U-shaped fusion convolutional transformer based workflow for fast optical coherence tomography angiography generation...

Area of Science:

• Biomedical Imaging
• Medical Diagnostics
• Oral Medicine

Background:

• Oral disorders, including cancer, present diagnostic challenges due to late detection and invasive biopsies.
• Current non-invasive imaging techniques like optical coherence tomography angiography (OCTA) offer potential but suffer from motion artifacts and noise.
• Robust image reconstruction is crucial for reliable OCTA in clinical settings.

Purpose of the Study:

• To develop a novel model for reconstructing high-quality, low-noise OCTA images from limited OCT scans.
• To address limitations in current OCTA imaging for oral disorder diagnosis.
• To improve the diagnostic capabilities of OCTA in oral medicine.

Main Methods:

• Proposed a U-shaped fusion convolutional transformer (UFCT) model integrating convolutional neural networks (CNNs) and transformers.
• Utilized two repeated OCT scans for image reconstruction.
• Conducted qualitative and quantitative analyses, comparative studies with CNN and transformer models, and ablation studies.

Main Results:

• The UFCT model successfully reconstructed high-quality, low-noise OCTA images from only two repeated B-scans.
• Performance surpassed traditional OCTA generation methods in both normal and pathological conditions.
• Ablation studies validated the effectiveness of the proposed UFCT strategies.

Conclusions:

• The UFCT model demonstrates significant potential for enhancing oral medicine clinical workflows.
• Facilitates earlier detection of oral disorders and reduces the need for invasive diagnostic procedures.
• Aims to improve diagnostic accuracy and overall patient outcomes in oral healthcare.