Multi-resolution and Multi-modal Feature Integration using Graph Neural Networks for Optical Coherence Tomography...

Area of Science:

• Ophthalmology and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Early diagnosis of retinal diseases is critical for effective management and prevention.
• Optical Coherence Tomography (OCT) is a key imaging modality for visualizing retinal structures and detecting abnormalities.
• Current OCT image analysis methods often rely on limited feature extraction and fusion techniques.

Purpose of the Study:

• To develop an advanced method for improved classification of retinal diseases from OCT images.
• To enhance the capture of complex lesion patterns and subtle abnormalities.
• To overcome limitations of existing feature extraction and fusion strategies in OCT analysis.

Main Methods:

• Incorporation of multi-modal semantic features extracted at multiple resolutions from OCT images.
• Development of a learnable graph-based effective feature fusion method to integrate diverse information.
• Evaluation using three publicly available datasets: OCTDL, OCTID, and OCT2014.

Main Results:

• The proposed method significantly surpasses the performance of current state-of-the-art models on retinal disease classification tasks.
• Demonstrated superior ability to capture complex lesion patterns and subtle abnormalities compared to existing approaches.
• Achieved robust and reliable classification performance across multiple OCT datasets.

Conclusions:

• The novel multi-modal feature fusion approach offers a significant advancement in OCT-based retinal disease classification.
• The learnable graph-based fusion effectively integrates information, leading to enhanced diagnostic accuracy.
• This method shows great promise for clinical application in the early and accurate detection of retinal diseases.