mAIcrobe validation datasets

mAIcrobe dataset collection — Overview This repository aggregates several bacterial microscopy datasets used for cell-cycle analysis, morphometrics, and deep-learning segmentation. Used in mAIcrobe https://maicrobe.henriqueslab.org https://github.com/HenriquesLab/mAIcrobe Folder map All folder contain a README.md with details on the dataset, acquisition, and references. JE2_WT SIM images of Staphylococcus aureus JE2 WT (membrane + DNA channels) LCML1_LCML262 SIM images of CRISPRi dnaA knockdown (LCML1) and control (LCML262) StarDist_Saureus Training and test data for StarDist segmentation on S. aureus SIM images UNetSPneumo Training and test data for U-Net segmentation on Streptococcus pneumoniae phase-contrast images JE2_WT 31 fields of view (FOVs) of S. aureus JE2 WT imaged by structured illumination microscopy (SIM). Channels: membrane (NileRed) and DNA (Hoechst 33342). Dataset used to validate eHooke’s cell-cycle model; repurposed to validate morphometrics in mAIcrobe. Reference: Saraiva BM, Krippahl L, Filipe SR, Henriques R, Pinho MG. Biological Imaging. 2021;1:e3. doi:10.1017/S2633903X21000027. LCML1_LCML262 Scope: Two S. aureus datasets, membrane-labelled and imaged by SIM. LCML1: CRISPRi-mediated knockdown of dnaA. LCML262: Control strain with dCas9 and markers but no sgRNA. Each dataset contains 5 FOVs. Reference (strain resource): Reed P et al. 2024. mBio 15:e02773-23. https://doi.org/10.1128/mbio.02773-23 StarDist_Saureus Goal: Data for training/testing StarDist instance segmentation on S. aureus SIM membrane images (used in mAIcrobe). Training set: Conditions: WT JE2 (10 FOVs) and WT JE2 + PC190723 (12 FOVs; FtsZ inhibitor causing enlarged bacteria). Contents: Fluorescence images + corresponding instance masks (Labels match image filenames per subset). Test set: Condition: LCML262 (3 FOVs). Contents: Images + instance masks. References: Ferreira MG et al. (ReScale4DL) bioRxiv, 2025. https://doi.org/10.1101/2025.04.09.647871 Reed P et al. mBio, 2024. https://doi.org/10.1128/mbio.02773-23 UNetSPneumo Goal: Data for training/testing a U‑Net model on phase-contrast images of Streptococcus pneumoniae (Pen6; PenR, unencapsulated; mosaic pbp alleles). Trained with ZeroCostDL4Mic. Size (from README): 16 FOVs for training 3 FOVs for testing Structure (with filename parity between images and masks): Train/Phase Train/Labels Test/Phase Test/Labels Citations Saraiva BM, Krippahl L, Filipe SR, Henriques R, Pinho MG. eHooke: Automated image analysis of spherical bacteria. Biological Imaging. 2021;1:e3. doi:10.1017/S2633903X21000027 Reed P, Sorg M, Alwardt D, Serra L, Veiga H, Schäper S, Pinho MG. 2024. A CRISPRi-based genetic resource to study essential Staphylococcus aureus genes. mBio 15:e02773-23. https://doi.org/10.1128/mbio.02773-23 Ferreira MG, Saraiva BM, Brito AD, Pinho MG, Henriques R, Gómez-de-Mariscal E. ReScale4DL: Balancing Pixel and Contextual Information for Enhanced Bioimage Segmentation. bioRxiv, 2025. https://doi.org/10.1101/2025.04.09.647871