Tracking Intermittent Particles with Self-Learned Visual Features

2026-07-10

AI Summary

This work addresses trajectory fragmentation in time-lapse fluorescence imaging caused by particle occlusion or intermittent invisibility. To robustly reconnect trajectory segments belonging to the same particle, the authors propose a trajectory stitching method based on self-supervised visual feature learning. This approach introduces self-supervised learning for the first time to single-particle trajectory association, jointly modeling particle appearance similarity and spatial motion consistency to construct a multimodal distance metric. Experiments on fluorescence imaging sequences of Hydra vulgaris neurons demonstrate that the proposed method substantially improves stitching accuracy, reducing error rates by 50% compared to existing algorithms.

Abstract

In time-lapse fluorescence imaging, single-particle-tracking is a powerful tool to monitor the dynamics of objects of interest, and extract information about biological processes. However, tracked particles can be subject to occlusion and intermittent detectability. When these phenomena persist over a few frames, tracking algorithms tend to produce multiple tracklets for the same particle. In this work, we introduce self-supervised learning of visual features to compare tracked particles, and we exploit both visual and positional distances to robustly stitch tracklets representing the same particle. We demonstrate the performance of our stitching framework on time-lapse fluorescence sequences of Hydra vulgaris neurons. Results show high stitching precision, and reduction of errors made by previous algorithms on the same data by a factor of two.

Problem

• single-particle tracking
• intermittent detectability
• occlusion
• tracklet stitching
• fluorescence imaging

Innovation

• self-supervised learning
• single-particle tracking
• tracklet stitching
• fluorescence imaging
• visual features