Real-time photoacoustic projection imaging using deep learning

Abstract

Photoacoustic tomography (PAT) is an emerging and non-invasive hybrid imaging modality for visualizing light absorbing structures in biological tissue. The recently invented PAT systems using arrays of 64 parallel integrating line detectors allow capturing photoacoustic projection images in fractions of a second. Standard image formation algorithms for this type of setup suffer from under-sampling due to the sparse detector array, blurring due to the finite impulse response of the detection system, and artifacts due to the limited detection view. To address these issues, in this paper we develop a new direct and non-iterative image reconstruction framework using deep learning. The proposed DALnet combines the universal backprojection (UBP) using dynamic aperture length (DAL) correction with a deep convolutional neural network (CNN). Both subnetworks contain free parameters that are adjusted in the training phase. As demonstrated by simulation and experiment, the DALnet is capable of producing high-resolution projection images of 3D structures at a frame rate of over 50 images per second on a standard PC with NVIDIA TITAN Xp GPU. The proposed network is shown to outperform state-of-the-art iterative total variation reconstruction algorithms in terms of reconstruction speed as well as in terms of various evaluation metrics.

Learned backprojection for sparse and limited view photoacoustic tomography

“…The presented results demonstrate, that for data given on a half circle, the learned FBP clearly improves the results compared the standard FBP. We note that the learnable backprojection could also be used as first layer in a deep CNN as proposed in [24]. Further, it is possible to improve the reconstruction quality by learning temporal filters in the UBP, or to use correction weights to reduce the error on arbitrary convex and bounded domains.…”

“…In [18] an iterative algorithm, updating these angle-depending weight factors has been proposed. Deep learning methods for artifact reduction in limited view PAT have been proposed in [24] and a framework to learn an extension of the data was introduced in [8]. In this paper we propose to learn weight factors from a big dataset of incomplete data with simulated directional detector sensitivity and the corresponding ground truth reconstructions.…”

Compensating for visibility artefacts in photoacoustic imaging with a deep learning approach providing prediction uncertainties

“…It is computed over several small windows of the image, quantifying the structure, contrast and luminescence similarities. The sSSIM [44] is used for obtaining a scaled and unbiased score which was not disadvantaging for the other reconstruction methods. For an overall performance estimation of the network, the mean and standard deviation values among the test set are presented.…”

Deep learning for biomedical photoacoustic imaging: A review

“…The central idea is to leverage the flexibility of deep learning to enhance already existing modelbased reconstruction algorithms, 69,70 by introducing learnable components. To this end, Schwab et al 71 proposed an extension of the weighted universal back-projection algorithm.…”