Fabian H. Reith

Fabian H. Reith, Jannik Franzen, Dinesh R. Palli et al.

Deep neural networks have become the go-to method for biomedical instance segmentation. Generalist models like Cellpose demonstrate state-of-the-art performance across diverse cellular data, though their effectiveness often degrades on domains that differ from their training data. While supervised fine-tuning can address this limitation, it requires annotated data that may not be readily available. We propose SelfAdapt, a method that enables the adaptation of pre-trained cell segmentation models without the need for labels. Our approach builds upon student-teacher augmentation consistency training, introducing L2-SP regularization and label-free stopping criteria. We evaluate our method on the LiveCell and TissueNet datasets, demonstrating relative improvements in AP0.5 of up to 29.64% over baseline Cellpose. Additionally, we show that our unsupervised adaptation can further improve models that were previously fine-tuned with supervision. We release SelfAdapt as an easy-to-use extension of the Cellpose framework. The code for our method is publicly available at https: //github.com/Kainmueller-Lab/self_adapt.

Claudia Winklmayr, Jerome Luescher, Nora Koreuber et al.

Digital pathology has seen the advent of a wealth of foundational models (FM), yet to date their performance on cell phenotyping has not been benchmarked in a unified manner. We therefore propose PhenoBench: A comprehensive benchmark for cell phenotyping on Hematoxylin and Eosin (H&E) stained histopathology images. We provide both PhenoCell, a new H&E dataset featuring 14 granular cell types identified by using multiplexed imaging, and ready-to-use fine-tuning and benchmarking code that allows the systematic evaluation of multiple prominent pathology FMs in terms of dense cell phenotype predictions in different generalization scenarios. We perform extensive benchmarking of existing FMs, providing insights into their generalization behavior under technical vs. medical domain shifts. Furthermore, while FMs achieve macro F1 scores > 0.70 on previously established benchmarks such as Lizard and PanNuke, on PhenoCell, we observe scores as low as 0.20. This indicates a much more challenging task not captured by previous benchmarks, establishing PhenoCell as a prime asset for future benchmarking of FMs and supervised models alike. Code and data are available on GitHub.

Fabian H. Reith, Brian A. Wandell

We investigate the performance of modern convolutional neural networks (CNN) and a linear support vector machine (SVM) with respect to spatial contrast sensitivity. Specifically, we compare CNN sensitivity to that of a Bayesian ideal observer (IO) with the signal-known-exactly and noise known statistically. A ResNet-18 reaches optimal performance for harmonic patterns, as well as several classes of real world signals including faces. For these stimuli the CNN substantially outperforms the SVM. We further analyzed the case in which the signal might appear in one of multiple locations and found that CNN spatial sensitivity continues to match the IO. However, the CNN sensitivity was far below optimal at detecting certain complex texture patterns. These measurements show that CNNs can have very large performance differences when detecting the presence of spatial patterns. These differences may have a significant impact on the performance of an imaging system designed to detect low contrast spatial patterns.