Point Positioning and Counting Network: A Deep Learning-Based Method for Automatic Axon Counting

doi:10.1016/j.exer.2026.110968

Point Positioning and Counting Network: A Deep Learning-Based Method for Automatic Axon Counting

2026-03-10

0

PRE

AI

DOI：10.1016/j.exer.2026.110968

Original

Original request for help

Save

Abstract

En 中文

• PPCNet is a point annotation-based deep learning model for axon center localization. • PPCNet integrates VGG16 feature fusion with dual-branch regression & classification. • Outperforms Axonet2.0 and AxonDeepSeg with MAE=45.93 and R2=0.939 on Large-animal (goat) dataset. • Demonstrates robustness across high and low axon density semi-thin sections.

Keywords:

PPCNet

axon center localization

deep learning

feature fusion

regression and classification

Point Positioning and Counting Network: A Deep Learning-Based Method for Automatic Axon Counting

Abstract

Journal

Organization