#Deepfocus crop images skin

However, this knowledge is of utmost importance for assisting systems to be implemented in the routine diagnostic workflow. Generalization of this information is difficult for models from different domains, tasks (classification, segmentation, detection), different neural network architectures, and models using different ground truth. Only few studies investigate how DL-based models will operate in case of artifacts.

The aim of the pathologist and of any assisting system (e.g., DL-based) is to provide top diagnostic accuracy irrespective of pre-analytical issues and artifacts. Such heterogeneity is prominent among different institutions and even within the same laboratory. Importantly, histological slides may be very heterogeneous with regard to staining, thickness of sections, and are especially prone to artifacts during tissue processing, cutting, staining, and digitization.

However, this is still a work in progress.

#Deepfocus crop images skin

Many studies to date addressed the feasibility of DL for diagnostic applications with high accuracy resulting in several commercial diagnostic products that are being developed for prostate cancer, breast cancer, skin diseases, and in other medical domains. The latter use unsupervised feature generation by convolving images using multiple different filters and aggregate the corresponding representations in different ways to achieve a prediction. Deep learning (DL) is a powerful technology for image analysis based on convolutional neural networks. One of the promises of digital pathology is an automated analysis of pathological specimens using deep learning-based models. At that, histological slides can be digitized using histological scanner and reviewed on the monitor without a microscope. Stress-testing of diagnostic models using synthetically generated artifacts might be an essential step during clinical validation of deep learning-based algorithms.ĭigital pathology is an emerging transformation of diagnostic pathology. Strategies for the prevention of diagnostic model accuracy losses in the context of artifacts are warranted.

We provide evidence that any histological artifact dependent on severity can lead to a substantial loss in model performance. Using six datasets from four different institutions digitized by different scanner systems, we systematically explore artifacts’ influence on the accuracy of the pre-trained, validated, deep learning-based model for prostate cancer detection in histological slides. In this study, we digitally reproduce major types of artifacts. Histological slides are very heterogeneous concerning staining, sections’ thickness, and artifacts arising during tissue processing, cutting, staining, and digitization. Digital pathology provides a possibility for computational analysis of histological slides and automatization of routine pathological tasks.