arXiv subjects

Guido Sauter

Publications and source records attributed to Guido Sauter.

4 recordsLinked to original sources

arXiv2026

EXPOSE: Explainable and Domain-Robust Embeddings from Pathology Vision Foundation Models using Sparse Autoencoders

Vision Foundation Models (VFMs) are widely used in computational pathology but remain sensitive to domain shifts arising from variations in staining, tissue preparation, and scanner hardware. A key limitation is that VFM embeddings entangle biological with domain-specific information, hindering cross-domain generalization. We propose Explainable Probing of Cross-Domain Sparse Embeddings (EXPOSE), a framework that uses Sparse Autoencoders (SAEs) as an explainable bottleneck to identify and suppress domain-specific components in VFM embeddings. We train a sparse representation of VFM features, use a linear classifier to identify domain-specific latent dimensions, and mask these features prior to downstream relapse prediction without retraining the backbone model. Experiments on a large prostate cancer dataset with multiple acquisition domains show that SAE features capture both domain- and task-specific information, which are partially disentangled in the latent space. Removing domain-specific features improves cross-domain performance and increases embedding robustness as measured by the Domain Robustness Index (DoRI). Code is available at https://github.com/imsb-uke/expose .

cs.CV↗

arXiv2025

Robust multicellular programs dissect the complex tumor microenvironment and track disease progression in colorectal adenocarcinomas

Colorectal cancer (CRC) is highly heterogeneous, with five-year survival rates dropping from $\sim$90% in localized disease to $\sim$15% with distant metastases. Disease progression is shaped not only by tumor-intrinsic alterations but also by the reorganization of the tumor microenvironment (TME). Metabolic, compositional, and spatial changes contribute to this progression, but considered individually they lack context and often fail as therapeutic targets. Understanding their coordination could reveal processes to alter the disease course. Here, we combined multiplexed ion beam imaging (MIBI) with machine learning to profile metabolic, functional and spatial states of 522 colorectal lesions with single-cell resolution. We observed recurrent stage-specific remodeling marked by a lymphoid-to-myeloid shift, stromal-cancer cooperation, and malignant metabolic shifts. Spatial organization of epithelial, stromal, and immune compartments provided stronger stratification of disease stage than tumor-intrinsic changes or bulk immune infiltration alone. To systematically model these coordinated changes, we condensed multimodal features into 10 latent factors of TME organization. These factors tracked disease progression, were conserved across cohorts, and revealed frequent multicellular metabolic niches and distinct, non-exclusive TME trajectories. Our framework MuVIcell exposes the elements that together drive CRC progression by grouping co-occurring changes across cell types and feature classes into coordinated multicellular programs. This creates a rational basis to therapeutically target TME reorganization. Importantly, the framework is scalable and flexible, offering a resource for studying multicellular organization in other solid tumors.

q-bio.QM↗

arXiv2021

Towards Explainable End-to-End Prostate Cancer Relapse Prediction from H&E Images Combining Self-Attention Multiple Instance Learning with a Recurrent Neural Network

Clinical decision support for histopathology image data mainly focuses on strongly supervised annotations, which offers intuitive interpretability, but is bound by expert performance. Here, we propose an explainable cancer relapse prediction network (eCaReNet) and show that end-to-end learning without strong annotations offers state-of-the-art performance while interpretability can be included through an attention mechanism. On the use case of prostate cancer survival prediction, using 14,479 images and only relapse times as annotations, we reach a cumulative dynamic AUC of 0.78 on a validation set, being on par with an expert pathologist (and an AUC of 0.77 on a separate test set). Our model is well-calibrated and outputs survival curves as well as a risk score and group per patient. Making use of the attention weights of a multiple instance learning layer, we show that malignant patches have a higher influence on the prediction than benign patches, thus offering an intuitive interpretation of the prediction. Our code is available at www.github.com/imsb-uke/ecarenet.

cs.CV↗

arXiv2021

Deep learning-based bias transfer for overcoming laboratory differences of microscopic images

The automated analysis of medical images is currently limited by technical and biological noise and bias. The same source tissue can be represented by vastly different images if the image acquisition or processing protocols vary. For an image analysis pipeline, it is crucial to compensate such biases to avoid misinterpretations. Here, we evaluate, compare, and improve existing generative model architectures to overcome domain shifts for immunofluorescence (IF) and Hematoxylin and Eosin (H&E) stained microscopy images. To determine the performance of the generative models, the original and transformed images were segmented or classified by deep neural networks that were trained only on images of the target bias. In the scope of our analysis, U-Net cycleGANs trained with an additional identity and an MS-SSIM-based loss and Fixed-Point GANs trained with an additional structure loss led to the best results for the IF and H&E stained samples, respectively. Adapting the bias of the samples significantly improved the pixel-level segmentation for human kidney glomeruli and podocytes and improved the classification accuracy for human prostate biopsies by up to 14%.

eess.IV↗