A lung structure and function information-guided residual diffusion model for predicting idiopathic pulmonary...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonary Medicine

Background:

• Idiopathic Pulmonary Fibrosis (IPF) is a progressive lung disease causing lung scarring and thickening.
• Accurate assessment of IPF progression is crucial for treatment planning and patient survival.
• Current methods rely on multiple CT scans, delaying diagnosis of disease progression.

Purpose of the Study:

• To develop a method for early prediction of IPF progression by generating future CT images from initial scans.
• To address the limitations of current multi-scan assessment protocols.

Main Methods:

• Proposed a lung structure and function information-guided residual diffusion model.
• Utilized a 2.5D generation strategy to reduce computational costs.
• Incorporated structural attention to handle spatial misalignment.
• Employed residual diffusion for accelerated training and focused lesion analysis.
• Integrated a CLIP-based module to extract and utilize lung function test information.

Main Results:

• The proposed model effectively predicts IPF progression.
• Achieved superior generation performance compared to existing state-of-the-art methods.
• Demonstrated the feasibility of generating follow-up CT images from initial scans.

Conclusions:

• The developed AI model offers a promising solution for early IPF progression prediction.
• This approach can potentially improve patient management and outcomes by enabling timely interventions.
• The method overcomes the diagnostic delay inherent in traditional multi-scan assessment.