A pseudo-time EnKF incorporating shape based reconstruction for diffuse optical tomography.

Area of Science:

• Biomedical optics
• Medical imaging
• Computational modeling

Background:

• Diffuse Optical Tomography (DOT) is a non-invasive imaging technique.
• Reconstructing complex biological structures from DOT data presents challenges.
• Accurate modeling of optical parameters is crucial for diagnostic accuracy.

Purpose of the Study:

• To develop a novel pseudo-time, sub-optimal stochastic filtering approach for DOT.
• To solve the inverse problem in DOT using a derivative-free Ensemble Kalman Filter (EnKF).
• To incorporate a shape-based reconstruction strategy for representing inhomogeneous tumor boundaries.

Main Methods:

• Utilized a derivative-free variant of the Ensemble Kalman Filter (EnKF).
• Employed circular harmonics (CH) for approximating optical parameter fields.
• Applied the pseudo-dynamic EnKF (PD-EnKF) to recover Fourier coefficients and scalar parameters.
• Used simulated and experimental photon fluence data from phantoms.

Main Results:

• PD-EnKF demonstrated superior reconstruction results compared to a deterministic Gauss-Newton algorithm.
• Successfully recovered optical parameters for inclusions with elliptical, annular ring, and dumbbell shapes.
• Validated reconstruction accuracy by matching experimental data with computed fluence data.
• Achieved accurate spatial mapping of absorption coefficients.

Conclusions:

• The PD-EnKF is accurate and robust for recovering absorption coefficients from DOT data.
• The shape-based representation and CH expansion effectively capture malignancy-representative shapes.
• The method shows low sensitivity to variations in introduced noise processes.
• This approach enhances medical diagnostic imaging capabilities through improved tumor boundary representation.