Skip to main content
. 2023 Aug 30;9:71. doi: 10.1038/s41523-023-00577-4

Fig. 8. Enhanced prediction of neoadjuvant chemotherapy response with deep learning (DL)-assisted stromal tumor-infiltrating lymphocyte (sTIL) evaluation.

Fig. 8

a, b The average sTIL score in the responder group (Miller–Payne grade 4 and 5) and the non-responder group (Miller–Payne grade 1–3) was interpreted by pathologists without (a) and with DL-assistance (b). c The DL standalone interpretation of sTIL score in the responder and the non-responder group (The center lines in the boxplot represent median values; the bounds of the boxplot represent the interquartile ranges; the whiskers represent the range of the data). d–f The odds ratio (OR) of sTIL high group (sTIL ≥ 50) and intermediate group (sTIL 1–49) compared to sTIL low group (sTIL < 10) in relation to the neoadjuvant chemotherapeutic response, d in the entire HER2-positive breast cancer and triple-negative breast cancer (TNBC) dataset, e in the HER2-positive breast cancers, and f in the TNBCs (error bars, 95% confidence interval). * denotes p < 0.05, and ** denotes p < 0.01. The p-values were calculated using a student’s t-test for (a–c) and binary logistic regression for (d–f).