Integrating Neuroscientific Priors into Spiking Neural Networks: ECSNN-SEG for Robust Brain ECS Segmentation from Low-SNR Cryo-electron Microscopy Data

2026-03-09

PRE

Zhang, Chao

Mingdong Wang

Yang, Shufan

Desheng Zhao

Ce Gao

Yin, Feng *

Ping Ren

Yusong Ge *

Xiaohong Wang *

DOI：10.1007/s11220-026-00744-4

Original

Original request for help

Save

Abstract

En 中文

Medical image segmentation is crucial for computer-aided diagnosis, but segmenting complex structures like brain extracellular space (ECS) remains challenging due to low contrast and noise. Accurate ECS segmentation is vital for identifying biomarkers in neurological diseases such as Alzheimer's. This study proposes ECSNN-SEG, a two-stage spiking neural network based on the biologically plausible ECS-LIF neuron model. It integrates a MUNet variant for coarse segmentation and an ECS-LIF-based refinement network to enhance accuracy. Evaluations on the cryo-EM ECSSeg dataset show ECSNN-SEG (BPTT) achieves superior performance (ACC: 0.9833 +/- 0.0017, F1-score: 0.8647 +/- 0.0079) compared with state-of-the-art models, as well as high robustness under noise.

Keywords:

Extracellular space (ECS)

Spiking neural network (SNN)

ECS-LIF neuron

Medical image segmentation

Cryo-electron microscopy

ECSNN-SEG

Integrating Neuroscientific Priors into Spiking Neural Networks: ECSNN-SEG for Robust Brain ECS Segmentation from Low-SNR Cryo-electron Microscopy Data

Abstract

Journal

Organization

Cited Papers

Citing Papers