ISOSNet: A Unified Framework for Cone Photoreceptor Detection and Inner Segment and Outer Segment Length Measurement from AO-OCT B-Scans

Adaptive-optics optical coherence tomography (AO-OCT) resolves individual cone photoreceptors in the living retina, but extracting quantitative structural measurements from the resulting B-scans has traditionally required painstaking manual annotation. ISOSNet unifies two tasks that were previously handled separately — detecting individual cone photoreceptors and measuring the lengths of their inner and outer segments — into a single deep-learning framework. The automated pipeline produces reproducible measurements at a scale that manual grading cannot match, supporting studies of photoreceptor structure in both healthy and diseased retinas.