QuPath image analysis

Using QuPath software⁴⁶ digital-scanned tissue sections of EGFR IHC were first preprocessed in the built-in visual stain editor using default settings for estimation of stain vectors to improve staining quality. In each tissue section, the mesothelium was annotated by a pathologist (H.D.). Using a watershed segmentation method, cells were automatically detected and manually reconfirmed by a pathologist (H.D.) based on histomorphological features including cellular and nuclear shape. A minimum of 800 cells and a total of 77 parameters per cell (including cell perimeter, cell circularity, staining OD etc.) was quantified for each tissue sample. Results were exported as.csv files for statistical analysis in R.

For collagen quantification, sections were stained with Masson Trichrome (Sigma, #HT15-1KT) according to the manufacturer’s instruction. Two independent and blinded investigators (JZ, JM) annotated the adhesion area and exported it to ImageJ⁴⁶. The RGB image was split into the respective red, green, and blue image components. Collagen formation was quantified by measuring the area percentage of blue channel above a threshold of 120 units (8-bit).