Enhancing interpretability and bias control in deep learning models for medical image analysis using generative AI

Carlos Minutti-Martinez; Boris Escalante-Ramírez; Jimena Olveres

doi:10.1117/12.3022263

18 June 2024 Enhancing interpretability and bias control in deep learning models for medical image analysis using generative AI

Carlos Minutti-Martinez, Boris Escalante-Ramírez, Jimena Olveres

Author Affiliations +

Proceedings Volume 12998, Optics, Photonics, and Digital Technologies for Imaging Applications VIII; 1299806 (2024) https://doi.org/10.1117/12.3022263
Event: SPIE Photonics Europe, 2024, Strasbourg, France

Abstract

Explainability and bias mitigation are crucial aspects of deep learning (DL) models for medical image analysis. Generative AI, particularly autoencoders, can enhance explainability by analyzing the latent space to identify and control variables that contribute to biases. By manipulating the latent space, biases can be mitigated in the classification layer. Furthermore, the latent space can be visualized to provide a more intuitive understanding of the model’s decision-making process. In our work, we demonstrate how the proposed approach enhances the explainability of the decision-making process, surpassing the capabilities of traditional methods like Grad-Cam. Our approach effectively identifies and mitigates biases in a straightforward manner, without necessitating model retraining or dataset modification, showing how Generative AI has the potential to play a pivotal role in addressing explainability and bias mitigation challenges, enhancing the trustworthiness and clinical utility of DL-powered medical image analysis tools.

Conference Presentation

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Carlos Minutti-Martinez, Boris Escalante-Ramírez, and Jimena Olveres "Enhancing interpretability and bias control in deep learning models for medical image analysis using generative AI", Proc. SPIE 12998, Optics, Photonics, and Digital Technologies for Imaging Applications VIII, 1299806 (18 June 2024); https://doi.org/10.1117/12.3022263

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available