Development of Photoacoustic Imaging Techniques Towards Clinical Translation

This thesis encompasses an introductory part and five papers related to developing techniques and phantoms required for the further growth of photoacoustic imaging (PAI). In PAI, images are created by detecting acoustic waves followed by the absorption of laser light, enabling high spatial resolution while maintaining high optical contrast. Due to the distinct absorption patterns of various chromophores in the body, a unique photoacoustic response can be detected by changing the wavelength of the laser light. This unique photoacoustic response, called photoacoustic spectrum, is employed in a wide range of biomedical applications such as skin cancer detection and estimation of the spatial distribution of oxygen in the body. Although there have been significant advancements in the research field, this technique has not yet been adopted in healthcare, and challenges need to be addressed for its translation into clinical practice. This thesis revolves around technical advancements in PAI in different aspects that support its pathway to the clinics. Paper I introduces a tuning approach of a stable tissue-mimicking phantom (TMP) for usage in PAI studies. The results show that artists’ oil-based inks dissolved in turpentine can be utilized to tune the optical absorption properties of SEBS-gel with high accuracy, creating various photoacoustic spectral shapes and amplitudes. In addition, the long-term stability investigation of these TMPs proved their effectiveness in preserving their optical properties. Paper II presents a novel automatic threshold selection (ATS) approach that can be applied to the adaptive matched filter spectral unmixing method to distinguish the target from the background. The approach was utilized on a TMP containing inclusions, and the feasibility of its use in detecting the border of malignant melanoma (MM) skin cancers was investigated. The thickness estimated by the ATS algorithm showed a root mean squared error of 0.26 mm for the phantom inclusions and 0.19 mm for the MM skin samples compared to the ground truth and histology examination. Paper III compares the performance of two previously proposed techniques to compensate for the spectral coloring. Both approaches build upon existing techniques and are modified to be adapted to the context of human in vivo and our PAI imaging system. Results showed that both methods led to similar oxygen saturation (sO2) estimates and minimized the depth-dependent variations in sO2 that are typically observed with linear unmixing, decreasing the gradient of saturation as a function of depth as physiologically expected in a normal situation. Paper IV introduces the center frequency (CF) spectra, the mean frequency of photoacoustic data across all wavelengths. The feasibility of using the CF spectra in separating the microspheres with various sizes and colors in a phantom study is investigated. The results showed microspheres with different pigmentation in the same size exhibited different CF spectra shapes across the utilized wavelength. Moreover, changing the size of microspheres resulted in a change in the CF offset while preserving the shape. Paper V presents the clinical application of CF spectra in PAI. A normalization was applied to remove the system dependency, defined as alpha spectra. The shape and the slope of a linear model applied to the alpha spectra were evaluated as potential biomarkers for distinguishing MM and basal cell carcinoma (BCC) from healthy tissue. The median of CF spectral slopes and spectral shapes showed significant separation, with higher alpha slope values for the tumors. In another application, changes in the alpha spectra related to variations in oxygenation of the finger during an occlusion-recovery model were investigated. Results showed significant differences in the alpha spectra, with an increase in the alpha slope observed during the occlusion phase.