Skip to main content
Fig. 5 | Brain Informatics

Fig. 5

From: DeepNeuron: an open deep learning toolbox for neuron tracing

Fig. 5

DeepNeuron axon reconstruction. (a) The same example image as shown in Fig. 2a. (b) 3D axon signals (red dots) were extracted with neurite signal detection module; local 3D connections (green lines) between signals were generated by neurite connection module. As shown in B, the proposed DMST in the connection module can formulate the correct neurite structure even when the detected signals from different neurites are spatially close to one another. In a more difficult case, however, the loss of true signal in the local area can make the information captured by the network outweighed by the distance, resulting in false connection. This could be avoided by adequately enlarging the examined local area. Note that unconnected fragments like those on the bottom left would be further processed by our non-deep-learning functions of the tool, which is out of the scope of this article

Back to article page