Deep Neurite Analysis Tool (DeNAT): a machine-learning framework for high-sensitivity automated neurite outgrowth measurement

Accurate quantification of neurite sprouting after injury is a critical step in axon regeneration research. Yet it remains a major bottleneck, as the current gold standard relies on manual counting by multiple blinded observers. This process is slow, labour-intensive, and prone to variability. Existing software can measure total neurite length in culture, but it is not designed to capture new growth in complex images from injury models, such as thoracic crush or pyramidotomy. Crucially, these tools lack the ability to selectively analyse growth within user-defined regions, a key requirement for injury paradigms. To address this need, we developed the Deep Neurite Analysis Tool (DeNAT), an accessible deep-learning based platform that automatically measures neurite growth after injury. DeNAT allows users to define regions of interest to specifically quantify sprouting in images from common spinal cord injury paradigms. We benchmarked its performance against manual scoring and conventional automated approaches. DeNAT achieved a near-perfect correlation (R = 0.9991) with manual ground truth, with a Sensitivity of 0.926, a Precision of 0.878, and a False Discovery Rate (FDR) of 0.12. DeNAT is optimized for high sensitivity, which is the primary requirement for neurotrauma researchers because missing a regenerating axon (false negative) carries greater risk than detecting background noise (false positive), while reducing variability and labour. By combining user-guided region selection with automated deep learning analysis, DeNAT offers an accurate, reproducible, and efficient solution for measuring neurite outgrowth in injury models.