text stringlengths 0 5.79k |
|---|
BRACS: A Dataset for BReAst Carcinoma Subtyping |
in H&E Histology Images |
Nadia Brancati1* , Anna Maria Anniciello2 , Pushpak Pati3,5 , Daniel Riccio1,4 , Giosuè |
Scognamiglio2 , Guillaume Jaume3,6 , Giuseppe De Pietro1 , Maurizio Di Bonito2 , Antonio |
Foncubierta3 , Gerardo Botti2 , Maria Gabrani3 , Florinda Feroce2 , and Maria Frucci1 |
1 Institute for High Performance Computing and Networking of the Research Council of Italy, ICAR-CNR, Naples, |
Italy |
arXiv:2111.04740v1 [q-bio.QM] 8 Nov 2021 |
2 National Cancer Institute – IRCCS – Fondazione Pascale, Naples, Italy |
3 IBM Research – Zurich, Switzerland |
4 University of Naples Federico II, Naples, Italy |
5 ETH Zurich, Switzerland |
6 EPFL Lausanne, Switzerland |
* Corresponding author: nadia.brancati@cnr.it |
ABSTRACT |
Breast cancer is the most commonly diagnosed cancer and registers the highest number of deaths for women with cancer. |
Recent advancements in diagnostic activities combined with large-scale screening policies have significantly lowered the |
mortality rates for breast cancer patients. However, the manual inspection of tissue slides by pathologists is cumbersome, |
time-consuming, and is subject to significant inter- and intra-observer variability. Recently, the advent of whole-slide scanning |
systems have empowered the rapid digitization of pathology slides, and enabled to develop digital workflows. These progress |
further enable to leverage Artificial Intelligence (AI) to assist, automate, and augment pathological diagnosis. But AI techniques, |
especially Deep Learning (DL), require a large amount of high-quality annotated data to learn from. Constructing such |
task-specific datasets poses several challenges, such as, data-acquisition level constrains, time-consuming and expensive |
annotations, and anonymization of patient information, etc. In this paper, we introduce the BReAst Carcinoma Subtyping |
(BRACS) dataset, a large cohort of annotated Hematoxylin & Eosin (H&E)-stained images to advance the characterization |
of breast lesions. BRACS contains 547 Whole-Slide Images (WSIs), and 4539 Regions of Interest (RoIs) extracted from the |
WSIs. Each WSI, and respective RoIs, are annotated by the consensus of three board-certified pathologists into different lesion |
categories. Specifically, BRACS includes three lesion types, i.e., benign, malignant and atypical, which are further subtyped |
into seven categories. The included RoIs exhibit large variability in dimensions, and incorporate the usual tissue-preparation |
and staining artifacts to bestow a realistic breast cancer diagnosis. It is, to the best of our knowledge, the largest annotated |
dataset for breast cancer subtyping both at WSI- and RoI-level. Further, by including the understudied atypical lesions, BRACS |
offers an unique opportunity for leveraging AI to better understand their characteristics. We encourage AI practitioners to |
develop and evaluate novel algorithms on the BRACS dataset to further breast cancer diagnosis and patient care. |
Background & Summary |
Histology images contain both complex and ambiguous information, thus challenging pathologists to perform a robust, |
reproducible and efficient analysis. Further, histology images are very large, which makes their analysis cumbersome and time- |
consuming. With advances in Computer-Aided-Diagnosis (CAD), AI techniques, especially Machine Learning (ML) and DL, |
have the potential to address the aforementioned bottlenecks1–4 . These techniques can identify discriminative morphological |
patterns from large datasets to diagnose histology images in a standardized and objective manner. However, there exist several |
challenges in adopting such techniques in digital pathology, such as, (i) the requirement of large annotated datasets, (ii) the need |
for sufficiently variable data to set up cross-patient experiments, (iii) the inclusion of diagnostically challenging lesions, that are |
generally difficult and expensive to acquire, (iv) the utilization of sub-region annotations to delineate RoI, (v) the coverage of |
diagnostic spectrum, and (vi) coping with data leakage and noisy annotations. Recent advancements in DL have demonstrated |
superior capabilities compared to classical ML approaches for CAD5–11 . The crucial advantage of DL approaches is their ability |
to learn task-specific salient features directly from the training data. However, this superiority comes at the cost of acquiring |
large, high-quality, variable, and unbiased annotated training datasets. Although several datasets for diagnosing breast histology |
images exist12–16 , they do not meet all the aforementioned criteria. For instance, some datasets focus on specific diseases that |
include only binary classes12, 14 , while others handling multiple classes13, 15 include only a small number of training samples |
(both at WSI- and RoI-level) collected from a few patients, thus limiting the dataset variability. Further, the set of considered |
classes in a dataset is crucial. Most of the public datasets aim to categorize lesions into benign and malignant classes, which do |
not depict the complete spectrum of classes in breast cancer diagnosis. Many of these datasets contain standardized images |
without clinical artifacts, e.g., staining anomalies, ink marks, tissue folding, blurred regions, tears etc. Consequently, these |
datasets do not comprehensively represent the real-world breast cancer diagnosis. Thus, it is necessary to develop a breast |
cancer dataset consisting of heterogeneous images across the diagnostic spectrum which is comparable to real-world diagnosis |
performed by the pathologists. |
To this end, we introduce BRACS, a large cohort of H&E-stained images to advance CAD of breast lesions. BRACS |
features the following advantages over the extant breast cancer image datasets, (i) it includes a large and heterogeneous set of |
realistic breast histology images (both at WSI- and RoI-level), (ii) RoIs range over variable dimensions by entirely including |
the diagnostic lesion, thus avoiding the loss of diagnostically relevant information, (iii) the images are acquired from a large |
number of patients encompassing large variability, and (iv) two atypical lesion categories, also known as precancerous lesions, |
are included along with other categories. In particular, we consider the following lesion types, Normal (N), Pathological Benign |
(PB), Usual Ductal Hyperplasia (UDH), Flat Epithelial Atypia (FEA), Atypical Ductal Hyperplasia (ADH), Ductal Carcinoma |
in Situ (DCIS), and Invasive Carcinoma (IC). Thus, BRACS represents a more realistic benchmark for breast cancer diagnosis |
by including several types of typical and atypical tissue samples over a wide variety of WSIs and RoIs extracted from a large |
number of patients. |
Methods |
The BRACS dataset is created to support the development of breast cancer diagnostic methods through the automatic analysis |
of histology images. The dataset was developed through the collaboration of the National Cancer Institute - Scientific Institute |
for Research, Hospitalization and Healthcare (IRCCS) "Fondazione G. Pascale", the Institute for High Performance Computing |
and Networking (ICAR) of National Research Council (CNR), and IBM Research – Zurich. The dataset was acquired from |
patients between 2019 and 2020, by board-certified pathologists of the Department of Pathology at the National Cancer Institute |
- IRCCS "Fondazione G. Pascale" in Naples (Italy). The samples were generated from H&E-stained breast tissue biopsy slides, |
and were selected based on the diagnostic reports of the patients. The age of the patients range from 16 to 86 years, with about |
61% of patients in the range of 40-60 years, and only a few patients aging less than 20 years or above 80 years. |
WSI- and RoI-level annotations |
The curation of rich and comprehensively annotated histology images is a complex and time-consuming task, while being prone |
to observer variability. The inclusion of atypical breast lesions at both WSI- and RoI-level further increases the task complexity |
by requiring annotations of specialized expert pathologists. Moreover, a WSI typically includes several lesions of different |
subtypes. To address the aforementioned challenges, we started by extracting and annotating RoIs in WSIs, and subsequently |
derived the WSI-level label as the most severe cancerous lesion detected within the slide. Specifically, the RoI-level annotations |
were conducted in a two-step procedure. First, a set of representative RoIs in each WSI was identified. Three board-certified |
pathologists independently annotated the RoIs, i.e., either as normal tissue or as one of the six lesion subtypes. Each extracted |
RoI corresponds to a unique category, and can include single or multiple glandular structures. Then, the annotations with |
disagreement were further discussed and re-annotated by the consensus of three pathologists. This process ensures reliable |
annotations, and allows to alleviate inter- and intra-observer variability. Figure 1 presents the annotation procedure for a sample |
End of preview. Expand in Data Studio
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Thesis data, release v4.2 (canonical)
The only data version for the thesis "Agentic Patient Similarity Retrieval". Hosted at https://huggingface.co/datasets/thanminh01/thesis-data and kept as data/ at the repo root (gitignored). Read-only: fixes ship as a new release, never as an edit. What changed from v4.1: CHANGES-v4.2.md. Rules: CONTRACT.md.
Contents
| Dataset | Bank (confirmed cases) | Queries | Label space |
|---|---|---|---|
tcga-nsclc |
842 (LUAD 427, LUSC 415) | 201 | LUAD vs LUSC |
bracs |
460 (train+val) | 87 (test) | fine: 7 classes; coarse: 3 classes |
ebr-coarse30 |
2,019 | 300 | 30 classes |
Vectors: TCGA 1,043, BRACS 547, EBR 2,319 slides. 675 queries in total (588 distinct slides). Exploratory results only: BRACS prompt work sees the test queries.
data/
<dataset>/ tcga-nsclc | bracs | ebr-coarse30
runtime/catalog.sqlite the ONLY thing the agent may open: label-blind queries, bank labels and bank documents
runtime/manifest.json catalog hash and size, release
bank.sqlite bank slides with labels and text (owner/builder use)
scorer/ query.sqlite, query_labels.csv: SCORER ONLY, never inside the agent
latents/<slide>.pt PRISM2 latents, 256x2560 bf16 (used by the morphology tool)
manifest.json
bracs/slides.csv all 547 slides with split (train 395 / val 65 / test 87) and labels
knowledge/ knowledge base, sources, gold questions (the runtime copy lives in APS-WSI/knowledge/)
docs/ rubric, scorer and build reports
CONTRACT.md CHANGES-v4.2.md SHA256.json
Get it
hf auth login # needs read access if the repo is private
hf download thanminh01/thesis-data --repo-type dataset --local-dir <repo>/data
About 5 GB, mostly latents/. Check integrity: every file's sha256 is in SHA256.json (paths relative to data/).
Use it
export APS_DATASETS_ROOT=<repo>/data # or pass --root ../data
cd APS-WSI
python -c "from pathlib import Path; from agent import preflight; from agent.config import Config; \
print(preflight.run(Config(datasets_root=Path('../data'), dataset='bracs', view='fine', provider='nim', model='m')))" # [] means clean
python -m agent.run_cli --root ../data --dataset bracs --view fine --model nvidia/nemotron-3-super-120b-a12b --finish-guard --no-llava-texts
python -m agent.scorer_cli baselines --root ../data --dataset bracs --view fine # scorer only
Rules that keep the benchmark honest:
- The agent reads
<dataset>/runtime/only.scorer/,latents/,docs/,bank.sqliteand the top-level manifest are denied byagent/runtime_paths.py. Open catalogs read-only (?immutable=1) or on a copy. - Queries carry no label and no stored text. Patient id is not used anywhere.
- WSI-LLaVA bank texts stay off (
--no-llava-texts). - Source-data documents (EBR
clinical_meta, BRACSroi_distribution) can be re-checked against the raw sheets withDocuments/plans/2026-10-02-final-run-autofix/verify-source-documents.py ANNOTATION_CSV BRACS_XLSX. - Knowledge text is checked against second sources, not clinically validated.
Archived (not here)
Older releases (v4.1, the v4 layout) are in the local archive/old-data-versions/ and on the Thesis drive (Thesis-datasets/releases/).
- Downloads last month
- 56