[1Win4-81] Improvement of detecting Age-related Macular Degeneration Using Deep Learning Method
Keywords:Age-related Macular Degeneration, Optical Coherence Tomography, YOLO
Purpose: This study aims to develop a novel deep learning model for detecting age-related macular degeneration (AMD) using optical coherence tomography (OCT) images. The research emphasizes the global health challenge of AMD and highlights the limitations of existing automated detection methods.
Methods: The proposed model incorporates YOLOv8n with advanced modules such as SEblock and CBAM to improve feature extraction and recognition, particularly in noisy and low-quality images. A redesigned architecture introduces a 160×160 detection layer to enhance small-target detection. The model is trained using publicly available datasets from UCSD and Duke University, which have been utilized in previous research.
Results: The new approach achieves superior performance with an accuracy of 99.2%, sensitivity of 98.4%, and specificity of 99.6%. It also maintains computational efficiency while reducing costs. Ablation experiments confirm the effectiveness of the design improvements in boosting detection rates and robustness.
Conclusion: This work presents a significant advancement in AMD detection, addressing challenges related to data noise and image complexity, and provides a more efficient and robust solution for automated AMD diagnosis.
Methods: The proposed model incorporates YOLOv8n with advanced modules such as SEblock and CBAM to improve feature extraction and recognition, particularly in noisy and low-quality images. A redesigned architecture introduces a 160×160 detection layer to enhance small-target detection. The model is trained using publicly available datasets from UCSD and Duke University, which have been utilized in previous research.
Results: The new approach achieves superior performance with an accuracy of 99.2%, sensitivity of 98.4%, and specificity of 99.6%. It also maintains computational efficiency while reducing costs. Ablation experiments confirm the effectiveness of the design improvements in boosting detection rates and robustness.
Conclusion: This work presents a significant advancement in AMD detection, addressing challenges related to data noise and image complexity, and provides a more efficient and robust solution for automated AMD diagnosis.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.