Inter-Patient Modelling of 2D Lung Variations from Chest X-Ray Imaging via Fourier Descriptors

Area of Science:

• Medical Image Processing
• Computational Anatomy
• Radiology

Background:

• Accurate anatomical lung variation knowledge is crucial for medical image processing.
• Inter-patient lung variations arise from factors like age, sex, and disease.

Purpose of the Study:

• To demonstrate inter-patient lung variations using Fourier descriptors.
• To develop a 2D statistical model of human lungs from chest X-ray images.

Main Methods:

• Utilized Fourier descriptors, a shape-based technique invariant to rotation, scale, and translation.
• Trained on 380 chest X-ray (CXR) images to build a statistical mean lung contour model.
• Registered lung contours, calculated dispersion (covariance matrix), and performed principal component analysis.

Main Results:

• Developed a 2D statistical model of human lungs from CXR archives.
• Extracted statistical parameters analyzing left and right lung shape variations.
• Demonstrated the utility of Fourier descriptors for lung shape characterization.

Conclusions:

• The proposed technique effectively models inter-patient lung variations.
• Statistical parameters derived are valuable for applications like segmentation and registration.
• This approach enhances understanding of lung shape variability in medical imaging.