Mu-Reg Challenge: Paired MR and TRUS Prostate Images

The mu-Reg challenge dataset provides over a hundred paired magnetic resonance imaging (MRI) and transrectal ultrasound (TRUS) images for prostate biopsy and focal therapy applications. This multimodal dataset is curated from real-world clinical practice and includes anatomical and pathological landmarks annotated by experts with more than 15 years of experience. It serves as a benchmark for developing and evaluating rigid, nonrigid, feature-based, and intensity-based image registration algorithms.

Use Cases

• Benchmarking rigid versus nonrigid registration algorithms based on paired MR-TRUS images.
• Training unsupervised or segmentation-driven learning models for prostate image fusion.
• Validating registration accuracy using expert-annotated anatomical and pathological landmarks.
• Researching biomechanical regularization and statistical motion modeling for surgical planning.

Strengths

• Contains more than a hundred paired MR and TRUS images, providing a sizable clinical cohort.
• Includes expert annotations from clinicians with over 15 years of domain experience in urology, radiology, and pathology.
• Curated from real-world clinical data, addressing a scarcity of high-quality, publicly-accessible multimodal prostate imaging.

Limitations

• Specific column names, row counts, file sizes, and last update dates are not provided in the available metadata.
• The dataset scale, while described as over a hundred pairs, lacks precise patient or image counts.
• Data freshness is unknown as no last updated information is available.

Provenance

Source

Zachary M. C. Baum, University College London

Collection Method

Collected from clinical practice for the mu-Reg challenge, with expert annotation.