Overview

Schema Version

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

Name

Multisite biparametric prostate MRI dataset for UDA-based lesion detection

Link

https://doi.org/10.1148/ryai.230521

Indexing

Keywords: prostate cancer, biparametric MRI, diffusion-weighted imaging, ADC, PI-RADS, b value, unsupervised domain adaptation, multisite

Content: GU, MR, RS

RadLex: RID12698, RID49502, RID50291, RID50317, RID50313, RID50307

Author(s)

Hao Li

Han Liu

Heinrich von Busch

Robert Grimm

Henkjan Huisman

Angela Tong

David Winkel

Tobias Penzkofer

Ivan Shabunin

Moon Hyung Choi

Qingsong Yang

Dieter Szolar

Steven Shea

Fergus Coakley

Mukesh Harisinghani

Ipek Oguz

Dorin Comaniciu

Ali Kamen

Bin Lou

Funding

Authors declared no funding for this work.

Ethical review

Study approved by local ethics committees at all participating institutions; written informed consent obtained or waived as appropriate.

Comments

Retrospective multicohort bpMRI dataset aggregated from nine clinical sites; includes lesion-based PI-RADS information and voxel-level lesion annotations. Vendor-provided ADC maps were excluded; ADC and DWI B-2000 were recomputed in a standardized manner.

Date

Published: 2024-08-21

References

[1] Li H, Liu H, von Busch H, Grimm R, Huisman H, Tong A, Winkel D, Penzkofer T, Shabunin I, Choi MH, Yang Q, Szolar D, Shea S, Coakley F, Harisinghani M, Oguz I, Comaniciu D, Kamen A, Lou B. "Deep Learning–based Unsupervised Domain Adaptation via a Unified Model for Prostate Lesion Detection Using Multisite Biparametric MRI Datasets". Radiology: Artificial Intelligence. 2024-08-21. doi:10.1148/ryai.230521. PMID: 39166972. PMCID: PMC11449150.

Dataset

Motivation

Assess whether harmonization of ADC and high-b-value DWI across varying acquisition protocols improves supervised prostate lesion detection and mitigates domain shift.

Sampling

All eligible bpMRI cases meeting inclusion/exclusion criteria from nine centers; all qualified b-value pairs per case included for methods using multiple domains.

Partitioning scheme

Training and independent testing splits by case; additional stratification by b-value groups for analysis.

Missing information

Public availability, file formats, spatial resolution, and de-identification specifics are not reported.

Relationships between instances

Multiple samples per case created from different low/high b-value pairs; each pair treated as a unique sample/domain.

External data

For t-SNE visualization, 100 cases from the PROSTATEx Challenge dataset were used as reference domain comparison.

Confidentiality

Retrospective clinical imaging data with lesion annotations; ethics approvals obtained.

Sensitive data

Contains health-related imaging and lesion annotations.