Conditional Invertible Neural Networks for Medical Imaging

Abstract

Over recent years, deep learning methods have become an increasingly popular choice for solving tasks from the field of inverse problems. Many of these new data-driven methods have produced impressive results, although most only give point estimates for the reconstruction. However, especially in the analysis of ill-posed inverse problems, the study of uncertainties is essential. In our work, we apply generative flow-based models based on invertible neural networks to two challenging medical imaging tasks, i.e., low-dose computed tomography and accelerated medical resonance imaging. We test different architectures of invertible neural networks and provide extensive ablation studies. In most applications, a standard Gaussian is used as the base distribution for a flow-based model. Our results show that the choice of a radial distribution can improve the quality of reconstructions.

PatchNR: Learning from Small Data by Patch Normalizing Flow Regularization

“…Assuming that we have given N example patches p 1 , ..., p N and corresponding additional information c 1 , ..., c N , we now aim to approximate the underlying conditional distribution of the patches. To this end, we train a conditional patchNR [5,14,20]. For two random variables X p and Y c , a conditional NF aims to approximate the conditional distributions P Xp|Yc=c for all possible observations of c. Formally, a conditional NF is a learned mapping…”

Invertible Neural Networks in Astrophysics

“…It has the additional advantage that for very high-dimensional or complex datasets y, we can include a feature extraction network in the conditioning block and fully integrate it in the training process [2,3]. This allows to employ cINNs in image processing tasks [13,32]. introduction of machine learning for solving regression, classification and clustering problems has revolutionized scientific research, and in particular has provided e↵ective methods for analyzing big astronomical data [17,24].…”

“…It has the additional advantage that for very high-dimensional or complex datasets y, we can include a feature extraction network in the conditioning block and fully integrate it in the training process [2,3]. This allows to employ cINNs in image processing tasks [13,32]. The particular example depicted here consists of eight affine coupling blocks interchanged with permutation layers.…”

Distilling Style from Image Pairs for Global Forward and Inverse Tone Mapping

“…We train our INN architecture in such that each pixel-wise conditioning vector (PCC) is concatenated with high level features extracted from the entire image using an additional feed-forward network 𝐻 . For this task, we use a pre-trained VGG network, similar to [Ardizzone et al 2020;Denker et al 2021] as the feed-forward network. The final conditioning vector for a given pixel is given as c 𝑝 = [C(y 𝑝 ), 𝐻 (y)], where the weights of 𝐻 (•) are simultaneously being updated alongside the weights of the INN.…”