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 Concept Videos

  1. Home
  3. Research Domains
  5. Information And Computing Sciences
  7. Artificial Intelligence
  9. Natural Language Processing
  11. Pathvlm-eval: Evaluation Of Open Vision Language Models In Histopathology.
  1. Home
  3. Research Domains
  5. Information And Computing Sciences
  7. Artificial Intelligence
  9. Natural Language Processing
  11. Pathvlm-eval: Evaluation Of Open Vision Language Models In Histopathology.

Related Experiment Video

Author Spotlight: Insights into Visual Cortex Research Through Wide-View fMRI Mapping
07:11

Author Spotlight: Insights into Visual Cortex Research Through Wide-View fMRI Mapping

Published on: December 8, 2023

1.9K

PathVLM-Eval: Evaluation of open vision language models in histopathology.

Nauman Ullah Gilal1, Rachida Zegour1, Khaled Al-Thelaya2

  • 1Qatar Computing Research Institute, Hamad Bin Khalifa University, Doha, Qatar.

Journal of Pathology Informatics
|July 21, 2025

View abstract on PubMed

Summary
This summary is machine-generated.

This study benchmarks over 60 vision language models (VLMs) on histopathology images using the PathMMU dataset. Qwen2-VL-72B-Instruct demonstrated superior performance, advancing AI in digital pathology diagnostics.

Keywords:
LLMs benchmarkingPathologyVLMsZero-shot evaluation

More Related Videos

In Vivo Dynamics of Retinal Microglial Activation During Neurodegeneration: Confocal Ophthalmoscopic Imaging and Cell Morphometry in Mouse Glaucoma
12:48

In Vivo Dynamics of Retinal Microglial Activation During Neurodegeneration: Confocal Ophthalmoscopic Imaging and Cell Morphometry in Mouse Glaucoma

Published on: May 11, 2015

10.7K
Utilizing 3D Printing Technology to Merge MRI with Histology: A Protocol for Brain Sectioning
15:53

Utilizing 3D Printing Technology to Merge MRI with Histology: A Protocol for Brain Sectioning

Published on: December 6, 2016

15.1K

Related Experiment Videos

Author Spotlight: Insights into Visual Cortex Research Through Wide-View fMRI Mapping
07:11

Author Spotlight: Insights into Visual Cortex Research Through Wide-View fMRI Mapping

Published on: December 8, 2023

1.9K
In Vivo Dynamics of Retinal Microglial Activation During Neurodegeneration: Confocal Ophthalmoscopic Imaging and Cell Morphometry in Mouse Glaucoma
12:48

In Vivo Dynamics of Retinal Microglial Activation During Neurodegeneration: Confocal Ophthalmoscopic Imaging and Cell Morphometry in Mouse Glaucoma

Published on: May 11, 2015

10.7K
Utilizing 3D Printing Technology to Merge MRI with Histology: A Protocol for Brain Sectioning
15:53

Utilizing 3D Printing Technology to Merge MRI with Histology: A Protocol for Brain Sectioning

Published on: December 6, 2016

15.1K

Area of Science:

  • Artificial Intelligence
  • Medical Imaging
  • Digital Pathology

Background:

  • Vision Language Models (VLMs) are advancing AI but lack domain-specific evaluations.
  • Digital pathology can greatly benefit from VLMs for histological interpretation and diagnosis.
  • Current VLM evaluations are limited to general datasets, hindering specialized applications.

Purpose of the Study:

  • To benchmark the performance of state-of-the-art VLMs on histopathology image understanding.
  • To provide an unbiased and contamination-free evaluation of VLMs using domain-specific datasets.
  • To identify top-performing VLMs for digital pathology applications and guide future research.

Main Methods:

  • Extensive zero-shot evaluation of over 60 VLMs on the PathMMU dataset.
  • Utilized VLMEvalKit for standardized and unbiased model performance assessment.
  • Included diverse subsets like PubMed, SocialPath, and EduContent featuring multiple-choice questions (MCQs).

    • Main Results:

    • Qwen2-VL-72B-Instruct achieved the highest average score of 63.97% across all PathMMU subsets.
    • Significant performance variations observed among evaluated VLMs.
    • The study expanded the range of evaluated VLMs compared to prior literature.

    Conclusions:

    • The comprehensive evaluation provides a valuable resource for developing next-generation VLMs for digital pathology.
    • Qwen2-VL-72B-Instruct shows strong potential for aiding pathologists in diagnostic reasoning.
    • The findings will foster advancements in AI-driven histopathology image analysis and healthcare services.