Dataset sources
Downloaded 5 October 2026. Original image pixels are retained. Source labels describe the supplied annotation; they are not HAEMARA predictions.
Barcelona PBC
17,092 cell crops across eight classes. This preview selects 200 per class, using a deterministic hash order. The archive and full manifest are retained in the delivery package.
Acevedo, Andrea; Merino, Anna; Alférez, Santiago; Molina, Ángel; Boldú, Laura; Rodellar, José (2020), “A dataset of microscopic peripheral blood cell images for development of automatic recognition systems”, Mendeley Data, V1, doi:10.17632/snkd93bnjr.1.
Original dataset · Associated paper · CC BY 4.0 licence
The paper describes research and educational use. These single-cell images have no supplied patient grouping or paired CBC results. They do not establish disease-detection performance or compatibility with a Canadian hospital scanner.
BCCD
364 smear fields with 4,888 supplied object annotations. All fields are available here. Original XML is retained in the archive. The interface excludes two zero-area boxes from drawing, without changing the source files.
Shenggan / BCCD Dataset, with upstream data and annotation credits to cosmicad and akshaylamba. Repository licence: MIT, copyright 2017 shenggan.
Original repository · Preserved MIT licence
The repository does not provide a documented clinical cohort or participant permissions. Its source history needs further review before commercial or clinical reuse. It is used here for annotation and image-navigation testing.
Bodzas WBC NEW
Nine original 1,200 × 1,200 pixel images, one per supplied class, from the 16,027-image dataset “A large-scale high-resolution WBC image dataset” by Bodzas, Kodytek and colleagues (2023). The preview uses the first archived image in each class and preserves original BMP files. This small selection is for workflow exploration; class names retain the source taxonomy.
Original dataset and CC0 licence · Associated paper · Open image workbench
NLM Malaria Screener — Sudan NEW
Courtesy of the U.S. National Library of Medicine. Two original smear images: one from the negative thin-smear group and one from the P. falciparum-positive thick-smear group. Selection uses the smallest intact file in each group; a damaged source image was excluded. Folder labels describe the source grouping and do not constitute a HAEMARA diagnosis or parasite count.
Yu H and colleagues (2023), “Patient-level performance evaluation of a smartphone-based malaria diagnostic application”, Malaria Journal. The source licence permits redistribution with its notice and acknowledgement. The original licence is included. No NLM endorsement is implied.
Original collection · Preserved source licence · Open image workbench
Use in this workspace
These public research images are kept separate from the constructed patient cases. No model has been trained or clinically validated in this work. Public availability and a reuse licence do not establish a hospital partnership, consent to a new clinical study, or permission to receive new patient data.
Reviews entered here are prototype records. There are no preloaded expert judgments. No patient information should be entered.