Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

Abdominal ultrasound images of focal liver lesions (Necker Hospital and Saint Louis Hospital, Paris; 2014-2019)

Link

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9152842/

Indexing

Keywords: Ultrasound, Abdomen/GI, Liver, Tissue Characterization, Computer-aided Diagnosis, Supervised Learning, Transfer Learning, Convolutional Neural Network, Detection Transformer, Faster R-CNN

Content: GI, US, OI

RadLex: RID30192, RID29602, RID10920, RID31739

SNOMED: 278527001, 109841003, 85057007, 25370001

Author(s)

Hind Dadoun

Anne-Laure Rousseau

Eric de Kerviler

Jean-Michel Correas

Anne-Marie Tissier

Fanny Joujou

Sylvain Bodard

Kemel Khezzane

Constance de Margerie-Mellon

Hervé Delingette

Nicholas Ayache

Organization(s)

Université Côte d’Azur, Inria, Epione Team, Sophia Antipolis, France

Department of Vascular Surgery, Georges Pompidou European Hospital APHP, Université de Paris, Paris, France

NHance.ngo, Saint Germain en Laye, France

Department of Radiology, Hôpital Saint Louis APHP, Université de Paris, Paris, France

Department of Adult Radiology, Université de Paris and Université de l’Hôpital Necker, Paris, France

Funding

Supported in part by the French government through the 3IA Côte d’Azur Investments in the Future project managed by the National Research Agency (ANR) (reference no. ANR-19-P3IA-0002).

Ethical review

Institutional review board–approved retrospective study (IRB00011591); informed consent waived.

Comments

Retrospective multicenter dataset of de-identified B-mode abdominal ultrasound images used to train and evaluate deep learning models for detection, localization, and characterization of focal liver lesions.

Date

Published: 2022

References

[1] Dadoun H, Rousseau AL, de Kerviler E, Correas JM, Tissier AM, Joujou F, Bodard S, Khezzane K, de Margerie-Mellon C, Delingette H, Ayache N. "Deep Learning for the Detection, Localization, and Characterization of Focal Liver Lesions on Abdominal US Images". Radiology: Artificial Intelligence. 2022. doi:10.1148/ryai.210110. PMID: 35652113. PMCID: PMC9152842.

Dataset

Motivation

To develop and evaluate deep learning methods to assist nonexpert caregivers in detecting, localizing, and characterizing focal liver lesions on noncontrast abdominal ultrasound.

Sampling

Adults ≥18 years. Inclusion for lesion cases required visible lesion(s), no prior local therapy, and definite pathologic diagnosis (cyst, angioma, focal nodular hyperplasia, adenoma, metastasis, or hepatocellular carcinoma). Non-lesion cases required definite absence of pathologic diagnosis.

Partitioning scheme

2014–2018 images for training/development (random 80%/20% split with similar class proportions); 2019 images for held-out test set from patients not present in 2014–2018.

Missing information

Demographics were not retained; exact per-class counts by split not reported.

Relationships between instances

Multiple images per patient; images within the same patient are not independent.

External data

Final diagnoses were assigned by cross-referencing with other imaging modalities (contrast-enhanced US, CT, or MRI) and biopsy when available; these external data were used for ground truth but are not part of the dataset.

Confidentiality

All images and clinical reports were de-identified within the centers; no demographic information was retained.

Re-identification

DICOM identifying data and metadata removed; upper information band cropped; on-image text/graphics removed and inpainted.

Sensitive data

Medical imaging data from clinical PACS; de-identified before use.