Shedding light on the black box of a neural network used to detect prostate cancer in whole slide images by occlusion-based explainability.

Diagnostic histopathology faces increasing demands due to aging populations and expanding healthcare programs. Semi-automated diagnostic systems employing deep learning methods are one approach to alleviate this pressure. The learning models for histopathology are inherently complex and opaque from the user's perspective. Hence different methods have been developed to interpret their behavior. However, relatively limited attention has been devoted to the connection between interpretation methods and the knowledge of experienced pathologists. The main contribution of this paper is a method for comparing morphological patterns used by expert pathologists to detect cancer with the patterns identified as important for inference of learning models. Given the patch-based nature of processing large-scale histopathological imaging, we have been able to show statistically that the VGG16 model could utilize all the structures that are observable by the pathologist, given the patch size and scan resolution. The results show that the neural network approach to recognizing prostatic cancer is similar to that of a pathologist at medium optical resolution. The saliency maps identified several prevailing histomorphological features characterizing carcinoma, e.g., single-layered epithelium, small lumina, and hyperchromatic nuclei with halo. A convincing finding was the recognition of their mimickers in non-neoplastic tissue. The method can also identify differences, i.e., standard patterns not used by the learning models and new patterns not yet used by pathologists. Saliency maps provide added value for automated digital pathology to analyze and fine-tune deep learning systems and improve trust in computer-based decisions.

Privacy risks of whole-slide image sharing in digital pathology.

Access to large volumes of so-called whole-slide images-high-resolution scans of complete pathological slides-has become a cornerstone of the development of novel artificial intelligence methods in pathology for diagnostic use, education/training of pathologists, and research. Nevertheless, a methodology based on risk analysis for evaluating the privacy risks associated with sharing such imaging data and applying the principle "as open as possible and as closed as necessary" is still lacking. In this article, we develop a model for privacy risk analysis for whole-slide images which focuses primarily on identity disclosure attacks, as these are the most important from a regulatory perspective. We introduce a taxonomy of whole-slide images with respect to privacy risks and mathematical model for risk assessment and design . Based on this risk assessment model and the taxonomy, we conduct a series of experiments to demonstrate the risks using real-world imaging data. Finally, we develop guidelines for risk assessment and recommendations for low-risk sharing of whole-slide image data.

Automated annotations of epithelial cells and stroma in hematoxylin-eosin-stained whole-slide images using cytokeratin re-staining.

The diagnosis of solid tumors of epithelial origin (carcinomas) represents a major part of the workload in clinical histopathology. Carcinomas consist of malignant epithelial cells arranged in more or less cohesive clusters of variable size and shape, together with stromal cells, extracellular matrix, and blood vessels. Distinguishing stroma from epithelium is a critical component of artificial intelligence (AI) methods developed to detect and analyze carcinomas. In this paper, we propose a novel automated workflow that enables large-scale guidance of AI methods to identify the epithelial component. The workflow is based on re-staining existing hematoxylin and eosin (H&E) formalin-fixed paraffin-embedded sections by immunohistochemistry for cytokeratins, cytoskeletal components specific to epithelial cells. Compared to existing methods, clinically available H&E sections are reused and no additional material, such as consecutive slides, is needed. We developed a simple and reliable method for automatic alignment to generate masks denoting cytokeratin-rich regions, using cell nuclei positions that are visible in both the original and the re-stained slide. The registration method has been compared to state-of-the-art methods for alignment of consecutive slides and shows that, despite being simpler, it provides similar accuracy and is more robust. We also demonstrate how the automatically generated masks can be used to train modern AI image segmentation based on U-Net, resulting in reliable detection of epithelial regions in previously unseen H&E slides. Through training on real-world material available in clinical laboratories, this approach therefore has widespread applications toward achieving AI-assisted tumor assessment directly from scanned H&E sections. In addition, the re-staining method will facilitate additional automated quantitative studies of tumor cell and stromal cell phenotypes.