Domain Adaptive Synapse Detection with Weak Point Annotations

TL;DR

This paper introduces AdaSyn, a domain adaptive synapse detection framework that uses weak point annotations and a two-stage segmentation approach to improve generalization across different datasets.

Contribution

The paper proposes a novel two-stage segmentation-based framework with pseudo-label regeneration for domain adaptive synapse detection using weak annotations.

Findings

Achieved first place in the WASPSYN challenge at ISBI 2023.

Demonstrated high accuracy in synapse detection across different datasets.

Improved model generalizability with pseudo label regeneration.

Abstract

The development of learning-based methods has greatly improved the detection of synapses from electron microscopy (EM) images. However, training a model for each dataset is time-consuming and requires extensive annotations. Additionally, it is difficult to apply a learned model to data from different brain regions due to variations in data distributions. In this paper, we present AdaSyn, a two-stage segmentation-based framework for domain adaptive synapse detection with weak point annotations. In the first stage, we address the detection problem by utilizing a segmentation-based pipeline to obtain synaptic instance masks. In the second stage, we improve model generalizability on target data by regenerating square masks to get high-quality pseudo labels. Benefiting from our high-accuracy detection results, we introduce the distance nearest principle to match paired pre-synapses and post-synapses. In the WASPSYN challenge at ISBI 2023, our method ranks the 1st place.