
OverviewThe FOCI dataset contains 446 fluorescence microscopy images of irradiated human lymphocyte nuclei stained for γ-H2AX, a biomarker of DNA double-strand breaks (DSBs). The dataset is designed for object counting and density estimation tasks in biomedical image analysis. Image acquisition and annotation were performed by the Bundesamt für Strahlenschutz, while preprocessing was carried out by the Umweltbundesamt. Data AcquisitionLymphocytes were isolated from whole blood samples of healthy donors, irradiated, and stained with γ-H2AX antibodies following the protocol described in Bucher et al. (2021). Images are provided in TIFF format with a resolution of 288×288 pixels, center-cropped to 220×220 pixels to remove preprocessing artifacts. Biological Contextγ-H2AX foci represent sites of radiation-induced DNA double-strand breaks. These appear as red fluorescent spots within blue-stained nuclei. Foci typically exhibit oval shapes with diffuse boundaries, with increasing overlap and signal diffusion at higher radiation doses. Annotations Bounding boxes for each detected focus Dot annotations derived from bounding box centers Only large, bright foci were annotated; small or dim signals were excluded Label Format xywh: YOLO format: <class_id> <x_center> <y_center> <width> <height> (normalized coordinates, single class “foci”) dots: JSON dot annotations with absolute pixel coordinates for each object center (center), plus a count field, suitable for single-class density map generation. Annotation UncertaintyAnnotation is subject to variability due to: ambiguous focus definition (intensity/size thresholds) overlapping foci non-symmetric shapes manual placement variability These uncertainties should be considered when training and evaluating models. Intended Use Object counting (cellular foci) Density map estimation
