Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Sep 18, 2025

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

693

Structured Transformation of Unstructured Prostate MRI Reports Using Large Language Models.

Luca Di Palma1, Fatemeh Darvizeh1, Marco Alì1

  • 1CDI Centro Diagnostico Italiano, Saint Bon 20, 20147 Milan, Italy.

Tomography (Ann Arbor, Mich.)
|June 25, 2025
PubMed
Summary

High-performing large language models (LLMs) demonstrated strong capabilities in extracting radiological features from prostate MRI reports. DeepSeek-R1-Llama3.3 achieved the highest performance, indicating potential for improved clinical workflows.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Correction: Cè et al. Decoding Radiomics: A Step-by-Step Guide to Machine Learning Workflow in Hand-Crafted and Deep Learning Radiomics Studies. <i>Diagnostics</i> 2024, <i>14</i>, 2473.

Diagnostics (Basel, Switzerland)·2026
Same author

Whole-body magnetic resonance imaging for cancer screening in asymptomatic adults: a multicenter study.

European journal of cancer prevention : the official journal of the European Cancer Prevention Organisation (ECP)·2026
Same author

XGeM: A multi-prompt foundation model for multimodal medical data generation.

Computerized medical imaging and graphics : the official journal of the Computerized Medical Imaging Society·2026
Same author

Biomarkers.

Alzheimer's & dementia : the journal of the Alzheimer's Association·2026
Same author

Quantifying energy savings in radiology: a simple approach to make the radiology department more sustainable.

BMJ leader·2025
Same author

The "iodine-derived perfusion" model applied to abdominal tissues with high arterial vascularization: pancreas, kidney cortex, and hepatocellular carcinoma.

The British journal of radiology·2025

Area of Science:

  • Radiology and Medical Imaging
  • Artificial Intelligence in Healthcare
  • Natural Language Processing

Background:

  • Prostate MRI reports contain crucial information for patient diagnosis and management.
  • Extracting key radiological features from unstructured text reports is time-consuming and prone to human error.
  • Large Language Models (LLMs) offer a potential solution for automated information extraction.

Purpose of the Study:

  • To evaluate the performance of open-weight LLMs in extracting specific radiological features from prostate MRI reports.
  • To compare the efficacy of different LLMs in identifying dimensions, volume, PSA density, and lesion characteristics.
  • To assess the impact of radiologist reporting variations on LLM performance.

Main Methods:

  • Five advanced LLMs (Llama3.3, DeepSeek-R1-Llama3.3, Phi4, Gemma-2, Qwen2.5-14B) were utilized.
Keywords:
LLMMRIinformation extractionstructured reportunstructured medical text

More Related Videos

A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
06:08

A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound

Published on: March 21, 2025

384
Use of MRI-ultrasound Fusion to Achieve Targeted Prostate Biopsy
09:11

Use of MRI-ultrasound Fusion to Achieve Targeted Prostate Biopsy

Published on: April 9, 2019

21.7K

Related Experiment Videos

Last Updated: Sep 18, 2025

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

693
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
06:08

A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound

Published on: March 21, 2025

384
Use of MRI-ultrasound Fusion to Achieve Targeted Prostate Biopsy
09:11

Use of MRI-ultrasound Fusion to Achieve Targeted Prostate Biopsy

Published on: April 9, 2019

21.7K
  • LLMs processed 250 free-text prostate MRI reports, with each model running three extraction prompts.
  • Manual annotation by an experienced radiologist served as the ground truth for performance assessment.

    • Main Results:

    • DeepSeek-R1-Llama3.3 achieved the highest average feature-level performance (98.6%).
    • All models demonstrated excellent performance in extracting PSA density (100%) and volume (≥98.4%).
    • Extraction of lesion characteristics showed more variability (88.4-94.0%), and performance differed across radiologists' reports.

    Conclusions:

    • Open-weight LLMs show significant promise for automated feature extraction from prostate MRI reports.
    • DeepSeek-R1-Llama3.3 emerged as the top-performing model in this evaluation.
    • Tailoring prompts to individual physician reporting styles is recommended to optimize LLM accuracy and clinical workflow integration.