Jove
Visualize
Contact Us

Related Concept Videos

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
  1. Home
  3. Ilam: Cross-fusion Of Latent And Attention Features For Explainable Medical Image Classification.
  1. Home
  3. Ilam: Cross-fusion Of Latent And Attention Features For Explainable Medical Image Classification.

Related Experiment Video

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.6K

ILAM: Cross-Fusion of Latent and Attention Features for Explainable Medical Image Classification.

Anshul Sharma, Utkarsh Varman, Vandana Bharti

    IEEE Journal of Biomedical and Health Informatics
    |April 14, 2025

    View abstract on PubMed

    Summary
    This summary is machine-generated.

    We developed the Integrated Latent and Attention Mapping (ILAM) framework to improve medical image analysis. ILAM enhances artificial intelligence (AI) model accuracy and explainability by combining local and global features for better interpretability.

    More Related Videos

    Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
    04:09

    Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

    Published on: October 10, 2018

    8.2K
    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
    04:48

    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

    Published on: July 5, 2024

    321

    Related Experiment Videos

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
    04:48

    Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

    Published on: November 30, 2022

    2.6K
    Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
    04:09

    Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

    Published on: October 10, 2018

    8.2K
    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
    04:48

    Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

    Published on: July 5, 2024

    321

    Area of Science:

    • Artificial Intelligence
    • Medical Imaging
    • Computer Vision

    Background:

    • Accurate and interpretable AI models are crucial for medical image analysis.
    • Existing explainable AI (XAI) methods often suffer from inconsistent interpretability.
    • There is a need for AI frameworks that enhance both classification accuracy and explainability in medical imaging.

    Purpose of the Study:

    • To introduce the Integrated Latent and Attention Mapping (ILAM) framework.
    • To improve the accuracy and interpretability of AI models in medical image analysis.
    • To provide a robust and interpretable solution for medical image classification.

    Main Methods:

    • ILAM fuses local and global feature representations using a custom Autoencoder (AE) and Vision Transformer (ViT).
  • The AE learns fine-grained local features via unsupervised patchwise image reconstruction in the latent space.
  • A modified attention rollout mechanism refines explainability by aggregating latent features and attention weights for stable activation maps.

    • Main Results:

    • ILAM demonstrated superior performance compared to ViT, DeiT, CvT, and SwinT on BreakHis, Chest X-Ray, and Retinal datasets.
    • The framework consistently generated detailed and reliable activation maps, enhancing visualization of critical image regions.
    • ILAM achieved improved classification accuracy and post hoc interpretability.

    Conclusions:

    • The ILAM framework effectively combines local and global feature fusion for robust medical image classification.
    • ILAM offers enhanced accuracy and superior interpretability over existing transformer-based models.
    • ILAM provides a promising solution for developing reliable and understandable AI in medical image analysis.