Deep brain stimulation, or DBS, is a treatment for movement disorders such as Parkinson’s disease, essential tremor, and dystonia. In this study, the researchers asked whether it is better to place the two DBS electrodes in separate surgeries, called staged implantation, or in a single surgery, called simultaneous implantation. They compared 252 patients who had staged procedures with 252 similar patients who had simultaneous procedures, using medical records from a large hospital database. The staged group tended to be older and had slightly lower body weight, and the two groups also differed somewhat in diagnosis and in the brain region targeted for stimulation. However, when the researchers compared outcomes, they found no meaningful differences between the two approaches in surgical complications, postoperative problems, delirium, hospital readmissions within 30 days, medication reduction after one year, or the need for another operation within five years. The main difference was practical: staged surgery required more anesthesia, more operating room time, and more hospital days, meaning it used more resources and took longer to complete. Overall, the study suggests that both approaches lead to similar clinical results, but doing both sides in one session may be less burdensome for patients and the healthcare system.

FIGURE 1.

Lead implantation surgery dates for staged vs simultaneous deep brain stimulation. Distribution of lead implantation surgery dates for staged and simultaneous cohorts. Each dot represents an individual patient; red horizontal lines denote median values; and red crosses indicate mean values.

]]> /valiant/2026/05/27/data-driven-fault-detection-and-isolation-enhanced-with-system-structural-relationships/ Wed, 27 May 2026 02:06:41 +0000 /valiant/?p=6824 Coursey, Austin.; Diaz-Gonzalez, Abel.; Quinones-Grueiro, Marcos.; Biswas, Gautam. (2025).��.��OpenAccess Series in Informatics, 136, 15.��

As modern machines become more complex, it is increasingly important to detect faults early and identify exactly where they are coming from. To support new methods that can work even when data are noisy, limited, and the mathematical model is incomplete, a fault-detection competition called DX 2025 LiU-ICE was created for diagnosing problems in the air path of an internal combustion engine. This paper describes the team’s winning solution. Their system first uses a semi-supervised Transformer Autoencoder, a type of machine-learning model trained to reconstruct normal engine behavior, to spot unusual patterns that may indicate a fault. To reduce false alarms, the detected signals are then passed through a rule-based filter that checks whether the problem persists long enough to be considered real. After a fault is confirmed, four neural networks estimate features from a partial system model, and the resulting residuals, meaning the differences between expected and observed behavior, are fed into a supervised classification network that estimates which fault is most likely. On the competition data, the system detected faults correctly 87% of the time with no false alarms, and it identified the correct fault with 73.8% probability on average. When tested on new driving data not seen during training, it still detected all faults and assigned the correct fault a 66.2% probability on average, although the autoencoder produced many false alarms because it did not transfer well to the new driving conditions. The authors discuss how future work could improve this weakness.

]]> /valiant/2026/05/27/generalizable-spinal-cord-multiple-sclerosis-lesion-segmentation-across-mri-contrasts-protocols-and-centers/ Wed, 27 May 2026 02:05:25 +0000 /valiant/?p=6821 Benveniste, Pierre-Louis.; Létourneau-Guillon, Laurent.; Araujo, David.; Chougar, Lydia.; Fetco, Dumitru.; Hori, Masaaki.; Kamiya, Kouhei.; Messina, Steven.; Tsagkas, Charidimos.; Audoin, Bertrand.; Bakshi, Rohit.; Bannier, Elise.; Blezek, Daniel.; Brisset, Jean-Christophe.; Callot, Virginie.; Charlson, Erik.; Chen, Michelle.; Ciccarelli, Olga.; Demortière, Sarah.; Edan, Gilles.; Filippi, Massimo.; Granberg, Tobias.; Granziera, Cristina.; Hemond, Christopher C.; Keegan, B. Mark.; Kerbrat, Anne.; Kirschke, Jan.; Kolind, Shannon.; Labauge, Pierre.; Lee, Lisa Eunyoung.; Liu, Yaou.; Mainero, Caterina.; McGinnis, Julian.; Laines Medina, Nilser.; Mühlau, Mark.; Nair, Govind.; O’Grady, Kristin P.; Oh, Jiwon.; Ouellette, Russell.; Prat, Alexandre.; Reich, Daniel S.; Rocca, Maria A.; Shepherd, Timothy M.; Smith, Seth A.; Stawiarz, Leszek.; Talbott, Jason.; Tam, Roger.; Tauhid, Shahamat.; Traboulsee, Anthony.; Treaba, Constantina Andrada.; Valsasina, Paola.; Vavasour, Zachary.; Yiannakas, Marios.; Lombaert, Hervé.; Cohen-Adad, Julien. (2026).��.��Multiple Sclerosis Journal.��

Magnetic resonance imaging, or MRI, is an important tool for finding and tracking spinal cord lesions in people with multiple sclerosis (MS), which are areas of damage caused by the disease. But automatic computer methods for detecting and outlining these lesions often work well only for one MRI type or one hospital’s scanning setup, which makes them less useful in real clinics where scan methods vary a lot. To address this, the researchers developed a more robust segmentation system, meaning a model that can automatically identify lesion boundaries, across many MRI contrasts and imaging sites. They trained and tested it on a large dataset of 4,428 annotated images from 1,849 people with MS across 23 imaging centers, using six different MRI contrast types and scans taken at 1.5, 3, and 7 tesla, which refers to the strength of the MRI scanner. Compared with existing methods that are designed for only one contrast type, the new model generalized better across different scan settings, according to neuroradiologist ratings. It also remained strong when tested across different spinal cord levels, image resolutions, threshold settings, and external datasets. Overall, the study shows that this approach can detect spinal cord MS lesions accurately and reliably across diverse MRI data, which is an important step toward making automated lesion analysis more useful in everyday clinical care.

Figure 1. Sankey diagram of annotated MRI scans across clinical sites. Line thickness is associated with the number of scans.

MRI scan distribution is clustered per acquisition type (3D, 2D sagittal, or 2D axial) and per MRI contrast, for each site.

Eclipsing binaries are pairs of stars that pass in front of each other from our point of view, and they are one of the best ways to measure basic stellar properties such as mass and radius. The challenge is that working out these properties has usually taken a lot of time and computing power, so only a small number of systems have been fully analyzed. To speed this up, the authors created a neural network, a type of artificial intelligence that learns patterns from data, which can use light curves from many common filters, radial velocity measurements for both stars, and information about the stars’ brightness across the spectrum to estimate the stars’ and orbit’s properties. The model was designed to handle messy real-world data, including extra light from nearby stars, starspots, and missing measurements, and it can also report uncertainty in each prediction. After training on simulated data, the researchers tested it on about 200 eclipsing binaries that had already been studied in detail. The model could estimate masses and radii to within about 20% and surface temperature to within about 500 K, and it did so much faster than traditional methods. Although it is not as precise as a detailed star-by-star analysis, it is well suited to the huge surveys now producing thousands of eclipsing binaries, helping researchers quickly find the most interesting systems for deeper study.

Figure 1.��Distribution of the parameter space covered by the synthetic EBs.

Magnetic resonance imaging, or MRI, is often collected as a stack of 2D slices because that can make scans faster and improve image quality for clinical use. But when software tries to analyze these scans as if they were full 3D images, it can struggle, especially when the slices are thick or have gaps between them. To address this, the researchers developed ECLARE, a new method that improves the resolution of these slice-based MRI scans without needing outside training data. ECLARE first estimates the shape of each slice’s signal, then learns from the image itself how to turn lower-resolution parts into higher-resolution ones, while also correcting for blur and making sure the image is resampled in a way that respects the original field of view. The method was tested on brain MRI data, including healthy T1-weighted scans and T2-FLAIR scans from people with multiple sclerosis, and compared with several existing image-enhancement methods. Across scans with slice thicknesses up to 5 mm and gaps up to 1.5 mm, ECLARE produced more accurate and visually similar images than the alternatives, including in important brain regions such as the ventricles, caudate, and white matter. Overall, the study suggests that ECLARE can make thick-slice MRI images more useful for 3D analysis, which could improve downstream tools that rely on detailed brain structure.

Fig.��1

Flowchart of our proposed method. The anisotropic input volume is fed independently into each of the three steps. First, in panel a (Sec.��), we estimate the slice excitation profile with ESPRESO.��Second, in panel b (Sec.��), we extract HR in-plane 2D patches and use the PSF estimated from panel a to create paired training data. This training data are used to train the network��𝑓𝜃��with supervised learning. Third, in panel c (Sec.��), we extract LR through-plane 2D slices and superresolve them with the trained network��𝑓𝜃��from panel b. The superresolved slices are stacked and averaged, yielding the superresolved output volume.

Spinal cord injury triggers a chain of biological changes, including inflammation in the nervous system, that strongly influence long-term recovery. Because of this, inflammation has become an important target for treatment. In this study, the researchers used riluzole, a drug that protects nerve tissue, to test whether two imaging methods could serve as reliable markers of injury severity, recovery over time, and response to treatment. They studied 16 male rats with a moderate contusion injury in the lower spinal cord and treated them with either riluzole or a control solution. The animals then underwent resting-state fMRI, which measures how different brain or spinal cord regions communicate when the body is at rest, and TSPO PET, a scan that can detect neuroinflammation, along with motor and sensory tests. After injury, the riluzole group showed stronger functional connectivity, meaning better communication between certain gray matter regions, above the injury site compared with the control group. In both groups, connectivity between many region pairs above and below the injury became weaker over time, and these changes matched the animals’ movement and sensory problems as well as their recovery. TSPO-PET also showed increased inflammatory activity at the injury site. Together, these results suggest that resting-state fMRI and TSPO-PET can track spinal cord injury severity and progression, and may be useful for evaluating new treatments in preclinical studies.

Fig 1.

High-resolution MTC-weighted spinal cord anatomical images at week 1 post-injury. (A–C) Representative MTC anatomical images of the spinal cord in sagittal (A), coronal (B), and axial (C) imaging plans from a Riluzole-treatment rat. Corresponding rsfMRI data were acquired using the same axial orientation field of view. Axial images of slices 2–4 are shown in (C). rsfMRI analysis was performed on slices 2 and 4, excluding slice 3 due to visible structural damage and bleeding/edema directly at the injury epicenter. Slices 1 and 5 were also excluded from the analysis due to lower SNR caused by greater distance from the MR coil isocenter and greater respiratory motion artifacts from proximity to the lungs. (D–F) Corresponding representative MTC anatomical imaging plans from a HBC-Vehicle rat.

Measuring which cell types are present in a tissue sample is important for understanding how tissues are organized and how diseases develop, but this is difficult when the data come from mixed samples. The problem is even harder because different cell types contain different amounts of RNA, the molecule used to read gene activity, and because data collected from different platforms do not always line up well. To address this, the researchers developed CSsingle, a new method for “deconvolution,” meaning it separates the mixed gene-expression signal from bulk or spatial transcriptomic data into the contributions from different cell types. CSsingle corrects for differences in cell size or RNA content, using either built-in spike-in controls or a computational estimate, and it is designed to work robustly across different data sources. Using single-cell reference data, the method estimates cell-type proportions more accurately than existing approaches. In tests on bulk data, it corrected known errors such as underestimating neutrophils in blood and tumor purity in breast cancer. When applied to spatial transcriptomics, which shows where genes are active within tissue sections, it helped map cell organization in the developing human pancreas and revealed distinct functional neighborhoods in colon cancer. Overall, CSsingle improves the analysis of complex tissues by accounting for cell-size differences and making results more reliable across platforms.

Graphical Abstract

Researchers wanted to know whether two blood proteins, von Willebrand factor (VWF) and ADAMTS13, could help predict brain aging, memory decline, and changes in brain structure. VWF helps blood clot, and ADAMTS13 is a protein that helps control VWF activity. The study followed 332 older adults in the 91������ Memory and Aging Project for about 6 years on average, with repeated blood tests, memory and thinking tests, and brain MRI scans. The main finding was that lower starting levels of ADAMTS13 were linked to faster decline in several thinking skills, including language, processing speed, executive function, memory, and visuospatial ability, as well as faster growth of white matter hyperintensities, which are bright spots on MRI that often reflect damage to the brain’s small blood vessels. These links were strongest in people who carried APOE-ε4, a gene variant known to increase Alzheimer’s disease risk. VWF by itself was not clearly linked to most outcomes, but it did interact with APOE-ε4: among people without the APOE-ε4 variant, higher VWF was associated with faster shrinkage of gray matter, the brain tissue made up mostly of neuron cell bodies. Overall, the study suggests that ADAMTS13 may be a promising blood marker of brain aging, while the role of VWF may depend on a person’s genetic background.

Figure 2. Baseline VWF×APOE‐ε4 status and longitudinal brain health outcomes.

Each datapoint represents a participant included in analytical models, illustrating each participant’s baseline plasma protein abundance (x��axis) and their unadjusted annual change in brain health outcomes of interest (yaxis). Datapoints are colored according to��APOE‐ε4 carrier status, with black datapoints representing noncarriers and red datapoints representing carriers. The solid line reflects the line of best‐fit from univariate linear regression of unadjusted annual change in brain health measures on baseline plasma protein abundance��Z��scores. Shading reflects 95% CI. For all outcomes illustrated, negative annual change indicates cognitive decline or gray matter atrophy over longitudinal follow‐up.��A, Boston Naming Test.��B, Hooper Visual Organization Test.��C, Episodic memory composite.��D, Total gray matter volume.��E, Frontal lobe gray matter volume.��F, Parietal lobe gray matter volume.��G, Temporal lobe gray matter volume.��H, Occipital lobe gray matter volume.��I, Hippocampal gray matter volume.��APOE‐ε4 indicates��apolipoprotein E‐ε4; NPX, Normalized Protein eXpression; and VWF, von Willebrand factor.

]]> /valiant/2026/05/27/distinct-oscillatory-fingerprints-of-language-and-default-mode-networks-support-language-comprehension-outcomes-a-fused-mri-eeg-study/ Wed, 27 May 2026 01:57:47 +0000 /valiant/?p=6803 Janson, Andrew.; Hong, Min Kyung.; Fotidzis, Tess S.; Koirala, Prasanna.; Aboud, Katherine. (2026).��.��NeuroImage, 333, 121940.��

Language comprehension is a complex mental process that depends on several brain networks working together over very short and longer time scales. One challenge in studying this process is that different brain imaging methods have different strengths: some show where activity happens better, while others show when it happens better. To get around this, the researchers combined functional MRI, which shows which brain areas are active, with EEG, which records the brain’s electrical activity, and used a mathematical tool called Continuous Wavelet Transform to examine changes in brain activity frequencies in the second after a word or sentence was presented. They compared natural language passages with scrambled words and found three brain network patterns that were more active during meaningful language processing. These included the main language network, a left-sided part of the default mode network, which is a set of brain regions often involved in internally directed thought, and another default mode subnetwork in both sides of the brain. Each network had its own “frequency fingerprint”: the language network was linked to longer-lasting theta activity along with bursts of beta and gamma activity, the first default mode network showed beta and gamma bursts, and the second default mode network was dominated by alpha activity. These patterns also related to language ability: differences in the language network’s frequency pattern were associated with how well people remembered what they read or heard, and reading comprehension depended partly on how strongly the language network and the alpha-dominant default mode network worked together. Overall, the study suggests that brain networks involved in language have distinct patterns of electrical activity that change over time and may help explain differences in language skill.

Fig. 1.��Stimuli presentation and fused fMRI-EEG frequency analysis. (A) Presentation of expository passages and non-sequential word baseline during both fMRI and EEG acquisition. (B) Fused fMRI-EEG analysis on subject-level inputs including passage (Pass) and word baseline (WB) to generate independent fused source components with subject weight loadings. (C) Continuous wavelet transform analysis on the EEG joint components to characterize frequency power over time throughout the post-stimulus window.

El-Sabawi, Bassim.; Huang, Xiaoning.; Lin, Phillip.; Anwar, Mohammad Yaser.; Betti, Michael.; Kim, Namju.; Perry, Andrew S.; Perera, B. Lakshitha A.; Gajjar, Priya.; Colangelo, Laura A.; Amancherla, Kaushik.; Sheng, Quanhu.; Zhao, Shilin.; Stolze, Lindsay.; Farber-Eger, Eric.; Landman, Joshua M.; Miller, Patricia E.; Liu, Gabrielle Y.; Das, Suman.; Wells, Quinn S.; Terry, James G.; Lloyd-Jones, Donald.; Das, Saumya.; Khan, Sadiya S.; North, Kari E.; Below, Jennifer.; Nayor, Matthew.; Kalhan, Ravi.; Carr, John Jeffrey.; Gamazon, Eric R.; Shah, Ravi V. (2026).��.��Arteriosclerosis, Thrombosis, and Vascular Biology, 1–15.��