Photoacoustic Imaging: Theory, Modelling and Experiments
暂无分享,去创建一个
In Photoacoustic (PA) imaging, a pulsed light excites ultrasound (US) waves in proportion to
the optical absorption of the tissue. This unique combination of light and sound brings together
optical contrast and resolution in a simple imaging modality. Many molecules in the tissue, such
as oxyhemoglobin, deoxyhemoglobin, water, etc., have a characteristic optical absorption at spe-
ci�c wavelengths. The intrinsic optical contrast of tissue molecules at speci�c wavelengths in the
near-infrared window enables PA imaging to be a potential modality in clinical applications like
early cancer diagnosis, metabolism imaging, etc. This advantage of PA imaging makes it ideal for
diagnostic applications. Over the past two decades, PA imaging has made it to several applications
including microscopy, small animal studies, pre-clinical studies and in vivo clinical studies.
An ideal clinical imaging system should be real-time, cost-e�ective and be a faithful diagnostic
tool. Following are the challenges in translating PA imaging for in vivo studies: (i) Massive data
acquisition hardware requirement for volumetric imaging. (ii) Non-realtime imaging due to com-
putationally complex reconstruction methods. (iii) Low signal strength with limited intensity laser
exposure resulting in poor image quality. In this work, these challenges have been addressed and
proposed methods for in vivo PA imaging. Speci�cally, two approaches were explored to address
these challenges. (i) To reduce hardware requirement, Compressed Sensing (CS) techniques were
also explored. (ii) To address the non real time and SNR issues, a physical acoustic lens based
solutions have been provided.
Under mild conditions, CS methods can recover a signal from highly undersampled measurements.
A simulation study was conducted on PA signals with random sensing and signal recovery using
CS techniques. Further, it was found that the inter-signal correlation between transducers is high.
This motivated the use of Distributed Compressed Sensing (DCS) signal recovery by combining in-
formation from neighboring transducer responses. Once the signal is recovered, any well known PA
reconstruction algorithm can be used to form an image. The results show that the image formed us-
ing DCS can provide better image quality than with the standard CS method. This method requires
a two-step reconstruction; �rst, the PA signal has to be recovered from compressed measurements,
and then the image has to be reconstructed. To directly form the image from compressed mea-
surements, model based imaging is employed. The model-based imaging in this context uses a PA
system matrix, and image reconstruction is done using matrix inversion. The literature is rich with
several PA system matrices, however, they do not always capture all the underlying phenomenon
(wave propagation, medium properties, and ultrasound transducer properties) involved in PA imag-
vii
ing. To address this shortcoming a pseudo spectral PA system matrix was proposed which model
wave propagation, medium properties, and ultrasound transducer properties. The purpose of using
the proposed matrix was to model the PA phenomena accurately and reconstruct the image from
undersampled measurements to improve image quality. It is evident from the results that recon-
struction using the proposed matrix provides better image quality than conventional reconstruction
algorithms in addition to enabling direct image reconstruction from undersampled measurements.
The proposed matrix was validated using experimental PA signal, and its e�ciency was tested
against other PA system matrices.
To address the challenge of non real-time imaging and low SNR, an acoustic lens based approach is
provided. An acoustic lens can be used to focus sound in a manner that is similar to how an optical
lens focuses light. An acoustic lens o�ers the following advantages: (i) Lens focusing can eliminate
reconstruction algorithms, which in turn enables real-time imaging. (ii) A larger lens aperture can
be used to focus the signal to a smaller transducer array, thereby reducing the number of transducers
required to collect the wavefront. This in turn, reduces the number of transducers, and acquisition
channels. (iii) The focusing action of the lens improves ultrasound signal strength which contributes
to improving the Signal to Noise Ratio (SNR). (iv) Further, an acoustic lens can be manufactured
using 3D printing which enables inexpensive solution. However, the use of acoustic lens was limited
to ex vivo studies as the lens can only focus an object plane to the image plane. Volumetric imaging
using an acoustic lens was impeded by defocusing at depths other than the focused object plane
and the aperture of the lens, which degrades the lateral resolution. In this thesis a volumetric PA
imaging was proposed using an acoustic lens. Two methods to overcome the above limitations were
proposed. First, a Fast Fourier Transform (FFT) based residual refocusing method was proposed
to simultaneously refocus the pressure signal from multiple depths. The second uses the concept
of spatial compounding, where a target object is viewed at multiple angles and combined to form
the �nal image. The refocusing method can improve the resolution to the di�raction limit of the
lens, while spatial compounding was shown to provide uniform and high resolution throughout the
imaging volume. Through phantom and tissue imaging studies, the e�cacy of these methods was
demonstrated.
In this thesis, challenges in in vivo PA imaging namely, hardware complexity, non real-time nature
and low SNR are addressed using theory, modeling and experimental studies. Finally, a novel in
vivo PA imaging system con�guration for thyroid imaging is proposed using an acoustic lens.