Abstract
Automated cone photoreceptor quantification in adaptive optics scanning light ophthalmoscopy (AOSLO) is often evaluated through global counting accuracy. However, an accurate scalar count does not necessarily imply that the predicted cone mosaic is spatially correct. We present a formulation-level empirical study of AOSLO cone quantification as a structured prediction problem by comparing three output formulations derived from the same centroid-level annotations: density estimation, semantic segmentation, and object detection. All newly implemented models are evaluated under a common train–validation–test split and repeated independent runs. Counting performance is assessed using mean absolute error, while mask-overlap metrics are reported for segmentation models. The results show that the choice of output representation affects both counting accuracy and spatial evidence, with CNN-based semantic segmentation providing the most favorable trade-off in the evaluated setting. These findings support evaluating AOSLO cone quantification beyond scalar counting alone and highlight the value of spatially explicit formulations for cellular-level retinal image analysis. The source code to replicate the experiments is publicly available at https://github.com/santiagotoco/AOSLO_Cone_Counting
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/OMIA_026.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=c3DFGlNuEt
BibTex
@InProceedings{TolSan_Counting_MICCAISAT2026,
author = { Toledo-Cortés, Santiago AND Martinez, Juan Andrés AND Niño-León, Manuel AND Cortez, Nicolay},
title = { { Counting Is Not Enough: Spatially Explicit Formulations for AOSLO Cone Photoreceptor Quantification } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17270},
month = {pending},
page = {pending}
}
