Related Experiment Video
Updated: Jul 2, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
A cognitive deep learning approach for medical image processing
Hussam N Fakhouri1, Sadi Alawadi2,3, Feras M Awaysheh4
1Department of Data Science and Artificial Intelligence, The University of Petra, Amman, Jordan.
Scientific Reports
|February 24, 2024
Summary
A new hybrid model, cognitive Deep Learning Retinal Blood Vessel Segmentation (CoDLRBVS), enhances ophthalmic diagnostics by accurately segmenting retinal blood vessels using U-Net and image processing. It sets a new benchmark for retinal vessel segmentation accuracy.
Area of Science:
- Ophthalmic diagnostics
- Medical image processing
- Computer vision
Background:
- Accurate retinal blood vessel segmentation is crucial for diagnosing eye conditions.
- Complex retinal image features present significant challenges to existing segmentation methods.
- Existing techniques often lack efficiency and precision in complex scenarios.
Purpose of the Study:
- To introduce a novel hybrid model, cognitive Deep Learning Retinal Blood Vessel Segmentation (CoDLRBVS), for precise retinal blood vessel segmentation.
- To enhance segmentation accuracy and efficiency by integrating deep learning with advanced image processing.
- To establish a new benchmark in retinal vessel segmentation performance.
Main Methods:
- Developed CoDLRBVS, a hybrid model combining U-Net architecture with image processing techniques.
- Integrated a matched filter (MF) for preprocessing and morphological techniques (MT) for post-processing.
- Incorporated multi-scale line detection and scale space methods within a cognitive computing framework.
Main Results:
- Achieved a mean accuracy of 96.7%, precision of 96.9%, sensitivity of 99.3%, and specificity of 80.4%.
- Demonstrated superior performance across multiple datasets (DRIVE, STARE, HRF, retinal blood vessel, Chase-DB1).
- Established a new benchmark, surpassing existing models in retinal vessel segmentation.
Conclusions:
- CoDLRBVS effectively overcomes challenges in retinal blood vessel segmentation.
- The hybrid approach offers human-like adaptability and reasoning for improved medical image analysis.
- CoDLRBVS shows significant potential for advancing ophthalmic diagnostics and medical image processing.
