A Study of Representational Properties of Unsupervised Anomaly Detection in Brain MRI

TL;DR

This paper investigates how different unsupervised anomaly detection methods for brain MRI utilize factorization properties, revealing their effectiveness in distinguishing normal from anomalous data through empirical analysis.

Contribution

It provides an empirical study of four existing models, highlighting the importance of factorization properties in unsupervised anomaly detection for brain MRI.

Findings

Factorization-related properties improve anomaly detection performance.

Models with these properties better distinguish normal and anomalous data.

Validation across multiple datasets confirms the observations.

Abstract

Anomaly detection in MRI is of high clinical value in imaging and diagnosis. Unsupervised methods for anomaly detection provide interesting formulations based on reconstruction or latent embedding, offering a way to observe properties related to factorization. We study four existing modeling methods, and report our empirical observations using simple data science tools, to seek outcomes from the perspective of factorization as it would be most relevant to the task of unsupervised anomaly detection, considering the case of brain structural MRI. Our study indicates that anomaly detection algorithms that exhibit factorization related properties are well capacitated with delineatory capabilities to distinguish between normal and anomaly data. We have validated our observations in multiple anomaly and normal datasets.