Contour-guided attention across YOLO architectures for optic structure detection

Background.Glaucoma is a leading cause of irreversible blindness, and the cup-to-disc ratio (CDR) used to assess it requires localising the optic disc and faintly bounded cup. Attention modules are frequently proposed for this task, but of six such studies meeting a pre-stated rule, none reported their number of training runs or any run-to-run variability.Methods.We evaluate a contour-guided attention module whose spatial map comes from a fixed, non-learned contour operator applied to the input image rather than an intermediate feature map, refined by zero-initialised convolutions and injected at P3, P4 and P5 so that each block is exactly the identity at initialisation. We measure the seed-noise floor over 310 baseline runs across seven YOLOv9, YOLOv12 and YOLO26 architectures, then run pre-registered, fixed-sample, fresh-seed paired comparisons and zero-shot evaluation on REFUGE.Results.The seed alone moves baseline mAP50 across a range of, and seed matching across arms gives essentially no variance reduction, so ten paired seeds resolve onlytomAP50, the order of gains those studies attribute to attention. Against that floor the module improved YOLOv12m in two pre-registered blocks offresh-seed pairs (mAP50,and,;mAP50-95 null in both), entirely optic-cup driven and confined to intersection over union (IoU), where cup AP is nearagainstat IoU.Conclusion.The module yields a small, replicated, optic-cup-specific gain at IoUon one architecture, with no evidence of improved boundary precision and none of generality. No clinical benefit follows: on a box-derived CDR proxy the module is a precise null. The noise floor is the more transferable result, fixing the sample sizes at which such gains become checkable and showing that single-run and few-seed comparisons cannot support them.