Wednesday 12 March 2025
A new approach has been developed for detecting retinal diseases, which are a leading cause of vision loss and blindness worldwide. The method, called WaveNet- SF, uses a combination of spatial and frequency domain learning to improve detection accuracy.
Retinal diseases such as age-related macular degeneration (AMD) and diabetic retinopathy (DR) can have devastating effects on patients’ quality of life. Early diagnosis is crucial for effective treatment, but traditional methods can be limited by issues like speckle noise, complex lesion shapes, and varying lesion sizes.
Optical Coherence Tomography (OCT) has become a standard imaging modality for retinal disease diagnosis, but even OCT images can be challenging to interpret. WaveNet-SF addresses these challenges by incorporating wavelet transforms to decompose the images into low- and high-frequency components.
The model’s core component is a multi-scale wavelet spatial attention module (MSW-SA), which enables the model to focus on regions of interest at multiple scales. This allows it to extract both global structural features and fine-grained details, improving its ability to detect complex lesions.
To further enhance performance, WaveNet-SF includes a high-frequency feature compensation block (HFFC). This module selectively extracts essential edge information from noisy high-frequency components, reducing the impact of noise on detection accuracy.
The authors tested WaveNet-SF on two publicly available datasets: OCT-C8 and OCT-2017. The results showed that WaveNet-SF achieved state-of-the-art performance, outperforming existing methods with classification accuracies of 97.82% and 99.58%, respectively.
One of the key strengths of WaveNet-SF is its ability to handle noisy data, a common issue in retinal OCT images. By incorporating frequency domain learning, the model can effectively suppress noise and preserve fine details crucial for lesion detection.
WaveNet-SF has significant implications for the diagnosis and treatment of retinal diseases. With its ability to accurately detect complex lesions, it could potentially enable earlier intervention and improved patient outcomes. The approach also demonstrates the potential of hybrid spatial-frequency domain learning in medical image analysis.
Further research is needed to fully explore the capabilities of WaveNet-SF and to develop more advanced models that can tackle the complexities of retinal disease diagnosis. However, this development marks an important step forward in the quest for more accurate and effective diagnostic tools.
Cite this article: “WaveNet-SF: A Novel Approach to Accurate Detection of Retinal Diseases”, The Science Archive, 2025.
Retinal Diseases, Optical Coherence Tomography, Wavenet-Sf, Age-Related Macular Degeneration, Diabetic Retinopathy, Medical Image Analysis, Hybrid Spatial-Frequency Domain Learning, Lesion Detection, Noise Suppression, Classification Accuracy.
