SRE-Conv: Symmetric Rotation Equivariant Convolution for Biomedical Image Classification

2025-01-16

AI Summary

Biomedical images often lack consistent orientation, and conventional CNNs lack rotational equivariance, leading to poor classification robustness under rotation. To address this, we propose the Symmetric Rotational Equivariant Convolution (SRE-Conv) module—a lightweight, plug-and-play convolutional layer grounded in group convolution theory. SRE-Conv is the first strictly rotationally equivariant operator compatible with arbitrary standard CNN architectures, requiring no data augmentation or post-processing. Its core innovation lies in explicit symmetry modeling, which simultaneously ensures mathematically rigorous rotational equivariance and reduces parameter count and memory footprint by 20–40%. Evaluated across all 16 classification tasks in MedMNISTv2—including both 2D and 3D biomedical imaging—models incorporating SRE-Conv demonstrate significantly improved rotational robustness and higher average classification accuracy. These results validate the effectiveness of jointly optimizing for equivariance and parameter efficiency.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Non-convex Optimization

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Security and Privacy: Privacy-enhancing technologies
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Convolutional neural networks (CNNs) are essential tools for computer vision tasks, but they lack traditionally desired properties of extracted features that could further improve model performance, e.g., rotational equivariance. Such properties are ubiquitous in biomedical images, which often lack explicit orientation. While current work largely relies on data augmentation or explicit modules to capture orientation information, this comes at the expense of increased training costs or ineffective approximations of the desired equivariance. To overcome these challenges, we propose a novel and efficient implementation of the Symmetric Rotation-Equivariant (SRE) Convolution (SRE-Conv) kernel, designed to learn rotation-invariant features while simultaneously compressing the model size. The SRE-Conv kernel can easily be incorporated into any CNN backbone. We validate the ability of a deep SRE-CNN to capture equivariance to rotation using the public MedMNISTv2 dataset (16 total tasks). SRE-Conv-CNN demonstrated improved rotated image classification performance accuracy on all 16 test datasets in both 2D and 3D images, all while increasing efficiency with fewer parameters and reduced memory footprint. The code is available at https://github.com/XYPB/SRE-Conv.

Problem

• Convolutional Neural Networks
• Biomedical Image Classification
• Rotation Invariance

Innovation

• SRE-Conv
• Rotation-Invariant
• Efficient Image Classification