Tumor-infiltrating lymphocytes (TILs) are prognostic in invasive breast cancer. However, their prognostic significance in ductal carcinoma in situ (DCIS) has been controversial. To investigate the prognostic role of TILs in DCIS outcome, we used different scoring methods for TILs in multi-national cohorts from Asian and European women. Self-described race was genetically confirmed using QC Infinium array combined with radmixture software. Stromal TILs, touching TILs, circumferential TILs, and hotspots were quantified on H&E-stained slides and correlated with the development of second breast cancer events (BCE) and other clinico-pathological variables. In univariate survival analysis, age older than 50 years, hormone receptor positivity and the presence of circumferential TILs were weakly associated with the absence of BCE at the 5-year follow-up in all cohorts (p < 0.03; p < 0.02; and p < 0.02, respectively, adjusted p = 0.11). In the multivariable analysis, circumferential TILs were an independent predictor of a better outcome (Wald test p = 0.01), whereas younger age was associated with BCE. Asian patients were younger with larger, higher grade, HR negative DCIS lesions, and higher TIL variables. The spatial arrangement of TILs may serve as a better prognostic indicator in DCIS cases than stromal TILs alone and may be added in guidelines for TILs evaluation in DCIS.

Standard clinicopathological parameters (age, growth pattern, tumor size, margin status, and grade) have been shown to have limited value in predicting recurrence in ductal carcinoma in situ (DCIS) patients. Early and accurate recurrence prediction would facilitate a more aggressive treatment policy for high-risk patients (mastectomy or adjuvant radiation therapy), and simultaneously reduce over-treatment of low-risk patients. Generative adversarial networks (GAN) are a class of DL models in which two adversarial neural networks, generator and discriminator, compete with each other to generate high quality images. In this work, we have developed a deep learning (DL) classification network that predicts breast cancer events (BCEs) in DCIS patients using hematoxylin and eosin (H & E) images. The DL classification model was trained on 67 patients using image patches from the actual DCIS cores and GAN generated image patches to predict breast cancer events (BCEs). The hold-out validation dataset (n = 66) had an AUC of 0.82. Bayesian analysis further confirmed the independence of the model from classical clinicopathological parameters. DL models of H & E images may be used as a risk stratification strategy for DCIS patients to personalize therapy.