arXiv subjects

Jens Sjölund

Publications and source records attributed to Jens Sjölund.

53 records · Page 3Linked to original sources

arXiv2023

Risk-sensitive Actor-free Policy via Convex Optimization

Traditional reinforcement learning methods optimize agents without considering safety, potentially resulting in unintended consequences. In this paper, we propose an optimal actor-free policy that optimizes a risk-sensitive criterion based on the conditional value at risk. The risk-sensitive objective function is modeled using an input-convex neural network ensuring convexity with respect to the actions and enabling the identification of globally optimal actions through simple gradient-following methods. Experimental results demonstrate the efficacy of our approach in maintaining effective risk control.

cs.LG↗

arXiv2023

Image Restoration with Mean-Reverting Stochastic Differential Equations

This paper presents a stochastic differential equation (SDE) approach for general-purpose image restoration. The key construction consists in a mean-reverting SDE that transforms a high-quality image into a degraded counterpart as a mean state with fixed Gaussian noise. Then, by simulating the corresponding reverse-time SDE, we are able to restore the origin of the low-quality image without relying on any task-specific prior knowledge. Crucially, the proposed mean-reverting SDE has a closed-form solution, allowing us to compute the ground truth time-dependent score and learn it with a neural network. Moreover, we propose a maximum likelihood objective to learn an optimal reverse trajectory that stabilizes the training and improves the restoration results. The experiments show that our proposed method achieves highly competitive performance in quantitative comparisons on image deraining, deblurring, and denoising, setting a new state-of-the-art on two deraining datasets. Finally, the general applicability of our approach is further demonstrated via qualitative results on image super-resolution, inpainting, and dehazing. Code is available at https://github.com/Algolzw/image-restoration-sde.

cs.LG↗

arXiv2023

Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Models

This work aims to improve the applicability of diffusion models in realistic image restoration. Specifically, we enhance the diffusion model in several aspects such as network architecture, noise level, denoising steps, training image size, and optimizer/scheduler. We show that tuning these hyperparameters allows us to achieve better performance on both distortion and perceptual scores. We also propose a U-Net based latent diffusion model which performs diffusion in a low-resolution latent space while preserving high-resolution information from the original input for the decoding process. Compared to the previous latent-diffusion model which trains a VAE-GAN to compress the image, our proposed U-Net compression strategy is significantly more stable and can recover highly accurate images without relying on adversarial optimization. Importantly, these modifications allow us to apply diffusion models to various image restoration tasks, including real-world shadow removal, HR non-homogeneous dehazing, stereo super-resolution, and bokeh effect transformation. By simply replacing the datasets and slightly changing the noise network, our model, named Refusion, is able to deal with large-size images (e.g., 6000 x 4000 x 3 in HR dehazing) and produces good results on all the above restoration problems. Our Refusion achieves the best perceptual performance in the NTIRE 2023 Image Shadow Removal Challenge and wins 2nd place overall.

cs.CV↗

arXiv2023

Probabilistic Estimation of Instantaneous Frequencies of Chirp Signals

We present a continuous-time probabilistic approach for estimating the chirp signal and its instantaneous frequency function when the true forms of these functions are not accessible. Our model represents these functions by non-linearly cascaded Gaussian processes represented as non-linear stochastic differential equations. The posterior distribution of the functions is then estimated with stochastic filters and smoothers. We compute a (posterior) Cramér--Rao lower bound for the Gaussian process model, and derive a theoretical upper bound for the estimation error in the mean squared sense. The experiments show that the proposed method outperforms a number of state-of-the-art methods on a synthetic data. We also show that the method works out-of-the-box for two real-world datasets.

stat.ML↗

arXiv2023

A Tutorial on Parametric Variational Inference

Variational inference uses optimization, rather than integration, to approximate the marginal likelihood, and thereby the posterior, in a Bayesian model. Thanks to advances in computational scalability made in the last decade, variational inference is now the preferred choice for many high-dimensional models and large datasets. This tutorial introduces variational inference from the parametric perspective that dominates these recent developments, in contrast to the mean-field perspective commonly found in other introductory texts.

stat.ML↗

arXiv2022

Unsupervised dynamic modeling of medical image transformation

Spatiotemporal imaging has applications in e.g. cardiac diagnostics, surgical guidance, and radiotherapy monitoring, In this paper, we explain the temporal motion by identifying the underlying dynamics, only based on the sequential images. Our dynamical model maps the inputs of observed high-dimensional sequential images to a low-dimensional latent space wherein a linear relationship between a hidden state process and the lower-dimensional representation of the inputs holds. For this, we use a conditional variational auto-encoder (CVAE) to nonlinearly map the higher-dimensional image to a lower-dimensional space, wherein we model the dynamics with a linear Gaussian state-space model (LG-SSM). The model, a modified version of the Kalman variational auto-encoder, is end-to-end trainable, and the weights, both in the CVAE and LG-SSM, are simultaneously updated by maximizing the evidence lower bound of the marginal likelihood. In contrast to the original model, we explain the motion with a spatial transformation from one image to another. This results in sharper reconstructions and the possibility of transferring auxiliary information, such as segmentation, through the image sequence. Our experiments, on cardiac ultrasound time series, show that the dynamic model outperforms traditional image registration in execution time, to a similar performance. Further, our model offers the possibility to impute and extrapolate for missing samples.

physics.med-ph↗

arXiv2022

Learn2Reg: comprehensive multi-task medical image registration challenge, dataset and evaluation in the era of deep learning

Image registration is a fundamental medical image analysis task, and a wide variety of approaches have been proposed. However, only a few studies have comprehensively compared medical image registration approaches on a wide range of clinically relevant tasks. This limits the development of registration methods, the adoption of research advances into practice, and a fair benchmark across competing approaches. The Learn2Reg challenge addresses these limitations by providing a multi-task medical image registration data set for comprehensive characterisation of deformable registration algorithms. A continuous evaluation will be possible at https://learn2reg.grand-challenge.org. Learn2Reg covers a wide range of anatomies (brain, abdomen, and thorax), modalities (ultrasound, CT, MR), availability of annotations, as well as intra- and inter-patient registration evaluation. We established an easily accessible framework for training and validation of 3D registration methods, which enabled the compilation of results of over 65 individual method submissions from more than 20 unique teams. We used a complementary set of metrics, including robustness, accuracy, plausibility, and runtime, enabling unique insight into the current state-of-the-art of medical image registration. This paper describes datasets, tasks, evaluation methods and results of the challenge, as well as results of further analysis of transferability to new datasets, the importance of label supervision, and resulting bias. While no single approach worked best across all tasks, many methodological aspects could be identified that push the performance of medical image registration to new state-of-the-art performance. Furthermore, we demystified the common belief that conventional registration methods have to be much slower than deep-learning-based methods.

eess.IV↗

arXiv2022

NUQ: A Noise Metric for Diffusion MRI via Uncertainty Discrepancy Quantification

Diffusion MRI (dMRI) is the only non-invasive technique sensitive to tissue micro-architecture, which can, in turn, be used to reconstruct tissue microstructure and white matter pathways. The accuracy of such tasks is hampered by the low signal-to-noise ratio in dMRI. Today, the noise is characterized mainly by visual inspection of residual maps and estimated standard deviation. However, it is hard to estimate the impact of noise on downstream tasks based only on such qualitative assessments. To address this issue, we introduce a novel metric, Noise Uncertainty Quantification (NUQ), for quantitative image quality analysis in the absence of a ground truth reference image. NUQ uses a recent Bayesian formulation of dMRI models to estimate the uncertainty of microstructural measures. Specifically, NUQ uses the maximum mean discrepancy metric to compute a pooled quality score by comparing samples drawn from the posterior distribution of the microstructure measures. We show that NUQ allows a fine-grained analysis of noise, capturing details that are visually imperceptible. We perform qualitative and quantitative comparisons on real datasets, showing that NUQ generates consistent scores across different denoisers and acquisitions. Lastly, by using NUQ on a cohort of schizophrenics and controls, we quantify the substantial impact of denoising on group differences.

eess.IV↗

arXiv2022

Graph-based Neural Acceleration for Nonnegative Matrix Factorization

We describe a graph-based neural acceleration technique for nonnegative matrix factorization that builds upon a connection between matrices and bipartite graphs that is well-known in certain fields, e.g., sparse linear algebra, but has not yet been exploited to design graph neural networks for matrix computations. We first consider low-rank factorization more broadly and propose a graph representation of the problem suited for graph neural networks. Then, we focus on the task of nonnegative matrix factorization and propose a graph neural network that interleaves bipartite self-attention layers with updates based on the alternating direction method of multipliers. Our empirical evaluation on synthetic and two real-world datasets shows that we attain substantial acceleration, even though we only train in an unsupervised fashion on smaller synthetic instances.

cs.LG↗

arXiv2020

The Elliptical Processes: a Family of Fat-tailed Stochastic Processes

We present the elliptical processes -- a family of non-parametric probabilistic models that subsumes the Gaussian process and the Student-t process. This generalization includes a range of new fat-tailed behaviors yet retains computational tractability. We base the elliptical processes on a representation of elliptical distributions as a continuous mixture of Gaussian distributions and derive closed-form expressions for the marginal and conditional distributions. We perform numerical experiments on robust regression using an elliptical process defined by a piecewise constant mixing distribution, and show advantages compared with a Gaussian process. The elliptical processes may become a replacement for Gaussian processes in several settings, including when the likelihood is not Gaussian or when accurate tail modeling is critical.

stat.ME↗

arXiv2020

Decentralized Differentially Private Segmentation with PATE

When it comes to preserving privacy in medical machine learning, two important considerations are (1) keeping data local to the institution and (2) avoiding inference of sensitive information from the trained model. These are often addressed using federated learning and differential privacy, respectively. However, the commonly used Federated Averaging algorithm requires a high degree of synchronization between participating institutions. For this reason, we turn our attention to Private Aggregation of Teacher Ensembles (PATE), where all local models can be trained independently without inter-institutional communication. The purpose of this paper is thus to explore how PATE -- originally designed for classification -- can best be adapted for semantic segmentation. To this end, we build low-dimensional representations of segmentation masks which the student can obtain through low-sensitivity queries to the private aggregator. On the Brain Tumor Segmentation (BraTS 2019) dataset, an Autoencoder-based PATE variant achieves a higher Dice coefficient for the same privacy guarantee than prior work based on noisy Federated Averaging.

cs.CV↗

arXiv2020

Registration by tracking for sequential 2D MRI

Our anatomy is in constant motion. With modern MR imaging it is possible to record this motion in real-time during an ongoing radiation therapy session. In this paper we present an image registration method that exploits the sequential nature of 2D MR images to estimate the corresponding displacement field. The method employs several discriminative correlation filters that independently track specific points. Together with a sparse-to-dense interpolation scheme we can then estimate of the displacement field. The discriminative correlation filters are trained online, and our method is modality agnostic. For the interpolation scheme we use a neural network with normalized convolutions that is trained using synthetic diffeomorphic displacement fields. The method is evaluated on a segmented cardiac dataset and when compared to two conventional methods we observe an improved performance. This improvement is especially pronounced when it comes to the detection of larger motions of small objects.

physics.med-ph↗

arXiv2019

A unified representation network for segmentation with missing modalities

Over the last few years machine learning has demonstrated groundbreaking results in many areas of medical image analysis, including segmentation. A key assumption, however, is that the train- and test distributions match. We study a realistic scenario where this assumption is clearly violated, namely segmentation with missing input modalities. We describe two neural network approaches that can handle a variable number of input modalities. The first is modality dropout: a simple but surprisingly effective modification of the training. The second is the unified representation network: a network architecture that maps a variable number of input modalities into a unified representation that can be used for downstream tasks such as segmentation. We demonstrate that modality dropout makes a standard segmentation network reasonably robust to missing modalities, but that the same network works even better if trained on the unified representation.

cs.CV↗

arXiv2018

A linear programming approach to inverse planning in Gamma Knife radiosurgery

Leksell Gamma Knife is a stereotactic radiosurgery system that allows fine-grained control of the delivered dose distribution. We describe a new inverse planning approach that both resolves shortcomings of earlier approaches and unlocks new capabilities. We fix the isocenter positions and perform sector-duration optimization using linear programming, and study the effect of beam-on time penalization on the trade-off between beam-on time and plan quality. We also describe two techniques that reduce the problem size and thus further reduce the solution time: dualization and representative subsampling. The beam-on time penalization reduces the beam-on time by a factor 2-3 compared with the naive alternative. Dualization and representative subsampling each leads to optimization time-savings by a factor 5-20. Overall, we find in a comparison with 75 clinical plans that we can always find plans with similar coverage and better selectivity and beam-on time. In 44 of these, we can even find a plan that also has better gradient index. On a standard GammaPlan workstation, the optimization times ranged from 2.3 to 26 s with a median time of 5.7 s. In conclusion, we present a combination of techniques that enables sector-duration optimization in a clinically feasible time frame.

physics.med-ph↗

arXiv2018

Bayesian uncertainty quantification in linear models for diffusion MRI

Diffusion MRI (dMRI) is a valuable tool in the assessment of tissue microstructure. By fitting a model to the dMRI signal it is possible to derive various quantitative features. Several of the most popular dMRI signal models are expansions in an appropriately chosen basis, where the coefficients are determined using some variation of least-squares. However, such approaches lack any notion of uncertainty, which could be valuable in e.g. group analyses. In this work, we use a probabilistic interpretation of linear least-squares methods to recast popular dMRI models as Bayesian ones. This makes it possible to quantify the uncertainty of any derived quantity. In particular, for quantities that are affine functions of the coefficients, the posterior distribution can be expressed in closed-form. We simulated measurements from single- and double-tensor models where the correct values of several quantities are known, to validate that the theoretically derived quantiles agree with those observed empirically. We included results from residual bootstrap for comparison and found good agreement. The validation employed several different models: Diffusion Tensor Imaging (DTI), Mean Apparent Propagator MRI (MAP-MRI) and Constrained Spherical Deconvolution (CSD). We also used in vivo data to visualize maps of quantitative features and corresponding uncertainties, and to show how our approach can be used in a group analysis to downweight subjects with high uncertainty. In summary, we convert successful linear models for dMRI signal estimation to probabilistic models, capable of accurate uncertainty quantification.

stat.AP↗

arXiv2016

Whole-brain diffusional variance decomposition (DIVIDE): Demonstration of technical feasibility at clinical MRI systems

Purpose: To assess the technical feasibility of whole-brain diffusional variance decomposition (DIVIDE) based on q-space trajectory encoding (QTE) at clinical MRI systems with varying performance. DIVIDE is used to separate diffusional heterogeneity into components that arise due to isotropic and anisotropic tissue structures. Methods: We designed imaging protocols for DIVIDE using numerically optimized gradient waveforms for diffusion encoding. Imaging was performed at systems with magnetic field strengths between 1.5 and 7 T, and gradient amplitudes between 33 and 80 mT/m. Technical feasibility was assessed from signal characteristics and quality of parameter maps in a single volunteer scanned at all systems. Results: The technical feasibility of QTE and DIVIDE was demonstrated at all systems. The system with the highest performance allowed whole-brain DIVIDE at 2 mm isotropic voxels. The system with the lowest performance required a spatial resolution of 2.5x2.5x4 mm3 to yield a sufficient signal-to-noise ratio. Conclusions: Whole-brain DIVIDE based on QTE is feasible at the investigated MRI systems. This demonstration indicates that tissue features beyond those accessible by conventional diffusion encoding may be explored on a wide range of MRI systems.

physics.med-ph↗

arXiv2016

Gaussian process regression can turn non-uniform and undersampled diffusion MRI data into diffusion spectrum imaging

We propose to use Gaussian process regression to accurately estimate the diffusion MRI signal at arbitrary locations in q-space. By estimating the signal on a grid, we can do synthetic diffusion spectrum imaging: reconstructing the ensemble averaged propagator (EAP) by an inverse Fourier transform. We also propose an alternative reconstruction method guaranteeing a nonnegative EAP that integrates to unity. The reconstruction is validated on data simulated from two Gaussians at various crossing angles. Moreover, we demonstrate on non-uniformly sampled in vivo data that the method is far superior to linear interpolation, and allows a drastic undersampling of the data with only a minor loss of accuracy. We envision the method as a potential replacement for standard diffusion spectrum imaging, in particular when acquistion time is limited.

stat.AP↗