REFLECT: Rectified Flows for Efficient Brain Anomaly Correction Transport

2025-08-04

AI Summary

Unsupervised anomaly detection in brain MRI faces challenges due to complex anatomical structures and scarcity of abnormal samples, leading to imprecise localization. Method: We propose a single-step anomaly correction and localization framework based on Rectified Flows, which learns an optimal transport mapping from anomalous to normal brain anatomy in one forward pass—bypassing iterative sampling inherent in diffusion models—and integrates MR image reconstruction residuals for pixel-level anomaly detection and precise localization. Contribution/Results: Evaluated on mainstream brain segmentation-based anomaly detection benchmarks, our method achieves state-of-the-art performance in detection accuracy, localization precision, and inference efficiency. It significantly outperforms existing unsupervised approaches, offering a highly efficient and reliable solution for clinical decision support.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Adversarial Attacks & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Unsupervised anomaly detection (UAD) in brain imaging is crucial for identifying pathologies without the need for labeled data. However, accurately localizing anomalies remains challenging due to the intricate structure of brain anatomy and the scarcity of abnormal examples. In this work, we introduce REFLECT, a novel framework that leverages rectified flows to establish a direct, linear trajectory for correcting abnormal MR images toward a normal distribution. By learning a straight, one-step correction transport map, our method efficiently corrects brain anomalies and can precisely localize anomalies by detecting discrepancies between anomalous input and corrected counterpart. In contrast to the diffusion-based UAD models, which require iterative stochastic sampling, rectified flows provide a direct transport map, enabling single-step inference. Extensive experiments on popular UAD brain segmentation benchmarks demonstrate that REFLECT significantly outperforms state-of-the-art unsupervised anomaly detection methods. The code is available at https://github.com/farzad-bz/REFLECT.

Problem

• Unsupervised anomaly detection in brain imaging without labeled data
• Accurate localization of anomalies in complex brain anatomy
• Efficient correction of brain anomalies using rectified flows

Innovation

• Rectified flows for direct anomaly correction
• Single-step inference via linear transport map
• Unsupervised learning without labeled data