Photoacoustic Imaging With Enhanced Sensitivity,Resolution And Speed

Photoacoustic imaging(PAI)is an emerging biomedical imaging modality that can non-invasively image tissue structures and molecules without radiations.PAI can provide structural and functional imaging at various spatial and temporal scales by combining the optical contrast and ultrasonic penetration depth.However,PAI has been facing many challenges when translated to clinical applications.In this work,we aim to improve the detection sensitivity,spatial resolution and imaging speed with focus on signal detection,system implementation and image processing.An ultrasound transducer with integrated amplifier has been developed to improve the sensitivity and imaging depth of PAI.Besides,a dual-transducer off-axis photoacoustic microscopy(PAM)system with enlarged depth of field(DoF)has been built.To restore the high-resolution image,a deconvolution algorithm has been developed to reduce the blurring in photoacoustic tomography(PAT).Last but not the least,a deep learning based super resolution method has been proposed for reconstruction of blurred and under-sampled photoacoustic microscopy.The main research contents are as follows:(1)A highly sensitive ultrasonic transducer with integrated front-end amplifier has been designed to improve the detection sensitivity of photoacoustic signals.According to the transmission line theory,putting the amplifier near the transducer element shortens the transmission line and reduces the noise figure,which leads to an enhanced SNR.The integrated amplifier circuits have a wide bandwidth ranging from 5 to 200 MHz and provides at least 25 dB gain from 10 to 50MHz,making it very suitable for photoacoustic detection.The acoustic characterization shows that the highly sensitive transducer achives a noise equivalent pressure(NEP)of 0.24 mPa/(?)and improves the SNR by 1OdB compared to the non-amplified transducer.Imaging tests demonstrate the feasibility of improving the imaging depth and quality using the highly sensitive transducer.(2)Detection sensitivity plays a key role in PAM that determines the detection range and the minium light dose.Here,we built a sensitivity enhanced PAM system using the amplified transducer that enables the use of a simplified opto-acoustic beam combiner and improves the contrast-to-noise ratio(CNR)by 10 dB.Further,a dual-transducer off-axis PAM system was developed to improve the DoF,which takes advantage of the wide reception field and high sensitivity of the unfocused amplified transducer and the high frequency of the focused transducer.Imaging tests show that the unfocused amplified transducer improves the DoF by 550 micrometers compared with the focused transducer.In the focal zone,the system leverages the high axial resolution of the focused transducer.Therefore,it combines the merits of high resolution and large detection range.(3)To reduce the blurring due to the large size and limited bandwidth of the transducer element,we developed a deconvolution algorithm with hybrid reweighted adaptive total variation(HRATV)regularization.It adaptively combines the edge preservation of total variation regularizer and smoothness of fourth-order partial differential equation.The proposed deconvolution method can enhance the tissue edges while restores the microstructures and vessels.Besides,we investigated the effectiveness of using Fourier ring correlation to evaluate the image resolution and monitor the deconvolution progress in PAT.Finally,we provided a practical guide for parameter selection and optimization of the proposed deconvolution method.(4)To correct the Gaussian blurring and reconstruct under-sampled PAM image,a deep learning based photoacoustic super resolution network(PASRnet)has been proposed.The channel-spatial attention module(CSAM)is incorporated in the network to improve the performance.The degradation kernels are combined with the low-resolution images for training and inference of the network to improve the generalization ability.The PASRnet takes in the blurred and under-sampled PAM image and outputs the recovered deblurred and up-sampled PAM image.It is able to fully reconstruct the high-resolution image using only 25%of the image data,implying that the PASRnet can attain the image quality while reducing the spatial sampling rate and imaging time.