Mathias Öttl

Sonja Kunzmann, Mathias Öttl, Prathmesh Madhu et al.

Image annotation is one essential prior step to enable data-driven algorithms. In medical imaging, having large and reliably annotated data sets is crucial to recognize various diseases robustly. However, annotator performance varies immensely, thus impacts model training. Therefore, often multiple annotators should be employed, which is however expensive and resource-intensive. Hence, it is desirable that users should annotate unseen data and have an automated system to unobtrusively rate their performance during this process. We examine such a system based on whole slide images (WSIs) showing lung fluid cells. We evaluate two methods the generation of synthetic individual cell images: conditional Generative Adversarial Networks and Diffusion Models (DM). For qualitative and quantitative evaluation, we conduct a user study to highlight the suitability of generated cells. Users could not detect 52.12% of generated images by DM proofing the feasibility to replace the original cells with synthetic cells without being noticed.

Frauke Wilm, Luis Carlos Rivera Monroy, Mathias Öttl et al.

Accurate detection of Plasmodium falciparum in Giemsa-stained blood smears is an essential component of reliable malaria diagnosis, especially in developing countries. Deep learning-based object detection methods have demonstrated strong potential for automated Malaria diagnosis, but their adoption is limited by the scarcity of datasets with detailed instance-level annotations. In this work, we present an enhanced version of the publicly available NIH malaria dataset, with detailed bounding box annotations in COCO format to support object detection training. We validated the revised annotations by training a Faster R-CNN model to detect infected and non-infected red blood cells, as well as white blood cells. Cross-validation on the original dataset yielded F1 scores of up to 0.88 for infected cell detection. These results underscore the importance of annotation volume and consistency, and demonstrate that automated annotation refinement combined with targeted manual correction can produce training data of sufficient quality for robust detection performance. The updated annotations set is publicly available via Zenodo: https://doi.org/10.5281/zenodo.17514694