XPCT Human Olfactory Bulb Dataset (XPCT-HOBD)

Data description The dataset consists of 300 consecutive slices of X-ray phase-contrast tomography (XPCT) reconstructions (2048×2048 px) of human olfactory bulb (OB). Each XPCT slice has expert-annotated mask containing: Background class. 5 classes corresponding to layers of the OB. To support data preprocessing and layer segmentation, we provide Python scripts for: Intensity normalization using fixed clipping bounds Inter-slice smoothing via 1D convolution across the slice axis Central ROI extraction (864×864 px) Inference using a trained neural network for layer segmentation This dataset is intended for research in biomedical image analysis, particularly for segmentation tasks where data is limited and reference annotations are inaccurate. Archive Contents slices/ - input TIFF images (used as --input_dir in scripts) masks/ - ground truth segmentation masks ob_layers_segmentation/: /scripts - preprocessing and inference scripts, model architectures /models - pre-trained models full_pipeline.py - script for running full pipeline README.md - data decsription and instructions Recommended project structure project_root/ data/ slices/ masks/ scripts/ model/ full_pipeline.py README.md Citation requirement An article related to this dataset is currently under review. Full reference will be provided upon official publication.