Patch-based Querying Identifies Structures of Interest in Electron Microscopy

2026-10-05

AI Summary

This study addresses the high annotation costs and low analytical efficiency associated with volumetric electron microscopy (EM) data by proposing a self-supervised learning-based patch retrieval framework. The proposed method integrates local image descriptors with an interactive querying mechanism, enabling the efficient localization of similar subcellular structures using only sparse annotations while supporting dynamic search refinement by domain experts. Experimental results demonstrate that this framework reliably identifies complex cellular architectures and exhibits strong cross-dataset generalization capabilities. By significantly reducing the search space for downstream analysis, this work presents an efficient new paradigm for the automated interpretation of large-scale EM datasets.

Abstract

Volume electron microscopy (vEM) has emerged as an essential sensing technique in biomedical research, allowing the three-dimensional imaging of biological cells and tissues at nanometer-scale resolution. The ability to generate extensive datasets has reached the limitations of downstream analysis processes, which depend significantly on the intervention of human experts for preprocessing and annotation. We propose an efficient and reliable patch-based retrieval framework based on self-supervised learning of local image descriptors to locate self-similar structures in vEM datasets. Given a few manual annotations of a given cellular structure, our method can retrieve similar structures across the EM volume. Our framework is interactive, allowing the human expert to refine the search queries and retrieve relevant image patches quickly and using little labeled data. Experiments on real-world vEM images of biological tissues demonstrate that our framework can reliably identify relevant cellular structures, generalize across different organelles and acquisition modalities, and substantially reduce the search space for downstream analysis.

Problem

• volume electron microscopy
• structure retrieval
• downstream analysis bottleneck
• manual annotation
• self-supervised learning

Innovation

• Patch-based Retrieval
• Self-supervised Learning
• Volume Electron Microscopy
• Local Image Descriptors
• Interactive Querying