Invertible Neural Networks for Uncertainty Quantification in Photoacoustic Imaging

  title={Invertible Neural Networks for Uncertainty Quantification in Photoacoustic Imaging},
  author={Jan-Hinrich N{\"o}lke and Tim J. Adler and Janek Gr{\"o}hl and Lynton Ardizzone and Carsten Rother and U. K{\"o}the and Lena Maier-Hein},
  booktitle={Bildverarbeitung f{\"u}r die Medizin},
Multispectral photoacoustic imaging (PAI) is an emerging imaging modality which enables the recovery of functional tissue parameters such as blood oxygenation. However, the underlying inverse problems are potentially ill-posed, meaning that radically different tissue properties may - in theory - yield comparable measurements. In this work, we present a new approach for handling this specific type of uncertainty by leveraging the concept of conditional invertible neural networks (cINNs… 

Convolutional Neural Networks for Radio Frequency Ray Tracing

By iteratively predicting each set of reflected rays, the model is able to predict sharp, distinct ray propagation with low error for low numbers of reflections, indicating further work is needed to capture the more complicated behavior that higher reflection numbers entail.



Uncertainty-aware performance assessment of optical imaging modalities with invertible neural networks

A novel approach to the assessment of optical imaging modalities, which is sensitive to the different types of uncertainties that may occur when inferring tissue parameters, is presented, which could help to optimize optical camera design in an application-specific manner.

Confidence Estimation for Machine Learning-Based Quantitative Photoacoustics

This manuscript proposes to reduce the systematic error of the ROI samples by additionally discarding those pixels for which the method estimates a high error and thus a low confidence, and uses a deep learning model to compute error estimates for optical parameter estimations of a qPAI algorithm.

Solving the visibility problem in photoacoustic imaging with a deep learning approach providing prediction uncertainties

The dropout Monte-Carlo procedure is applied to estimate a pixel-wise degree of confidence on each predicted picture of the neural network predictions, and the possibility to use transfer learning with simulated data in order to drastically limit the size of the experimental dataset is addressed.

Image reconstruction with uncertainty quantification in photoacoustic tomography.

The results show that the Bayesian approach can be used to provide accurate estimates of the initial pressure distribution, as well as information about the uncertainty of the estimates.

Bayesian Image Reconstruction in Quantitative Photoacoustic Tomography

The noise of optical data is modelled as Gaussian distributed with mean and covariance approximated by solving several acoustic inverse initial value problems using acoustic noise samples as data and Bayesian approximation error modelling is applied to compensate for the modelling errors in the optical data caused by the acoustic solver.

Deep learning for photoacoustic imaging: a survey

This review performs an overview of some new developments and challenges in the application of machine learning to medical image analysis, with a special focus on deep learning in photoacoustic imaging.

Light in and sound out: emerging translational strategies for photoacoustic imaging.

In this review, the basics of PAI and its recent advances in biomedical research are described, followed by a discussion of strategies for clinical translation of the technique.

Estimating optical absorption, scattering, and Grueneisen distributions with multiple-illumination photoacoustic tomography.

It is demonstrated that multiple-illumination photoacoustic tomography (MI-PAT) can alleviate ill-posedness due to absorption-scattering nonuniqueness and produce quantitative high-resolution reconstructions of optical absorption, scattering, and Gruneisen parameter distributions.

Guided Image Generation with Conditional Invertible Neural Networks

This work introduces a new architecture called conditional invertible neural network (cINN), which combines the purely generative INN model with an unconstrained feed-forward network, which efficiently preprocesses the conditioning input into useful features.

Analyzing Inverse Problems with Invertible Neural Networks

It is argued that a particular class of neural networks is well suited for this task -- so-called Invertible Neural Networks (INNs), and it is verified experimentally that INNs are a powerful analysis tool to find multi-modalities in parameter space, to uncover parameter correlations, and to identify unrecoverable parameters.