#38 · ML / AI methods

Automated pipelines quantify subcellular and tissue structures in 3D microscopy data

The common thread across these items is packaging segmentation and quantification of microscopy images into usable workflows. OCTOPUS applies a UNet-based model to segment retinal dysplasia in mouse OCT scans and is delivered as a cross-platform open-source app, reported as validated against expert assessment. CiliAI provides a web-based deep-learning workflow that segments primary cilia substructures in 3D confocal stacks and quantifies cilium length and compartment-specific fluorescence. The Amira watershed pipeline, by contrast, is semi-automatic rather than deep-learning based, quantifying Liprin-α/SYD1 seeding sites in 3D confocal data from Drosophila synapses. The items differ in target, modality, and degree of automation. What is new is emphasis on accessibility and integration through apps and web interfaces, rather than a distinct algorithmic advance evident in these summaries.