go-v.orgdoi

Helicobacter Pylori Detection in Immunohistochemically stained histological WSI images

Débora Gil, Pau Cano, Eva Musulén · CORA.Repositori de Dades de Recerca · 2026

  • DOI 10.34810/data3729
  • Publisher Consorci de Serveis Universitaris de Catalunya
  • Type dataset
  • Issued 2026

Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.

CORA.Repositori de Dades de Recerca · Consorci de Serveis Universitaris de Catalunya · dataset · 2026

Abstract

This database, made in collaboration with Quiron Salud, has 245 WSI, scored by a pathologist according to H. pylori density as NEGATIVE (Healthy), LOW DENSITY and HIGH DENSITY. Of the 245 patients included in the study, 117 (47.8% of the total) are classified as NEGATIVE, while 128 are classified as POSITIVE (LOW and HIGH DENSITY) for the presence of H. pylori. Each WSI contains 2 sections of several gastric mucosa samples and one section extracted from an external positive control to check the validity of the immunohistochemical staining. 1. Images 1.1 Patches Patches of size 256x256 were cropped along the border of the tissue samples of one of the sections.

Cropped images are inside a folder identified by the PatID and the tissue section as PatID_Section#. Patches can be explored at http://ivendis.cvc.uab.cat/Digipatics/, with user isbi_user and password isbi_pass. From a subset of 123 cases (77 positive ones), an expert pathologist annotated 1211 patches (161 being positive ones) extracted from the first slide. The number of positive annotated patches has been augmented by cropping windows along tissue border from the point the annotated patches were extracted from. They are labelled with the tag ‘_Aug’. The set of annotated cases was complemented with images from negative cases in order to balance the set at diagnosis level.

These subset of cases with annotated patches can be used for CrossValidation of methods, while the remaining cases are an independent HoldOut set for verification of reproducibility. 2. Folder Structure The CrossValidation folder contains patches from patients with annotations of presence of helicobacter in patches, as well as, global diagnosis of the sample. This set should be used for training and validation of models for, both, patch classification and patient diagnosis. The HoldOut folder contains patches cropped from patients that only have global diagnosis. This cases can be used to assess reproducibility.

The diagnosis of the whole set of patients is given in the file PatientDiagnosis.csv. The CrossValidation has 2 sets of patches, the Annotated ones that have identified the presence of helicobacter and the total set of patches Cropped from the tissue sample. The annotated ones are a subset of the Cropped. Annotations of presence of the bacteria in this set is given in the file HP_WSI-CoordAllAnnotatedPatches.xlsx. For positive patches, the set of augmentations extracted from neighboring pixels are also included. The folder also contains the extra negative patches extracted from healthy patients. This cases are not annotated in HP_WSI-CoordManualAnnotatedPatches.xlsx.

This excel only contains the manual annotations made by experts. Both folders contain one subfolder for each patient’s patches. Each patient folder is identified by the patient code or PatID (B22-#) and the tissue section number (1,0): PatID_#Section. The HolsOut set also follows the same distribution in subfolders containing the patches cropped for each patient. 3. Metadata 3.1 Annotated Patches We have 2 files containing information about the annotated patches: HP_WSI-CoordManualAnnotatedPatches.xlsx, HP_WSI-CoordAllAnnotatedPatches.xlsx. Both files have the same data for each patch: • Pat_ID, Section_ID, Window_ID: Patient, Section and patch identifiers.

Window_ID is the patch image filename. Pat_ID_Section_ID is the name of the folder containing the image (.png format) • i,j,h,w define the position of the patch in the .tiff image with respect the left upper corner of the image. Values are in matrix coordinates with access [i:i+h, j:j+w]. • Presence indicates whether H. pylori was observed in that patch (1) or not (-1). A 0 represents a patch that was not possible to classify. The file HP_WSI-CoordManualAnnotatedPatches.xlsx, contains information only for the patches inspected by the pathologists. The information for the whole set in the folder Annotated is given in the file HP_WSI-CoordAnnotatedAllPatches.xlsx.

The positive patches that are an augmentation of the annotated original ones are labelled with Window_ID=XXX_AugY. 3.2 Patient Diagnosis The file PatientDiagnosis.csv contains information about the presence of Helicobacter pylori on the sampled tissue. The first column, CODI, has the Pat_ID of each WSI image, and the second column indicates the degree of H. pylori presence: Negative (NEGATIVA), Low (BAIXA), or High (ALTA). The positive cases can be group into a single group for binary classification (positive/negative) of the patient diagnosis.

Open the resolved locator · 10.34810/data3729

Metadata licence: OpenAlex metadata (CC0)

Registry
OpenAlex DOI index + API
Fetched
2026-11-20 01:17 UTC
Cache state
not cached yet (bulk-index+openalex)
Locator
https://dataverse.csuc.cat/dataset.xhtml
Metadata licence
OpenAlex metadata (CC0)
Scope
Metadata as returned by the resolver; abstracts are never generated here.