Branch Points Detection in Automated Reconstruction of Morphological Neural Circuits Via Supervised Machine Learning

Automated reconstruction of morphological neural circuits plays a vital role in advancing our understanding of brain function and development, facilitating the discovery of treatments for neurological disorders, and driving progress in artificial intelligence. Manual tracing, due to its labor-intensive and time-consuming nature, is not a scalable solution. However, automated tracing encounters a significant challenge when two axons intersect and merge, leading to substantial errors, particularly in low-resolution microscopy images.

In this study, we introduce a supervised machine learning approach designed to detect branch points in low-resolution 3D stacks. The primary objective of this technique is to enhance the precision of neural circuit reconstruction and mitigate potential errors. Our method significantly reduces the occurrence of merged crossing dendrites, thereby improving the overall accuracy of the reconstruction process.