PARROT, an open multilingual radiology reports dataset

<h2>Abstract</h2><h3>Aims</h3> To develop and validate PARROT (Polyglottal Annotated Radiology Reports for Open Testing), a multicentric, open-access dataset of fictional radiology reports spanning multiple languages for testing natural language processing applications in radiology. <h3>Methods</h3> From May to September 2024, radiologists were invited to contribute fictional radiology reports following their standard reporting practices. Contributors provided at least 20 reports with associated metadata including anatomical region, imaging modality, clinical context, and for non-English reports, English translations. All reports were assigned ICD-10 codes. A human vs. AI report differentiation study was conducted with 154 participants (radiologists, healthcare professionals, and non-healthcare professionals) assessing whether reports were human-authored or AI-generated. <h3>Results</h3> The dataset comprises 2,658 radiology reports from 76 authors across 21 countries and 13 languages. Reports cover multiple imaging modalities (CT: 36.1%, MRI: 22.8%, radiography: 19.0%, ultrasound: 16.8%) and anatomical regions, with chest (19.9%), abdomen (18.6%), head (17.3%), and pelvis (14.1%) being most prevalent. In the differentiation study, participants achieved 53.9% accuracy (95% CI: 50.7%-57.1%) in distinguishing between human and AI-generated reports, with radiologists performing significantly better (56.9%, 95% CI: 53.3%-60.6%, p<0.05) than other groups. <h3>Conclusion</h3> PARROT represents the largest open multilingual radiology report dataset, enabling testing and validation of natural language processing applications across linguistic, geographic, and clinical boundaries without privacy constraints.

Authors

Institutions

Publication Details

Journal
European Journal of Radiology Artificial Intelligence
Published
2025-12-22
DOI
https://doi.org/10.1016/j.ejrai.2025.100066
Citations
1
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
article
Field-Weighted Citation Impact
0.65
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

PARROT, an open multilingual radiology reports dataset

Quentin Holay, Keno K. Bressem, Arnaldo Stanzione, Rafał Poręba et al.
1 citations
European Journal of Radiology Artificial Intelligence
Artificial Intelligence in Healthcare and Education
0.65
article

PARROT, an open multilingual radiology reports dataset

Quentin Holay, Keno K. Bressem, Arnaldo Stanzione, Rafał Poręba, Natalia Gorelik, Raphaël Bentegeac, Andrzej Wysocki, Elif Can, Katarzyna Resler, Barbara Dziadkowiec-Macek, Federica Piccione, Aymen Meddeb, Tristan Lemke, V. Rafailidis, L Saba, Eugen Divjak, S. L. Zhang, Radia Ait Chalal, Sebastian Ziegelmayer, Felix Busch, Tugba Akinci D’Antonoli, Ali Fuat Tekin, Martyna Gachowska, Maximilian Lindholz, Claudia Fossataro, Kokou Adambounou, Federica Fossataro, Małgorzata Poręba, Feno Hasina Rabemanorintsoa, C. Chavihot, Claudio Benzoni, Salvatore Claudio Fanni, Grégory Kuchcinski, L Müller, Renato Cuocolo, Gordana Ivanac, Dominik Krupka, P. Amouyel, J PRUVO, Marco Gatti, Thibault Agripnidis, Michail E. Klontzas, Johan Wassélius, Armando Ugo Cavallo, Liz Toapanta‐Yanchapaxi, Piotr Macek, Paweł Gać, Adriana Ojeda, Federica Vernuccio, Matthaios Triantafyllou, Philipp Prucker, Krzysztof Kraik, Sung Soo Ahn, Lina Xu, A Blandino, Henrik Andersson, Ignacio García-Juárez, Weilang Wang, Carlos Ferrarotti, Luigi Mannacio, Yae Won Park, Ezann Siebert, Nicolas Fanantenana Herinirina, Evangelia E Vassalou, Alexia Maria Goulianou, Armel Elogne, Anna Kompanowska, Beatrice Nguema Edzang, Jan Rotkegel, Ekaterini Tsaoulia, Rosa Alba Pugliesi, Riccardo Cau, Marcus Makowski, Szymon Włodarczak, Adrian Włodarczak, Tomasz Zatoński, Felipe Kitamura, Bastien Le Guellec, Aghiles Hamroun, Lisa C. Adams, Michał Fułek, Rafał Kompanowski, Alexandre Lefèvre, Andrea Ponsiglione, Antonio Natale, Katarzyna Fułek, Szymon Urban, Erwin Chiquete
article en
1 citations

Abstract

<h2>Abstract</h2><h3>Aims</h3> To develop and validate PARROT (Polyglottal Annotated Radiology Reports for Open Testing), a multicentric, open-access dataset of fictional radiology reports spanning multiple languages for testing natural language processing applications in radiology. <h3>Methods</h3> From May to September 2024, radiologists were invited to contribute fictional radiology reports following their standard reporting practices. Contributors provided at least 20 reports with associated metadata including anatomical region, imaging modality, clinical context, and for non-English reports, English translations. All reports were assigned ICD-10 codes. A human vs. AI report differentiation study was conducted with 154 participants (radiologists, healthcare professionals, and non-healthcare professionals) assessing whether reports were human-authored or AI-generated. <h3>Results</h3> The dataset comprises 2,658 radiology reports from 76 authors across 21 countries and 13 languages. Reports cover multiple imaging modalities (CT: 36.1%, MRI: 22.8%, radiography: 19.0%, ultrasound: 16.8%) and anatomical regions, with chest (19.9%), abdomen (18.6%), head (17.3%), and pelvis (14.1%) being most prevalent. In the differentiation study, participants achieved 53.9% accuracy (95% CI: 50.7%-57.1%) in distinguishing between human and AI-generated reports, with radiologists performing significantly better (56.9%, 95% CI: 53.3%-60.6%, p<0.05) than other groups. <h3>Conclusion</h3> PARROT represents the largest open multilingual radiology report dataset, enabling testing and validation of natural language processing applications across linguistic, geographic, and clinical boundaries without privacy constraints.

European Journal of Radiology Artificial IntelligenceVol. 5
Università Cattolica del Sacro Cuore (IT), University of Pisa (IT), Bunker Hill Community College (US), University of Salerno (IT), University of Crete (GR), Inserm (FR), University of Freiburg (DE), Yonsei University (KR), Johannes Gutenberg University Mainz (DE), Aristotle University of Thessaloniki (GR), Université de Lille (FR), Libreville Hospital (GA), University Hospital Dubrava (HR), Institut Pasteur de Lille (FR), AHEPA University Hospital (GR), McGill University Health Centre (CA), Sir Charles Gairdner Hospital (AU), University Hospital of Basel (CH), TUM Klinikum (DE), Omar Bongo University (GA), University of Toamasina (MG), Wroclaw Medical University (PL), National University of Rosario (AR), Assistance Publique Hôpitaux de Marseille (FR), Lille’s Cardiology Hospital (FR), Wodociągi Kłodzkie (Poland) (PL), University Medical Center of the Johannes Gutenberg University Mainz (DE), Université d’Antsiranana (MG), Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán (MX), Hôpital Roger Salengro (FR), Istanbul Metropolitan Municipality (TR), Zhongda Hospital Southeast University (CN), Dresden International University (DE), Istituto Dermopatico dell'Immacolata (IT), Société de Développement des Forêts (CI), Azienda Ospedaliero-Universitaria Cagliari (IT), University Children’s Hospital Basel (CH), Wroclaw University of Health and Sport Sciences (PL), Skåne University Hospital (SE), University Hospital of Heraklion (GR), Centro de Educación Médica e Investigaciones Clínicas Norberto Quirno (AR), Hôpital d'Instruction des Armées Sainte-Anne (FR), Istituti di Ricovero e Cura a Carattere Scientifico (IT), Centre Hospitalo-Universitaire Bab El Oued (DZ), ASST Fatebenefratelli Sacco (IT), Mossakowski Medical Research Institute, Polish Academy of Sciences (PL), University of Lomé (TG), İstanbul Başakşehir Çam ve Sakura Şehir Hastanesi, University of Turin (IT), Technical University of Munich (DE), University of Naples Federico II (IT), Charité - Universitätsmedizin Berlin (DE), Universidade Federal de São Paulo (BR), University of Palermo (IT)
Quality Education
Openalex Percentile: Top 18%
Artificial Intelligence in Healthcare and Education
0.65
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.