Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: May 24, 2026

Assessing Early Stage Open-Angle Glaucoma in Patients by Isolated-Check Visual Evoked Potential

Assessing Early Stage Open-Angle Glaucoma in Patients by Isolated-Check Visual Evoked Potential

Published on: May 25, 2020

Evaluating Open and Accessible Visual Language Models for Optical Character Recognition in Clinical Case Report

Giovanna Nicora¹, Marco Germanotta², Irene Giovanna Aprile²

¹Dept. of Electronic, Computer and Biomedical Engineering, University of Pavia, Italy.

Studies in Health Technology and Informatics

|May 23, 2026

Summary

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

When Robots Learn: Artificial Intelligence and the Next Human-Centered Era of Neurorehabilitation.

Brain sciences·2026

Same author

Social and Behavioral Correlates of Self-Perceived Psychological Distress in Celiac Disease During the COVID-19 Pandemic: An Exploratory Cross-Sectional Study (COVIMPACT).

Nutrients·2026

Same author

Correction: Effects of a tailored rehabilitation treatment in lower limb Soft Tissue Sarcomas reconstruction: a case series.

Journal of neuroengineering and rehabilitation·2026

Same author

Elevated microbially-derived metabolites in autism: a possible diagnostic screening test for a distinct ASD phenotype.

Molecular psychiatry·2026

Same author

ViTMARE - A Vision Transformer Pipeline for Anomaly Detection in 3D Brain MRI.

Studies in health technology and informatics·2026

Same author

Redesigning Python Programming Course in the Generative AI Era: An Italian Case Study.

Studies in health technology and informatics·2026

This summary is machine-generated.

Open-source visual language models (VLMs) show promise for automating clinical trial data entry from paper forms. However, challenges remain, particularly with extracting handwritten information.

Area of Science:

Artificial Intelligence
Clinical Research Informatics
Medical Imaging

Background:

Paper-based case report forms are prevalent in clinical trials, necessitating manual data transcription.
Current automated methods for text and mark recognition struggle with template rigidity and generalizability.
Visual Language Models (VLMs) offer a novel approach by integrating image and text processing for data extraction.

Purpose of the Study:

To benchmark the performance of three open-source, locally executable VLMs (Qwen 2.5, Mistral Small 3.1, Granite 3.2 Vision) for information extraction from clinical trial case report forms.
To evaluate the efficacy of VLMs in identifying form titles, handwritten Record IDs, handwritten dates, and marked checkboxes.
To assess the potential of VLMs in reducing workload and errors in clinical research data management.

Keywords:

LLMs Optical Mark Recognition REDCap information extraction zero-shot learning

More Related Videos

Using Optical Coherence Tomography and Optokinetic Response As Structural and Functional Visual System Readouts in Mice and Rats

Using Optical Coherence Tomography and Optokinetic Response As Structural and Functional Visual System Readouts in Mice and Rats

Published on: January 10, 2019

Related Experiment Videos

Last Updated: May 24, 2026

Assessing Early Stage Open-Angle Glaucoma in Patients by Isolated-Check Visual Evoked Potential

Assessing Early Stage Open-Angle Glaucoma in Patients by Isolated-Check Visual Evoked Potential

Published on: May 25, 2020

Using Optical Coherence Tomography and Optokinetic Response As Structural and Functional Visual System Readouts in Mice and Rats

Using Optical Coherence Tomography and Optokinetic Response As Structural and Functional Visual System Readouts in Mice and Rats

Published on: January 10, 2019

Main Methods:

Benchmarking three open-source VLMs: Qwen 2.5, Mistral Small 3.1, and Granite 3.2 Vision.
Utilizing 80 smartphone-acquired images of printed case report forms from an Italian stroke trial.
Evaluating performance on tasks including form title identification, handwritten Record ID extraction, handwritten date extraction, and checkbox detection.

Main Results:

Qwen 2.5 demonstrated superior performance in form title recognition (91%) and handwritten date extraction (75%).
Mistral Small 3.1 achieved higher accuracy in Record ID extraction (53%) and checkbox detection (80%).
Granite 3.2 Vision was the fastest but frequently failed to adhere to output formatting, and handwritten fields posed significant challenges for all models.

Conclusions:

Open-source VLMs show potential for automating information extraction from clinical trial forms, streamlining workflows.
Current VLM performance, especially for handwritten data, indicates limitations that require further development.
VLMs represent a promising but not yet fully realized solution for enhancing efficiency and accuracy in clinical research data management.