Anomaly Detection Techniques and Applications

Abnormality Detection for Medical Images UsingSelf-Supervision and Negative Samples

This work explores creating negative samples for medical image abnormality detection using self-supervised self-distillation, showing improved performance on pneumonia, polyp, and glaucoma detection.

Nonparametric Anomaly Detection on Time Series of Graphs

This paper introduces a nonparametric anomaly detection framework for dynamic networks, demonstrating improved accuracy on simulated and real-world data including fMRI and communication graphs.

A Clustering Approach for Detecting Implausible Observation Values in Electronic Health Records Data

This paper develops a clustering-based method to detect implausible values in EHR data, showing it has high specificity and fewer false positives than conventional approaches.

Semi-supervised Encoding for Outlier Detection in Clinical Observation Data

This paper evaluates a semi-supervised encoding approach for outlier detection in clinical observation data from EHRs, finding that semi-supervised encoders outperform conventional autoencoders.