Fig. 1: Construction framework of the collagen signature.
a Selection of the region of interest by comparing H&E staining and multiphoton imaging. The samples of all enroled 343 patients are used, and five regions of interest with a field of view of 500āĆā500 Ī¼m per sample within the invasive region of the gastric serosa are randomly selected for multiphoton imaging. Scale bars: 1500 Ī¼m, 150 Ī¼m, 50 Ī¼m and 50 Ī¼m, respectively. b A total of 146 collagen features for each enroled patient are extracted from multiphoton images, and predictive collagen features are selected using LASSO regression in the training cohort, and the collagen signature is constructed. The collagen signature of the validation cohort is calculated from the collagen signature calculation formula obtained in the training cohort. GLCM, grey-level co-occurrence matrix; H&E, hematoxylin and eosin; LASSO, least absolute shrinkage and selection operator.
