Individual transducer impulse response characterization method to improve image quality of array-based handheld optoacoustic tomography

Abstract

The physical properties of each transducer element play a vital role in the quality of images generated in optoacoustic (photoacoustic) tomography using transducer arrays. Thorough experimental characterization of such systems is often laborious and impractical. A shortcoming of the existing impulse response correction methods, however, is the assumption that all transducers in the array are identical and therefore share one electrical impulse response (EIR). In practice, the EIRs of the transducer elements in the array vary, and the effect of this element-to-element variability on image quality has not been investigated so far, to the best of our knowledge. We hereby propose a robust EIR derivation for individual transducer elements in an array using sparse measurements of the total impulse response (TIR) and by solving the linear system for temporal convolution. Thereafter, we combine a simulated spatial impulse response with the derived individual EIRs to obtain a full characterization of the TIR, which we call individual synthetic TIR. Correcting for individual transducer responses, we demonstrate significant improvement in isotropic resolution, which further enhances the clinical potential of array-based handheld transducers.

Multispectral optoacoustic tomography for in vivo detection of lymph node metastases in oral cancer patients using an EGFR-targeted contrast agent and intrinsic tissue contrast: A proof-of-concept study

“…The regularization parameter was tuned to be 0.01 using an L -curve. The model accounted for the acoustic and electrical properties of the probe summarized as total impulse response correction [44] , [45] . To improve the signal-to-noise ratio, three consecutive MSOT frames were reconstructed and averaged [24] .…”

Disentangling the Frequency Content in Optoacoustics

“…Fig. 1a reports the simulated electrical impulse response in receive-only mode in the time and frequency domain for a single transducer obtained with the KLM (Krimholtz, Leedom and Matthaei) model [28][29][30].…”

Deep-learning-based electrical noise removal for localized spectral optoacoustic contrast in deep tissue

“…Additionally, patient or operator movement impedes accurate averaging. Therefore, instead of experimentally acquiring noise-free optoacoustic sinograms, we generated samples of 𝑃 𝑂𝐴 via simulation by applying an accurate acoustic forward model of the scanner [20,21] to publicly available images from the PASCAL VOC2012 dataset [22], a diverse collection of over 17 000 images covering a large range of features. Utilizing these images as underlying initial pressure distributions in the simulations enables us to account for a broad range of potential features in optoacoustic sinograms and should yield a good approximation of the empirical distribution of 𝑃 𝑂𝐴 .…”

“…To evaluate the effects of the denoising method visually and quantitatively on optoacoustic images, we reconstructed the initial pressure 𝑝 0 of all breast scans in Dataset-BC, both with and without denoising the recorded sinograms with the trained neural network, using a modelbased reconstruction algorithm [20,21]. We added two regularization terms to address the two main causes of the ill-posedness of the inverse problem: simple Tikhonov regularization to mitigate limited view noise and Laplacian-based regularization to mitigate sub-resolution noise.…”