Gold166 v1

研究团队收集了大约3万个跨物种的、广泛多样的三维神经元图像数据集，并将它们用于自动追踪算法的基准测试，生成了140万个追踪结果，得到了迄今最大的神经元重建基准测试数据集。从中选取了166个神经元图像，并通过专家手动标注得到了其“金标准”重建形态，称为Gold166数据集，用于自动追踪算法的后验基准测试。

The research team collected approximately 30,000 cross-species, broadly diverse 3D neuronal image datasets. These datasets were utilized for benchmarking automatic tracing algorithms, generating 1.4 million tracing results, thereby constructing the largest neuronal reconstruction benchmark dataset to date. From this corpus, 166 neuronal images were selected, and their "gold standard" reconstructed morphologies were obtained through expert manual annotation, forming the Gold166 dataset intended for post-hoc benchmarking of automatic tracing algorithms.

搜集汇总

背景与挑战

背景概述

Gold166 v1是一个神经科学领域的专家标注数据集，包含166个精选的神经元图像，用于自动追踪算法的基准测试。数据集由东南大学脑科学与智能技术研究院发布，是BigNeuron项目的一部分。