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: Jan 7, 2026

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Published on: June 21, 2024

Integrating Foundation Model Features into Graph Neural Network and Fusing Predictions with Standard Fine-Tuned

Nematollah Saeidi¹, Nima Torbati¹, Ramona Woitek¹

¹Research Center for Medical Image Analysis and Artificial Intelligence, Department of Medicine, Faculty of Medicine and Dentistry, Danube Private University, 3500 Krems an der Donau, Austria.

Bioengineering (Basel, Switzerland)

|December 30, 2025

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

Special Issue on "Machine Learning-Aided Medical Image Analysis".

Bioengineering (Basel, Switzerland)·2026

Same author

Radiological assessment of Sella Turcica morphology correlates with skeletal classes in an Austrian population: an observational study.

Oral radiology·2024

Same author

NuInsSeg: A fully annotated dataset for nuclei instance segmentation in H&E-stained histological images.

Scientific data·2024

Same author

Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer.

Nature communications·2023

Same author

Integrating Artificial Intelligence Tools in the Clinical Research Setting: The Ovarian Cancer Use Case.

Diagnostics (Basel, Switzerland)·2023

Same author

Hyperpolarized Carbon-13 MRI in Breast Cancer.

Diagnostics (Basel, Switzerland)·2023

Same journal

Correction: Komatsu et al. Three-Dimensional Visualization and Detection of the Pulmonary Venous-Left Atrium Connection Using Artificial Intelligence in Fetal Cardiac Ultrasound Screening. <i>Bioengineering</i> 2026, <i>13</i>, 100.

Bioengineering (Basel, Switzerland)·2026

Same journal

Comparison of CO<sub>2</sub> Laser and Microdebrider in the Surgical Treatment of Pediatric Recurrent Respiratory Papillomatosis: A Retrospective Analysis.

Bioengineering (Basel, Switzerland)·2026

Same journal

Toward More Translational Tumor Models: Breast dECM-Based 3D Systems Capture Native Microenvironmental Cues.

Bioengineering (Basel, Switzerland)·2026

Same journal

Postural Stability Changes During the 4 Phases of the Half Squat: Kinematics Profile of the Center of Pressure and Center of Mass in High-Performance Weightlifters-A Pilot Study.

Bioengineering (Basel, Switzerland)·2026

Same journal

Definite Implant Position as Novel Readout for Effectiveness of Ridge Preservation Indicates to Beneficial Effect of Combined Treatment with Platelet-Rich Fibrin (PRF) and Xenogenic Biomaterial in Bone Regeneration.

Bioengineering (Basel, Switzerland)·2026

Same journal

Trueness and Precision of Intraoral Scanners for 3D-Printed Orthodontic Models with Attachments: An In Vitro Comparative Study.

Bioengineering (Basel, Switzerland)·2026

See all related articles

This summary is machine-generated.

Foundation models enhance graph neural networks (GNNs) for histopathological image classification, outperforming traditional methods. Combining GNNs and fine-tuned models with a prediction fusion strategy yields superior diagnostic accuracy.

Area of Science:

Computational pathology
Artificial intelligence in medicine
Machine learning for histopathology

Background:

Histopathological image classification is crucial for disease diagnosis.
Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) are common but limited by pre-training datasets.
Graph Neural Networks (GNNs) offer an alternative but require effective feature extraction.

Purpose of the Study:

To evaluate the performance of GNNs using foundation models for feature extraction in histopathology.
To compare GNNs with foundation features against fine-tuned CNNs and ViTs.
To investigate a prediction fusion strategy combining GNN and fine-tuned model outputs.

Main Methods:

Integrated foundation models as feature extractors within a lightweight GNN architecture.

Keywords:

computational pathology deep learning foundation model graph neural network image classification medical image analysis

More Related Videos

A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports

A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports

Published on: October 13, 2023

Related Experiment Videos

Last Updated: Jan 7, 2026

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Author Spotlight: Enhanced Multiplex Immunofluorescent Microscopy Protocol for Neuroscience Research

Published on: June 21, 2024

A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports

A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports

Published on: October 13, 2023

Compared GNNs with foundation features against standard fine-tuned CNN and ViT models.

Implemented a prediction fusion approach combining outputs from the best GNN and fine-tuned models.

Main Results:

GNNs with foundation model features outperformed those using CNN or ViT features.
Performance of GNNs with foundation features was comparable to standard fine-tuned CNN and ViT models.
The prediction fusion strategy achieved the highest overall performance across three datasets.

Conclusions:

Foundation models significantly improve GNN performance in histopathological image analysis.
A prediction fusion strategy combining complementary representations offers a powerful approach for enhanced classification accuracy.
The proposed methods demonstrate high efficacy on public datasets like PanNuke, BACH, and BreakHis.