I Detect What I Don't Know: Incremental Anomaly Learning with Stochastic Weight Averaging-Gaussian for Oracle-Free Medical Imaging

2025-11-05

AI Summary

Medical imaging anomaly detection is hindered by scarce annotations and high costs of expert supervision. This paper proposes an unsupervised incremental learning framework that requires no abnormal labels or prior knowledge. It enables safe model updating via dynamic expansion of the normal sample set, coupled with dual-probability gating and SWAG-Gaussian uncertainty calibration to prevent model drift and error accumulation. The architecture employs a frozen pretrained vision backbone augmented with lightweight convolutional adapters, integrated with core-set storage, k-nearest-neighbor-based anomaly scoring, z-score thresholding, and cognitive uncertainty estimation. Extensive evaluation across multiple medical imaging benchmarks demonstrates substantial performance gains: COVID-CXR (ROC-AUC 0.9982, F1 0.9746), Pneumonia CXR (ROC-AUC 0.8968), and Brain MRI ND-5 (ROC-AUC 0.7269, PR-AUC 0.8211).

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• 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
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Unknown anomaly detection in medical imaging remains a fundamental challenge due to the scarcity of labeled anomalies and the high cost of expert supervision. We introduce an unsupervised, oracle-free framework that incrementally expands a trusted set of normal samples without any anomaly labels. Starting from a small, verified seed of normal images, our method alternates between lightweight adapter updates and uncertainty-gated sample admission. A frozen pretrained vision backbone is augmented with tiny convolutional adapters, ensuring rapid domain adaptation with negligible computational overhead. Extracted embeddings are stored in a compact coreset enabling efficient k-nearest neighbor anomaly (k-NN) scoring. Safety during incremental expansion is enforced by dual probabilistic gates, a sample is admitted into the normal memory only if its distance to the existing coreset lies within a calibrated z-score threshold, and its SWAG-based epistemic uncertainty remains below a seed-calibrated bound. This mechanism prevents drift and false inclusions without relying on generative reconstruction or replay buffers. Empirically, our system steadily refines the notion of normality as unlabeled data arrive, producing substantial gains over baselines. On COVID-CXR, ROC-AUC improves from 0.9489 to 0.9982 (F1: 0.8048 to 0.9746); on Pneumonia CXR, ROC-AUC rises from 0.6834 to 0.8968; and on Brain MRI ND-5, ROC-AUC increases from 0.6041 to 0.7269 and PR-AUC from 0.7539 to 0.8211. These results highlight the effectiveness and efficiency of the proposed framework for real-world, label-scarce medical imaging applications.

Problem

• Detects unknown anomalies in medical imaging without labeled data
• Incrementally builds normal sample set using uncertainty-gated admission
• Achieves oracle-free adaptation using frozen backbone with tiny adapters

Innovation

• Uses lightweight convolutional adapters for domain adaptation
• Employs k-nearest neighbor scoring with compact coreset
• Implements dual probabilistic gates for safe sample admission