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

TL;DR

This paper presents a deep learning method to improve photoacoustic imaging quality by compensating for view and bandwidth limitations, while also quantifying prediction confidence and exploring transfer learning to reduce data needs.

Contribution

It introduces a neural network approach that enhances image reconstruction quality and provides uncertainty estimates, using both experimental and simulated data for training.

Findings

Significantly improved image quality over conventional methods

Effective uncertainty quantification with dropout Monte-Carlo

Transfer learning reduces the need for large experimental datasets

Abstract

Conventional photoacoustic imaging may suffer from the limited view and bandwidth of ultrasound transducers. A deep learning approach is proposed to handle these problems and is demonstrated both in simulations and in experiments on a multi-scale model of leaf skeleton. We employed an experimental approach to build the training and the test sets using photographs of the samples as ground truth images. Reconstructions produced by the neural network show a greatly improved image quality as compared to conventional approaches. In addition, this work aimed at quantifying the reliability of the neural network predictions. To achieve this, the dropout Monte-Carlo procedure is applied to estimate a pixel-wise degree of confidence on each predicted picture. Last, we address the possibility to use transfer learning with simulated data in order to drastically limit the size of the experimental dataset.