Improving Quality Control of Whole Slide Images by Explicit Artifact Augmentation

TL;DR

This work addresses the problem of artifacts in whole slide image acquisition by proposing a method dedicated to augmenting whole slide images with artifacts, which seamlessly generates and blends artifacts from an external library to a given histopathology dataset.

Abstract

The problem of artifacts in whole slide image acquisition, prevalent in both clinical workflows and research-oriented settings, necessitates human intervention and re-scanning. Overcoming this challenge requires developing quality control algorithms, that are hindered by the limited availability of relevant annotated data in histopathology. The manual annotation of ground-truth for artifact detection methods is expensive and time-consuming. This work addresses the issue by proposing a method dedicated to augmenting whole slide images with artifacts. The tool seamlessly generates and blends artifacts from an external library to a given histopathology dataset. The augmented datasets are then utilized to train artifact classification methods. The evaluation shows their usefulness in classification of the artifacts, where they show an improvement from 0.10 to 0.01 AUROC depending on the artifact type. The framework, model, weights, and ground-truth annotations are freely released to facilitate open science and reproducible research.

Figures (11)

• Figure 1: Aggregation of a whole slide image patch-level classification. Color meaning: (yellow) marker, (red) ink, (blue) dust, (green) tissue folding.
• Figure 2: Example of failure of a quality control algorithm in identifying artifacts within an IHC image: (left) coverslip edge, (right) small objects (e.g., dust).
• Figure 3: Examples of artifacts from the considered datasets: a) focus, b) tissue, c) dust, d) ink, e) air, f) marker.
• Figure 4: ROC curve for classification, evaluated on additional ACROBAT annotations unseen during training. (left) model trained on $\mathbf{ACR}$. (right) model trained on augmented $\mathbf{ACR'}$.
• Figure 5: Chart of the validation loss during training on $\mathbf{ACR}$ and $\mathbf{ACR'}$ datasets.