Related Experiment Video
Updated: Jul 5, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
DBPF-net: dual-branch structural feature extraction reinforcement network for ocular surface disease image
Cheng Wan1, Yulong Mao1, Wenqun Xi2
1College of Electronic Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, China.
Frontiers in Medicine
|January 19, 2024
Summary
A new dual-branch network reinforced by PFM block (DBPF-Net) accurately diagnoses ocular surface diseases like pterygium and subconjunctival hemorrhage from images. This AI model shows high accuracy, aiding clinical decisions for these common eye conditions.
Area of Science:
- Ophthalmology and Medical Imaging AI
Background:
- Pterygium and subconjunctival hemorrhage are prevalent ocular surface diseases causing patient distress.
- Accurate and timely diagnosis is crucial for effective patient management and treatment planning.
Purpose of the Study:
- To develop and evaluate a novel deep learning model for the automated classification of ocular surface diseases.
- To compare the proposed model's performance against established deep learning architectures.
Main Methods:
- A dataset of 2855 ocular surface images was curated across four categories: normal, subconjunctival hemorrhage, observable pterygium, and surgical pterygium.
- A dual-branch network reinforced by a patch merging and FReLU layer (PFM block) was designed as a four-way classifier.
- The DBPF-Net model was trained and validated against VGG16, ResNet50, EfficientNetB7, and Conformer models.
Main Results:
- The DBPF-Net achieved an average accuracy of 0.9789 and a kappa coefficient of 0.9681.
- For pterygium requiring surgery, the model demonstrated sensitivity (0.9210), specificity (0.9905), F1-score (0.9292), and AUC (0.9776).
- The model also showed high performance in diagnosing pterygium to be observed and subconjunctival hemorrhage.
Conclusions:
- The proposed DBPF-Net model offers a highly accurate and reliable method for diagnosing common ocular surface diseases from images.
- This AI-driven diagnostic tool has significant potential for clinical application, improving the recognition of these conditions.
