Bayesian-Optimized Superpixel-GrabCut for Traceable Optic Disc Segmentation

Shraddha Changune; Vivek Noel Soren; Gautam Das; Tapan Kumar Gandhi

cs.CVAug 29, 2026

Bayesian-Optimized Superpixel-GrabCut for Traceable Optic Disc Segmentation

Authors: Shraddha Changune, Vivek Noel Soren, Gautam Das, Tapan Kumar Gandhi

Organizations: Electrical Engineering Indian Institute of Technology Delhi, India · Electronics and Communication Engineering Gauhati University Guwahati, India

Abstract

Optic disc (OD) segmentation is essential for diagnosing ophthalmic pathologies from retinal fundus images. However, prevailing deep learning approaches operate as opaque black boxes, lacking the inference-stage mathematical traceability--a critical requirement for algorithmic auditing and failure analysis in clinical workflows. This paper presents a fully algorithmically traceable and trainable segmentation pipeline that jointly combines superpixel decomposition, hybrid brightness-proximity superpixel scoring, morphological regularization, iterative GrabCut refinement, and elliptical shape fitting. The hyperparameter optimization is formulated as an objective function and solved via Bayesian optimization to eliminate manual parameter tuning. A quantitative evaluation on the Drishti-GS dataset demonstrates that our method achieves a Dice coefficient of 0.9536, matching state-of-the-art performance. By maintaining explicit mathematical transparency across all processing stages, our framework offers a deterministic, traceable alternative to black-box architectures for medical review and debugging.

Figures & tables

Fig. 1 : Framework for OD Segmentation and Hyperparameter Optimization

Parameter	Symbol	Range
clahe_clip †	$\gamma$	[1.0, 4.0]
n_segments ∗	$K$	[300, 700]
compactness †	$m$	[1, 30]
dist_weight †	$w_{d}$	[0.3, 0.9]
bright_weight †	$w_{b}$	[0.1, 0.7]
superpixel_count ∗	$k$	[7, 15]

TABLE I : Tunable parameters for OD segmentation pipeline (* refers to the parameters

\in\mathbb{Z}

\dagger

refers to the parameters

\in\mathbb{R})

Preprocessing
Configuration	Dice	Sens.	Spec.	Prec.
Full Pipeline	0.9536	0.9255	0.9995	0.9811
w/o CLAHE Enhancement	0.8956	0.9077	0.9956	0.8879
Segmentation
w/o Proximity Weighting	0.9048	0.8551	0.9995	0.9826
w/o Brightness Weighting	0.9456	0.9279	0.999	0.9683

TABLE II : Component-wise Analysis for OD Segmentation

Panel of Figure 4 — Fig. 2 : Qualitative segmentation results on the Drishti-GS dataset. (a), (d) Input fundus images; (b), (e) Corresponding ground-truth masks; (c), (f) Corresponding predictions with Dice scores of 0.972 and 0.897, respectively. The bottom row depicts the intermediate pipeline outputs: (g) SLIC superpixel decomposition, (h) initial coarse mask, and (i) final refined mask after GrabCut and elliptical regularization.

Traditional methods
Method	Dice	Sens.	Spec.	Prec.
Adams et al. [ 16 ]	0.88	0.83	0.88	0.87
Adhikari et al. [ 17 ]	0.89	0.87	0.91	0.91
Elazab et al. [ 18 ]	0.91	0.87	0.91	0.91
Thakur et al. [ 7 ]	0.92	0.89	0.92	0.92
Haleem et al. [ 6 ]	0.95	–	–	–

TABLE III : Comparison of our proposed framework with the existing state-of-the-art traditional methods

Optic Disc Segmentation

See all 8 papers

Optic Disc Segmentation in Fundus Images: From Classical Image Processing and Deformable Models to Modern AI
Aug 18, 2026Buket D. BarkanaOptic Disc Segmentation Ophthalmic Image Analysis
Biomedical Engineering Department, The University of Akron, Akron, OH 44325, USA
Nested Radially Monotone Polar Occupancy Estimation: Clinically-Grounded Optic Disc and Cup Segmentation for Glaucoma Screening
Apr 10, 2026Rimsa Goperma, Rojan Basnet, Liang ZhaoOptic Disc Segmentation Glaucoma Detection
Graduate School of Advanced Integrated Studies in Human Survivability (GSAIS), Kyoto University, Kyoto 606-8306, Japan
OptiModNet: A UNet-Transformer Hybrid with Grouped-Query and Channel Attention for Optic Disc and Cup Segmentation
Aug 19, 2026Soumili Ghosh, Debapriya Roy, Aryan Das +1Optic Disc Segmentation Ophthalmic Image Analysis
IEM Saltlake,India · Techno Main Salt Lake, India · VIT Bhopal University, India +1

Aug 18, 2026J/K move · Enter open · S save

Optic Disc Segmentation in Fundus Images: From Classical Image Processing and Deformable Models to Modern AI

Buket D. Barkana

Biomedical Engineering Department, The University of Akron, Akron, OH 44325, USA

Optic Disc Segmentation Ophthalmic Image Analysis Medical Image Segmentation Fundus Imaging Medical Imaging Foundation Models

Open paper AskPDF arXiv

Accurate localization and segmentation of the optic disc (OD) are important for retinal image analysis and glaucoma assessment, yet remain challenging due to variations in illumination, pathology, vascular interference, and poorly defined boundaries. This structured methodological review examines the evolution of OD segmentation from classical image-processing and deformable models to contemporary artificial intelligence (AI)-based approaches. A structured literature search and study-selection process was used to identify representative studies spanning major methodological developments. The review first summarizes commonly used fundus-image datasets, then organizes classical methods by principal mechanisms, including intensity and thresholding, histogram and entropy analysis, morphology, geometric and Hough-transform methods, filtering and feature operators, texture- and region-based approaches, and active-contour and level-set models. This paper pays particular attention to the assumptions, strengths, limitations, and complementary roles of these methods in OD localization and boundary delineation. Representative AI approaches are subsequently examined to illustrate the transition from handcrafted features and explicitly defined priors to learned representations, Transformer-based segmentation, boundary- and shape-aware learning, promptable segmentation, and retinal foundation models. Across these methodological generations, several core segmentation principles persist, including region-of-interest localization, multiscale representation, geometric and anatomical constraints, and boundary regularization, although their implementation has shifted from predefined operators to learned modules, losses, prompts, and pretrained representations. The review further identifies boundary ambiguity, anatomical variability, domain shift, and cross-dataset generalization as continuing challenges.

Abstract

Optic Disc Segmentation

Explore similar work

Optic Disc Segmentation in Fundus Images: From Classical Image Processing and Deformable Models to Modern AI

Nested Radially Monotone Polar Occupancy Estimation: Clinically-Grounded Optic Disc and Cup Segmentation for Glaucoma Screening

OptiModNet: A UNet-Transformer Hybrid with Grouped-Query and Channel Attention for Optic Disc and Cup Segmentation