Deep Learning Maps Osteocyte Networks And Distinguishes Age

Researchers publish a PLoS Computational Biology study (Jan 27, 2026) applying deep learning to confocal microscopy to segment osteocyte lacunar-canalicular networks, using Attention U-Net and vision-transformer variants. The model segments 81.8% of osteocytes and 42.1% of dendritic processes, and reduces analysis time from 130 hours to 10 seconds; it distinguishes 2-month and 36-month mouse bones and partially captures genetic degeneration.