A Statistical Shape-Constrained Reconstruction Framework for Electrical Impedance Tomography.

Area of Science:

• Medical Imaging
• Computational Biology
• Biophysics

Background:

• Electrical Impedance Tomography (EIT) is a valuable imaging technique for monitoring lung function.
• Accurate reconstruction of lung geometry is crucial for reliable EIT analysis.
• Existing EIT reconstruction methods often struggle with precise anatomical representation, especially in dynamic or pathological conditions.

Purpose of the Study:

• To introduce a novel Statistical Shape-Constrained Reconstruction (SSCR) framework for lung EIT.
• To leverage statistical prior information from a large dataset of human lung shapes to improve reconstruction accuracy.
• To evaluate two SSCR implementation approaches: one-step and iterative methods.

Main Methods:

• Extraction of statistical prior models of human lung shapes from 8000 chest CT scans (800 patients).
• Implementation of a one-step SSCR for rapid reconstruction of healthy lung shapes.
• Development of an iterative SSCR for simultaneous reconstruction of pre-injured and injured lung states.
• Validation using simulated thorax imaging data and experimental laboratory data, incorporating difference imaging.

Main Results:

• Significant improvement in the accuracy of lung shape reconstruction compared to conventional methods.
• The one-step SSCR demonstrated fast and accurate reconstructions for healthy lungs.
• The iterative SSCR effectively estimated both pre-injured and injured lung regions simultaneously.
• The SSCR approaches showed robustness against measurement noise, domain boundary errors, and regularization parameter selection.

Conclusions:

• The SSCR framework effectively integrates statistical prior lung shape information into EIT reconstruction.
• Both one-step and iterative SSCR methods offer significant advantages in accuracy and robustness for lung EIT.
• This approach holds promise for more precise monitoring and diagnosis of lung conditions using EIT.