whole slide image; Whole slide image; Image matching; Tumor heterogeneity; Serial slice; Breast cancer; Multiscale ring encoder; Pathology image
• ISSN: 1746-8094
• DOI: 10.1016/j.bspc.2024.106825

Tumor diagnosis and heterogeneity assessment often need pathologists to interpret multiple serial whole slide images (WSIs). However, the task of accurately locating common regions at maximum magnification of multiple serial WSIs is challenging and laborious due to the large data volume and heterogeneity of images. Here we demonstrate the automated matching work of multi-stained serial WSIs by the proposed multiscale-ring-encoder. Three Giga-pixel serial WSIs stained with Hematoxylin-eosin (H&E) and two immunohistochemical (IHC) stains are registered in 9.36 min with a spatial uncertainty of less than 200μm, and only 0.68 s are needed to locate corresponding regions among the WSIs, more than 30 times faster than pathologists. Benefitting from the accurate matching of the method, quantification of Ki-67 across the whole breast cancer slide is achieved, enabling the quantification of intratumor heterogeneity benefiting personalized treatment. This method has the potential to significantly enhance cancer diagnosis and treatment by efficiently reducing the workload of pathologists and ensuring more precise tumor heterogeneity assessments. The code and software system are available at https://github.com/WHU-lab/MTH. •We match multiple-stain serial WSIs, including H&E and two IHC.•We precisely match WSIs at maximum magnification for diagnosis.•We quantify tumor heterogeneity of breast cancer Ki-67.•We match pathologists’ needs for accuracy, success rate, efficiency.