Use this URL to cite or link to this record in EThOS:
Title: Gradient-based methods for quantitative photoacoustic tomography
Author: Soonthornsaratoon, T.
ISNI:       0000 0004 5358 9090
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Photoacoustic tomography (PAT) is showing its potential as a non-invasive biomedical imaging modality, and interest in the field is growing rapidly. The images possess excellent contrast, high spatial resolution and good specificity, however, they are largely qualitative and not directly representative of the optically absorbing structures of interest. Quantitative PAT (QPAT) aims to determine quantitatively accurate spatial maps of the underlying tissue chromophores, in order to obtain highly-resolved images of functional information such as blood oxygen saturation and haemoglobin concentration. PAT images are inherently three-dimensional (3D), and their high resolution means that the data sets are of an extremely large scale; a typical problem can easily possess 10⁷ unknowns. Existing methods for QPAT have failed to address their applicability to real, 3D PAT images, either by making restrictive approximations to the light model or by using computational intensive techniques which are impractical for large-scale data sets. This thesis develops a practical inversion method for the full and general QPAT problem, in which the tissue geometry is arbitrary, the optical coefficients are unknown and the data is large-scale. The accuracy of the inversion method is ensured by use of the radiative transfer equation (RTE), which provides a highly accurate description of the propagation of light within biological tissue. Using the RTE, a thorough investigation into the effects of errors in the scattering coefficient on the reconstructed absorption coefficient is performed. Computational efficiency in the inversion is provided through an adjoint-assisted, gradient-based minimisation scheme, which iteratively adjusts the parameters of interest until the model prediction matches the measured data. Since the RTE proves too computationally intensive for large data sets, an extension to 3D simulated data is facilitated by the incorporation of the δ-Eddington approximation, thereby providing an accurate, efficient inversion method for QPAT that may be readily applied to experimental data.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available