Related Experiment Video
Updated: Oct 7, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Vessel-based hybrid optic disk segmentation applied to mobile phone camera retinal images
Tin Tin Khaing1,2, Pakinee Aimmanee3, Stanislav Makhanov1
1Sirindhorn International Institute of Technology, Thammasat University, 131 Moo 5, Tiwanont Road, Bangkadi, Meung, Pathum Thani, 12000, Thailand.
Medical & Biological Engineering & Computing
|January 6, 2022
Summary
A new hybrid method accurately detects the optic disk (OD) in smartphone retinal images, crucial for remote diabetic retinopathy screening. This automated approach achieves high accuracy, enabling efficient diagnosis for a large diabetic population.
Area of Science:
- Ophthalmology
- Medical Imaging
- Computer Vision
Background:
- Diabetic retinopathy diagnosis requires precise optic disk (OD) detection.
- Smartphone retinal imaging offers a promising, accessible approach for remote screening.
- Low quality and limited field of view in smartphone images pose challenges for OD detection.
Purpose of the Study:
- To develop and validate a fully automatic hybrid method for optic disk (OD) localization and segmentation using smartphone retinal images.
- To address the challenges of low image quality and incomplete vessel structures in mobile retinal imaging.
- To provide an efficient and reliable tool for remote screening of diabetic retinopathy.
Main Methods:
- A hybrid method (HLM) combining an exclusion method and a novel line detection method for OD localization.
- Integration of an active contour model and circle fitting for OD segmentation.
- Validation on both mobile camera datasets and standard fundus imaging equipment datasets.
Main Results:
- The HLM achieved 98% average accuracy for OD localization on mobile camera datasets.
- Segmentation routine yielded an average precision of 92.64% and recall of 82.38%.
- Comparable performance was observed against state-of-the-art methods on standard datasets.
Conclusions:
- The proposed HLM is effective for optic disk localization and segmentation in smartphone retinal images.
- This automated framework facilitates efficient and remote screening for diabetic retinopathy.
- The method demonstrates robustness across diverse retinal image datasets.
