Spot the Difference: Detection of Topological Changes via Geometric Alignment

TL;DR

This paper introduces an unsupervised method using a conditional variational auto-encoder to detect topological changes in images during geometric alignment, addressing a common challenge in domain adaptation and biomedical imaging.

Contribution

It presents the first unsupervised algorithm for detecting topological changes during image registration, applicable to diverse fields like microscopy and brain imaging.

Findings

Effective detection of topological changes in microscopy images.

Successful unsupervised anomaly detection in brain imaging.

Validated on two distinct datasets with promising results.

Abstract

Geometric alignment appears in a variety of applications, ranging from domain adaptation, optimal transport, and normalizing flows in machine learning; optical flow and learned augmentation in computer vision and deformable registration within biomedical imaging. A recurring challenge is the alignment of domains whose topology is not the same; a problem that is routinely ignored, potentially introducing bias in downstream analysis. As a first step towards solving such alignment problems, we propose an unsupervised algorithm for the detection of changes in image topology. The model is based on a conditional variational auto-encoder and detects topological changes between two images during the registration step. We account for both topological changes in the image under spatial variation and unexpected transformations. Our approach is validated on two tasks and datasets: detection of topological changes in microscopy images of cells, and unsupervised anomaly detection brain imaging.