Sitemap
A list of all the pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
patents
System and Method for Estimating Sleep Stage
Published:
A mechanism for estimating sleep stages of an subject’s sleep session using a machine-learning algorithm. An EOG signal, produced during the sleep session, is obtained. For each of a plurality of samples, a subset of one or more other samples is obtained. The subset is identified by determining temporal dependencies between the sample and the other sample(s) using temporal information trained into the machine-learning algorithm. Each sample is processed using its corresponding subset of one or more other samples to produce an estimate sleep stage for the sample. This produces a plurality of sleep stages for the sleep session of the subject.
Download here
System and Method for Generating Output Data Using a Generative Model
Published:
A mechanism for generating output data using a generative model. The generative model comprises a plurality of score-based neural networks, each configured to generate an initial score processable, using a sampling technique, to produce an instance of example data. The initial scores are combined to define a combined score. The combined score is processed using the sampling technique to generate the output data. Each score-based neural network is trained using a different set of training data.
Download here
publications
Dose Distribution as Outcome Predictor for Gamma Knife Radiosurgery on Vestibular Schwannoma
Published in Medical Imaging, 2019
Vestibular schwannomas are benign brain tumors that can be treated radiosurgically with the Gamma Knife in order to stop tumor progression. However, in some cases tumor progression is not stopped and treatment is deemed a failure. At present, the reason for these failed treatments is unknown. Clinical factors and MRI characteristics have been considered as prognostic factors. Another confounder in the success of treatment is the treatment planning itself. It is thought to be very uniformly planned, even though dose distributions among treatment plans are highly inhomogeneous. This paper explores the predictive value of these dose distributions for the treatment outcome. We compute homogeneity indices (HI) and three-dimensional histogram-of-oriented gradients (3D-HOG) and employ support vector machine (SVM) paired with principal component analysis (PCA) for classification. In a clinical dataset, consisting of 20 tumors that showed treatment failure and 20 tumors showing treatment success, we discover that the correlation of the HI values with the treatment outcome presents no statistical evidence of an association (52.5% accuracy employing linear SVM and no statistical significant difference with t-tests), whereas the 3D-HOG features concerning the dose distribution do present correlations to the treatment outcome, suggesting the influence of the treatment on the outcome itself (77.5% accuracy employing linear SVM and PCA). These findings can provide a basis for refining towards personalized treatments and prediction of treatment efficiency. However, larger datasets are needed for more extensive analysis.
Download here
Active Deep Probabilistic Subsampling
Published in ICML, 2021
Subsampling a signal of interest can reduce costly data transfer, battery drain, radiation exposure and acquisition time in a wide range of problems. The recently proposed Deep Probabilistic Subsampling (DPS) method effectively integrates subsampling in an end-to-end deep learning model, but learns a static pattern for all datapoints. We generalize DPS to a sequential method that actively picks the next sample based on the information acquired so far; dubbed Active-DPS (A-DPS). We validate that A-DPS improves over DPS for MNIST classification at high subsampling rates. Moreover, we demonstrate strong performance in active acquisition Magnetic Resonance Image (MRI) reconstruction, outperforming DPS and other deep learning methods.
Download here
Image Denoising with Deep Unfolding and Normalizing Flows
Published in ICASSP, 2022
Many application domains, spanning from low-level computer vision to medical imaging, require high-fidelity images from noisy measurements. State-of-the-art methods for solving denoising problems combine deep learning with iterative model-based solvers, a concept known as deep algorithm unfolding or unrolling. By combining a-priori knowledge of the forward measurement model with learned proximal image-to-image mappings based on deep networks, these methods yield solutions that are both physically feasible (data-consistent) and perceptually plausible (consistent with prior belief). However, current proximal mappings based on (predominantly convolutional) neural networks only implicitly learn such image priors. In this paper, we propose to make these image priors fully explicit by embedding deep generative models in the form of normalizing flows within the unfolded proximal gradient algorithm, and training the entire algorithm in an end-to-end fashion. We demonstrate that the proposed method outperforms competitive baselines on image denoising.
Download here
Representations of Temporal Sleep Dynamics: Review and Synthesis of the Literature
Published in Sleep Medicine Reviews, 2022
Sleep is characterized by an intricate variation of brain activity over time. Measuring these temporal sleep dynamics is relevant for elucidating healthy and pathological sleep mechanisms. The rapidly increasing possibilities for obtaining and processing sleep registrations have led to an abundance of data, which can be challenging to analyze and interpret. This review provides a structured overview of approaches to represent temporal sleep dynamics, categorized based on the way the source data is compressed. For each category of representations, we describe advantages and disadvantages. Standard human-defined 30-s sleep stages have the advantages of standardization and interpretability. Alternative human-defined representations are less standardized but offer a higher temporal resolution (in case of microstructural events such as sleep spindles), or reflect non-categorical information (for example spectral power analysis). Machine-learned representations offer additional possibilities: automated sleep stages are useful for handling large quantities of data, while alternative sleep stages obtained from clustering data-driven features could aid finding new patterns and new possible clinical interpretations. While newly developed sleep representations may offer relevant insights, they can be difficult to interpret in for example a clinical context. Therefore, there should always be a balance between developing these sophisticated sleep analysis techniques and maintaining clinical explainability.
Download here
Deep Unfolding with Normalizing Flow Priors for Inverse Problems
Published in IEEE Transactions on Signal Processing, 2022
Many application domains, spanning from computational photography to medical imaging, require recovery of high-fidelity images from noisy, incomplete or partial/compressed measurements. State-of-the-art methods for solving these inverse problems combine deep learning with iterative model-based solvers, a concept known as deep algorithm unfolding or unrolling. By combining a-priori knowledge of the forward measurement model with learned proximal image-to-image mappings based on deep networks, these methods yield solutions that are both physically feasible (data-consistent) and perceptually plausible (consistent with prior belief). However, current proximal mappings based on (predominantly convolutional) neural networks only implicitly learn such image priors. In this paper, we propose to make these image priors fully explicit by embedding deep generative models in the form of normalizing flows within the unfolded proximal gradient algorithm, and training the entire algorithm end-to-end for a given task. We demonstrate that the proposed method outperforms competitive baselines on various image recovery tasks, spanning from image denoising to inpainting and deblurring, effectively adapting the prior to the restoration task at hand.
Download here
Certainty About Uncertainty in Sleep Staging: A Theoretical Framework
Published in Sleep, 2022
Sleep stage classification is an important tool for the diagnosis of sleep disorders. Because sleep staging has such a high impact on clinical outcome, it is important that it is done reliably. However, it is known that uncertainty exists in both expert scorers and automated models. On average, the agreement between human scorers is only 82.6%. In this study, we provide a theoretical framework to facilitate discussion and further analyses of uncertainty in sleep staging. To this end, we introduce two variants of uncertainty, known from statistics and the machine learning community: aleatoric and epistemic uncertainty. We discuss what these types of uncertainties are, why the distinction is useful, where they arise from in sleep staging, and provide recommendations on how this framework can improve sleep staging in the future.
Download here
Aleatoric Uncertainty Estimation of Overnight Sleep Statistics Through Posterior Sampling Using Conditional Normalizing Flows
Published in ICASSP, 2023
In sleep staging, a polysomnography is visually scored by a human expert, who creates a hypnogram that classifies the measurement into a sequence of sleep stages, from which overnight sleep statistics, such as total sleep time, are derived. Because inter-scorer agreement between humans is limited, deep learning methods trained to automate sleep staging have aleatoric uncertainty about both hypnogram and overnight statistics. We would like to estimate this aleatoric uncertainty, which can be achieved by means of posterior sampling. Current approaches model the hypnogram through a time-based factorization of categorical distributions over sleep stages. This discards time-dependent information, invalidating posterior sampling of the overnight statistics. Instead of factorizing, we propose to jointly model the sequence of sleep stages, by introducing U-Flow, a conditional normalizing flow network. We compare U-Flow to factorized baselines, leveraging 921 recordings, and show that it achieves similar performance in terms of accuracy and Cohen’s kappa on the majority voted hypnograms, while outperforming in terms of uncertainty estimation of the overnight sleep statistics.
Download here
Modeling the Impact of Inter-Rater Disagreement on Sleep Statistics Using Deep Generative Learning
Published in IEEE Journal of Biomedical and Health Informatics, 2023
Sleep staging is the process by which an overnight polysomnographic measurement is segmented into epochs of 30 seconds, each of which is annotated as belonging to one of five discrete sleep stages. The resulting scoring is graphically depicted as a hypnogram, and several overnight sleep statistics are derived, such as total sleep time and sleep onset latency. Gold standard sleep staging as performed by human technicians is time-consuming, costly, and comes with imperfect inter-scorer agreement, which also results in inter-scorer disagreement about the overnight statistics. Deep learning algorithms have shown promise in automating sleep scoring, but struggle to model inter-scorer disagreement in sleep statistics. To that end, we introduce a novel technique using conditional generative models based on Normalizing Flows that permits the modeling of the inter-rater disagreement of overnight sleep statistics, termed U-Flow. We compare U-Flow to other automatic scoring methods on a hold-out test set of 70 subjects, each scored by six independent scorers. The proposed method achieves similar sleep staging performance in terms of accuracy and Cohen’s kappa on the majority-voted hypnograms. At the same time, U-Flow outperforms the other methods in terms of modeling the inter-rater disagreement of overnight sleep statistics. The consequences of inter-rater disagreement about overnight sleep statistics may be great, and the disagreement potentially carries diagnostic and scientifically relevant information about sleep structure. U-Flow is able to model this disagreement efficiently and can support further investigations into the impact inter-rater disagreement has on sleep medicine and basic sleep research.
Download here
Studying Sleep: Towards the Identification of Hypnogram Features That Drive Expert Interpretation
Published in Sleep, 2023
Study Objectives
Hypnograms contain a wealth of information and play an important role in sleep medicine. However, interpretation of the hypnogram is a difficult task and requires domain knowledge and “clinical intuition”. This study aimed to uncover which features of the hypnogram drive interpretation by physicians. In other words, make explicit which features physicians implicitly look for in hypnograms.
Download here
Model-Based Diffusion for Mitigating Automotive Radar Interference
Published in ICASSP, 2024
Mitigating automotive radar-to-radar interference is a challenging task, especially when the observed signal is densely corrupted with highly correlated interference signals. In this paper, we propose to remove this interference using joint-conditional posterior sampling with score-based diffusion models. These models use three individual scores: a target score, an interference score, and a joint data consistency score. Leveraging the sparsity of clean target signals in the Fourier domain, we propose a model-based score estimator for the target signals, derived from the proximal step of the ℓ1 -norm. For the interference score, we use a neural network with denoising score-matching, given that it is difficult to obtain analytical statistical models of the interference signals. Lastly, the target and interference scores are connected by a data-consistency score. Experimental results show that our solution results in superior performance over state-of-the-art methods, in terms of normalized mean squared error (NMSE) and receiver operating characteristic (ROC) curves.
Download here
Single-Channel EOG Sleep Staging on a Heterogeneous Cohort of Subjects with Sleep Disorders
Published in Physiological Measurement, 2024
Objective
Sleep staging based on full polysomnography is the gold standard in the diagnosis of many sleep disorders. It is however costly, complex, and obtrusive due to the use of multiple electrodes. Automatic sleep staging based on single-channel electro-oculography (EOG) is a promising alternative, requiring fewer electrodes which could be self-applied below the hairline. EOG sleep staging algorithms are however yet to be validated in clinical populations with sleep disorders.
Download here
The Tree of Diffusion Life: Evolutionary Embeddings to Understand the Generation Process of Diffusion Models
Published in arXiv preprint, 2024
Diffusion models generate high-quality samples by corrupting data with Gaussian noise and iteratively reconstructing it with deep learning, slowly transforming noisy images into refined outputs. Understanding this data evolution is important for interpretability but is complex due to its high-dimensional evolutionary nature. While traditional dimensionality reduction methods like t-distributed stochastic neighborhood embedding (t-SNE) aid in understanding high-dimensional spaces, they neglect evolutionary structure preservation. Hence, we propose Tree of Diffusion Life (TDL), a method to understand data evolution in the generative process of diffusion models. TDL samples a diffusion model’s generative space via instances with varying prompts and employs image encoders to extract semantic meaning from these samples, projecting them to an intermediate space. It employs a novel evolutionary embedding algorithm that explicitly encodes the iterations while preserving the high-dimensional relations, facilitating the visualization of data evolution. This embedding leverages three metrics: a standard t-SNE loss to group semantically similar elements, a displacement loss to group elements from the same iteration step, and an instance alignment loss to align elements of the same instance across iterations. We present rectilinear and radial layouts to represent iterations, enabling comprehensive exploration. We assess various feature extractors and highlight TDL’s potential with prominent diffusion models like GLIDE and Stable Diffusion with different prompt sets. TDL simplifies understanding data evolution within diffusion models, offering valuable insights into their functioning.
Download here
Hypnogram and Hypnodensity Analysis of REM Sleep Behaviour Disorder Using Both EEG and HRV-Based Sleep Staging Models
Published in Journal of Sleep Research, 2025
Rapid-eye-movement (REM) sleep behaviour disorder (RBD) is a primary sleep disorder strongly associated with Parkinson’s disease. Assessing sleep structure in RBD is important for understanding the underlying pathophysiology and developing diagnostic methods. However, the performance of automated sleep stage classification (ASSC) models is considered suboptimal in RBD, for both models utilising neurological signals (“ExG”: EEG, EOG, and chin EMG) and heart rate variability combined with body movements (HRVm). Here, we explore this underperformance through the categorical representation of sleep macrostructure (i.e., hypnogram) and a representation that leverages the underlying probability distribution of ASSCs (i.e., hypnodensity). By comparing the RBD population (n=36) to a sex- and age-matched group of OSA patients chosen for their anticipated similarly decreased sleep stability, we confirm lower 4-stage classification performance in both ExG-based ASSC (RBD: κ=0.74, OSA: κ=0.80) and HRVm-based ASSC (RBD: κ=0.50, OSA: κ=0.63). Stages showing lower agreement in RBD, namely, N1+N2 and REM sleep, exhibited elevated ambiguity in the hypnodensity, indicating more ambiguous classification distributions. Limited differences in bout durations between RBD and OSA suggested sleep instability is not necessarily driving lower agreement in RBD. However, stage transitions in OSA showed more abrupt changes in the underlying probability distribution, while RBD transitions had a more continuous profile, possibly complicating classification. Although both ExG-based and HRVm-based automated sleep staging in RBD remain challenging, hypnodensity analysis is informative for the characterisation of (RBD) sleep and can capture potential drivers of classification disagreement.
Download here
Learning Structured Compressed Sensing with Automatic Resource Allocation
Published in ICASSP, 2025
Multidimensional data acquisition often requires extensive time and poses significant challenges for hardware and software regarding data storage and processing. Rather than designing a single compression matrix as in conventional compressed sensing, structured compressed sensing yields dimension-specific compression matrices, reducing the number of optimizable parameters. Recent advances in machine learning (ML) have enabled task-based supervised learning of subsampling matrices, albeit at the expense of complex downstream models. Additionally, the sampling resource allocation across dimensions is often determined in advance through heuristics. To address these challenges, we introduce Structured COmpressed Sensing with Automatic Resource Allocation (SCOSARA) with an information theory-based unsupervised learning strategy. SCOSARA adaptively distributes samples across sampling dimensions while maximizing Fisher information content. Using ultrasound localization as a case study, we compare SCOSARA to state-of-the-art ML-based and greedy search algorithms. Simulation results demonstrate that SCOSARA can produce high-quality subsampling matrices that achieve lower Cramér-Rao Bound values than the baselines. In addition, SCOSARA outperforms other ML-based algorithms in terms of the number of trainable parameters, computational complexity, and memory requirements while automatically choosing the number of samples per axis.
Download here
Deep Sylvester Posterior Inference for Adaptive Compressed Sensing in Ultrasound Imaging
Published in ICASSP, 2025
Ultrasound images are commonly formed by sequential acquisition of beam-steered scan-lines. Minimizing the number of required scan-lines can significantly enhance frame rate, field of view, energy efficiency, and data transfer speeds. Existing approaches typically use static subsampling schemes in combination with sparsity-based or, more recently, deep-learning-based recovery. In this work, we introduce an adaptive subsampling method that maximizes intrinsic information gain in-situ, employing a Sylvester Normalizing Flow encoder to infer an approximate Bayesian posterior under partial observation in real-time. Using the Bayesian posterior and a deep generative model for future observations, we determine the subsampling scheme that maximizes the mutual information between the subsampled observations, and the next frame of the video. We evaluate our approach using the EchoNet cardiac ultrasound video dataset and demonstrate that our active sampling method outperforms competitive baselines, including uniform and variable-density random sampling, as well as equidistantly spaced scan-lines, improving mean absolute reconstruction error by 15%. Moreover, posterior inference and the sampling scheme generation are performed in just 0.015 seconds (66Hz), making it fast enough for real-time 2D ultrasound imaging applications.
Download here
Removing Structured Noise Using Diffusion Models
Published in Transactions on Machine Learning Research (TMLR), 2025
Solving ill-posed inverse problems requires careful formulation of prior beliefs over the signals of interest and an accurate description of their manifestation into noisy measurements. Handcrafted signal priors based on e.g. sparsity are increasingly replaced by data-driven deep generative models, and several groups have recently shown that state-of-the-art scorebased diffusion models yield particularly strong performance and flexibility. In this paper, we show that the powerful paradigm of posterior sampling with diffusion models can be extended to include rich, structured, noise models. To that end, we propose a joint conditional reverse diffusion process with learned scores for the noise and signal-generating distribution. We demonstrate strong performance gains across various inverse problems with structured noise, outperforming competitive baselines using normalizing flows, adversarial networks and various posterior sampling methods for diffusion models. This opens up new opportunities and relevant practical applications of diffusion modeling for inverse problems in the context of non-Gaussian measurement models.
Download here
SSC-SleepNet: A Siamese-Based Automatic Sleep Staging Model with Improved N1 Sleep Detection
Published in IEEE Journal of Biomedical and Health Informatics, 2025
Automatic sleep staging from single-channel electroencephalography (EEG) using artificial intelligence (AI) is emerging as an alternative to costly and time-consuming manual scoring using multi-channel polysomnography. However, current AI methods, mainly deep learning models such as convolutional neural network (CNN) and long short-term memory (LSTM), struggle to detect the N1 sleep stage, which is challenging due to its rarity and ambiguous nature compared to other stages. Here we propose SSC-SleepNet, an automatic sleep staging algorithm aimed at improving the learning of N1 sleep. SSC-SleepNet employs a pseudo-Siamese neural network architecture owing to its capability in one- or few-shot learning with contrastive loss. Which we selected due to its strong capability in one- or few-shot learning with a contrastive loss function. SSC-SleepNet consists of two branches of neural networks: a squeeze-and-excitation residual network branch and a CNN-LSTM branch. These two branches are used to generate latent features of the EEG epoch. The adaptive loss function of SSC-SleepNet uses a weighing factor to combine weighted cross-entropy loss and focal loss to specifically address the class imbalance issue inherent in sleep staging. The proposed new loss function dynamically assigns a higher penalty to misclassified N1 sleep stages, which can improve the model’s learning capability for this minority class. Four datasets were used for sleep staging experiments. In the Sleep-EDF-SC, Sleep-EDF-X, Sleep Heart Health Study, and Haaglanden Medisch Centrum datasets, SSC-SleepNet achieved macro F1-scores of 84.5%, 89.6%, 89.5%, and 85.4% for all sleep stages, and N1 sleep stage F1-scores of 60.2%, 58.3%, 57.8%, and 55.2%, respectively. Our proposed deep learning model outperformed the most existing models in automatic sleep staging using single-channel EEG signals. In particular, N1 detection performance has been markedly improved compared to the state-of-art models.
Download here
Neural Augmented Kalman Filters for Road Network Assisted GNSS Positioning
Published in ICML Workshop on Machine Learning for Wireless Communication and Networks, 2025
The Global Navigation Satellite System (GNSS) provides critical positioning information globally, but its accuracy in dense urban environments is often compromised by multipath and non-line-of-sight errors. Road network data can be used to reduce the impact of these errors and enhance the accuracy of a positioning system. Previous works employing road network data are either limited to offline applications, or rely on Kalman Filter (KF) heuristics with little flexibility and robustness. We instead propose training a Temporal Graph Neural Network (TGNN) to integrate road network information into a KF. The TGNN is designed to predict the correct road segment and its associated uncertainty to be used in the measurement update step of the KF. We validate our approach with real-world GNSS data and open-source road networks, observing a 29% decrease in positioning error for challenging scenarios compared to a GNSS-only KF. To the best of our knowledge, ours is the first deep learning-based approach jointly employing road network data and GNSS measurements to determine the user position on Earth.
Download here
A Deep Generative Model for Five-Class Sleep Staging with Arbitrary Sensor Input
Published in IEEE Journal of Biomedical and Health Informatics, 2025
Gold-standard sleep scoring is based on epoch-based assignment of sleep stages based on a combination of EEG, EOG and EMG signals. However, a polysomnographic recording consists of many other signals that could be used for sleep staging, including cardio-respiratory modalities. Leveraging this signal variety would offer important advantages, for example increasing reliability, resilience to signal loss, and application to long-term non-obtrusive recordings. We developed a deep generative model for automatic sleep staging from a plurality of sensors and any -arbitrary- combination thereof. We trained a score-based diffusion model using a dataset of 1947 expert-labelled overnight recordings with 36 different signals, and achieved zero-shot inference on any sensor set by leveraging a novel Bayesian factorization of the score function across the sensors. On single-channel EEG, the model reaches the performance limit in terms of polysomnography inter-rater agreement (5- class accuracy 85.6%, Cohen’s kappa 0.791). Moreover, the method offers full flexibility to use any sensor set, for example finger photoplethysmography, nasal flow and thoracic respiratory movements, (5-class accuracy 79.0%, Cohen’s kappa of 0.697), or even derivations very unconventional for sleep staging, such as tibialis and sternocleidomastoid EMG (5-class accuracy 71.0%, kappa 0.575). Additionally, we propose a novel interpretability metric in terms of information gain per sensor and show this is linearly correlated with classification performance. Finally, our model allows for post- hoc addition of entirely new sensor modalities by merely training a score estimator on the novel input instead of having to retrain from scratch on all inputs.
Download here
Deep Generative Modeling in Sleep Diagnostics
Published in PhD thesis, Eindhoven University of Technology, 2026
Summary
According to clinical standards, the diagnosis of many sleep disorders requires an over-night sleep measurement known as a polysomnography (PSG). During a PSG, multiple physiological signals—such as brain activity (EEG), eye movements (EOG), and muscle tone (EMG)—are recorded to assess sleep structure and detect abnormalities. One of the most widely used methods to analyze the recorded data is sleep staging. Following the guidelines of the American Academy of Sleep Medicine (AASM) manual, gold-standard sleep staging is performed by trained human experts who divide the night into 30-second segments (known as epochs) and classify each as belonging to one of several distinct sleep stages by visually inspecting the recorded signals.
Download here
Automatic Sleep Staging and Atonia Detection in RBD Using Single-Channel Chin EMG
Published in Journal of Parkinson’s Disease, 2026
Study Objectives
The diagnosis of rapid eye movement (REM) sleep behavior disorder (RBD) is of clinical interest, as there is a high rate of future conversion to Parkinson’s disease and related neurodegenerative disorders. The clinical standard for diagnosis of RBD relies on human scoring of REM sleep without atonia (RSWA) via polysomnography, which is labor-intensive, costly, and unsuitable for long-term monitoring. Existing (semi-)automatic detection methods often depend on prior manual annotation of REM or suffer from low REM classification performance in this population. We here propose a fully automated system using only a single-channel chin electromyography (EMG) for both REM scoring and RSWA quantification.
Download here
The Geometry of LISTA
Published in IEEE Signal Processing Magazine, 2026
The celebrated iterative soft-thresholding algorithm (ISTA) and its accelerated variant, fast ISTA (FISTA), are classical signal processing methods used to solve the least absolute shrinkage and selection operator (LASSO) problem [1]. This problem spans various applications, including medical imaging, direction-ofarrival estimation, astronomy, and sparse coding. Within the broader trend in signal processing of transitioning from classical model-based approaches to deep learning-based methods, the learned ISTA (LISTA) algorithm was proposed as a way to solve the LASSO problem using deep learning while preserving the original ISTA structure [2]. LISTA learns a fast approximation to the LASSO problem by casting the weight matrices of the ISTA algorithm as learnable parameters and making them unique for each iteration, a technique now known as deep unfolding [3]. LISTA achieves superior reconstruction results in fewer iterations compared to ISTA. This is due to two main factors. First, by learning its weights, LISTA addresses potential modeling mismatches, such as imperfect knowledge of noise behavior or the forward model. Second, even with an accurately known model, the sparsifying basis might be too complex to implement efficiently using classical methods. Here, we consider a geometric interpretation to gain insight into why LISTA performs well with significantly fewer iterations (or folds) compared to ISTA. To that end, we use the fact that both models are continuous piecewise linear (CPWL) functions.
Download here
Ultrasound Operator Guidance Using World Modeling and Retrieval Based Action Planning
Published in arXiv preprint, 2026
Ultrasound is widely used, but acquisition quality is heavily dependent on the operator’s knowledge and expertise. With demand for examinations outpacing the supply of trained sonographers, operator-guidance systems aim to close this gap by instructing a less trained user how to move the probe toward a target view. In this paper, we propose a retrieval-induced latent transition model for ultrasound acquisition dynamics, formulating ultrasound operator guidance as multi-step planning and retrieval in a world model. Using a V-JEPA 2.1 backbone, observations are first encoded into a latent space where anatomically related views lie close together. We then retrieve similar views from a reference database containing encoded latent states and corresponding probe positions and orientations. Rather than learning a parametric transition function, we directly use physically executed transitions from the database to establish our nonparametric, retrieval-induced transition model that supports receding-horizon planning. At deployment, guidance is generated from the live ultrasound image feed alone, without any probe tracking hardware. Applied to carotid ultrasound, the proposed planner reaches the target view in 86% of retrospective closed-loop episodes, versus 52% and 43% for representative baselines, outperforming both on every target view, including the challenging longitudinal internal and external carotid artery views. A prospective feasibility study on unseen volunteers, run in real time on a CPU using distillation, reaches 83% target-view reachability. Because planning is driven by proximity to any encodable goal latent, the same world model can navigate back to any previously acquired, patient-specific frame, supporting reproducible longitudinal imaging for e.g. perioperative or follow-up monitoring.
Download here