Quantitative Whole Slide Assessment of Tumor-Infiltrating CD8-Positive Lymphocytes in ER-Positive Breast Cancer in Relation to Clinical Outcome.

There is a need for a reliable and reproducible quantification of the immune infiltrate within the heterogeneous microenvironment of tumors in order to support therapy selection in oncology. Here we present an automated, modular method for whole-slide image analysis of the spatial distribution of tumor-infiltrating CD8-positive lymphocytes. The method uses a deep learning tissue-type classification algorithm on the hematoxylin eosin (HE) stained tissue section to identify the central tumor (CT) and invasive margin (IM) of the tumor. A CD8-positive cell detection algorithm using a deep learning-based nucleus detection is applied to a sequential immunohistochemistry (IHC)-stained tissue section. Image registration then allows obtaining IHC-derived CD8 scores for the HE-derived CT and the IM, respectively. Both, the mean and the standard deviation of the spatial CD8-positive density distributions were determined for the CT and IM in a cohort of post-menopausal, estrogen receptor-positive invasive breast cancer patients who received adjuvant tamoxifen therapy. Spatial density distributions were found to be highly heterogeneous. In contrast to previous studies, CD8 density in the IM and CT correlated positively with clinical outcome. However, statistical significance was only achieved for the standard deviation of the CD8 density distribution. We hypothesize that this is due to the positive contribution of local high-density areas. The IM/CT density ratio did not correlate with outcome. In view of the clinical relevance of our finding, we would like to encourage a study with a larger cohort. Our modular pipeline approach allows a robust and objective scoring of CD8 infiltrate based on routine pathology staining and should contribute to clinical adoption of computational pathology.

Deep learning assessment of breast terminal duct lobular unit involution: Towards automated prediction of breast cancer risk.

Terminal duct lobular unit (TDLU) involution is the regression of milk-producing structures in the breast. Women with less TDLU involution are more likely to develop breast cancer. A major bottleneck in studying TDLU involution in large cohort studies is the need for labor-intensive manual assessment of TDLUs. We developed a computational pathology solution to automatically capture TDLU involution measures. Whole slide images (WSIs) of benign breast biopsies were obtained from the Nurses' Health Study. A set of 92 WSIs was annotated for acini, TDLUs and adipose tissue to train deep convolutional neural network (CNN) models for detection of acini, and segmentation of TDLUs and adipose tissue. These networks were integrated into a single computational method to capture TDLU involution measures including number of TDLUs per tissue area, median TDLU span and median number of acini per TDLU. We validated our method on 40 additional WSIs by comparing with manually acquired measures. Our CNN models detected acini with an F1 score of 0.73±0.07, and segmented TDLUs and adipose tissue with Dice scores of 0.84±0.13 and 0.87±0.04, respectively. The inter-observer ICC scores for manual assessments on 40 WSIs of number of TDLUs per tissue area, median TDLU span, and median acini count per TDLU were 0.71, 0.81 and 0.73, respectively. Intra-observer reliability was evaluated on 10/40 WSIs with ICC scores of >0.8. Inter-observer ICC scores between automated results and the mean of the two observers were: 0.80 for number of TDLUs per tissue area, 0.57 for median TDLU span, and 0.80 for median acini count per TDLU. TDLU involution measures evaluated by manual and automated assessment were inversely associated with age and menopausal status. We developed a computational pathology method to measure TDLU involution. This technology eliminates the labor-intensiveness and subjectivity of manual TDLU assessment, and can be applied to future breast cancer risk studies.