Anıl Yüce

Luca Deininger, Bernhard Stimpel, Anil Yuce et al.

Recently, vision transformers were shown to be capable of outperforming convolutional neural networks when pretrained on sufficient amounts of data. In comparison to convolutional neural networks, vision transformers have a weaker inductive bias and therefore allow a more flexible feature detection. Due to their promising feature detection, this work explores vision transformers for tumor detection in digital pathology whole slide images in four tissue types, and for tissue type identification. We compared the patch-wise classification performance of the vision transformer DeiT-Tiny to the state-of-the-art convolutional neural network ResNet18. Due to the sparse availability of annotated whole slide images, we further compared both models pretrained on large amounts of unlabeled whole-slide images using state-of-the-art self-supervised approaches. The results show that the vision transformer performed slightly better than the ResNet18 for three of four tissue types for tumor detection while the ResNet18 performed slightly better for the remaining tasks. The aggregated predictions of both models on slide level were correlated, indicating that the models captured similar imaging features. All together, the vision transformer models performed on par with the ResNet18 while requiring more effort to train. In order to surpass the performance of convolutional neural networks, vision transformers might require more challenging tasks to benefit from their weak inductive bias.

Tomé Albuquerque, Anil Yüce, Markus D. Herrmann et al.

Interpreting machine learning model decisions is crucial for high-risk applications like healthcare. In digital pathology, large whole slide images (WSIs) are decomposed into smaller tiles and tile-derived features are processed by attention-based multiple instance learning (ABMIL) models to predict WSI-level labels. These networks generate tile-specific attention weights, which can be visualized as attention maps for interpretability. However, a standardized evaluation framework for these maps is lacking, questioning their reliability and ability to detect spurious correlations that can mislead models. We herein propose a framework to assess the ability of attention networks to attend to relevant features in digital pathology by creating artificial model confounders and using dedicated interpretability metrics. Models are trained and evaluated on data with tile modifications correlated with WSI labels, enabling the analysis of model sensitivity to artificial confounders and the accuracy of attention maps in highlighting them. Confounders are introduced either through synthetic tile modifications or through tile ablations based on their specific image-based features, with the latter being used to assess more clinically relevant scenarios. We also analyze the impact of varying confounder quantities at both the tile and WSI levels. Our results show that ABMIL models perform as desired within our framework. While attention maps generally highlight relevant regions, their robustness is affected by the type and number of confounders. Our versatile framework has the potential to be used in the evaluation of various methods and the exploration of image-based features driving model predictions, which could aid in biomarker discovery.