To push the speed and resolution limit of in vivo quantitative imaging and enable estimation of quantitative tissue parameters of subtle brain structures that were previously difficult to assess. This study implemented an efficient quantitative imaging approach, 3D-SPI MRF, on the NexGen 7T scanner equipped with a high-performance head-only gradient and 96-channel receiver array. To address challenges associated with performing rapid mesoscale MRF on this system, acquisition and reconstruction mitigation methods were developed and incorporated into the MRF framework, including: (i) flip-angle-aware dictionary fitting to account for both B1 + inhomogeneity and voxel-specific RF frequency response, (ii) gradient imperfection corrections via Skope measurements that incorporates a new per-TR trajectory rewinder compensation, (iii) incorporation of rapid B1 + and B0 mappings into the MRF sequence, and (iv) high-temporal motion navigation. Whole-brain T1 and T2 maps were obtained at 560-μm isotropic resolution within 4 min, where ablation studies demonstrated the necessity of the various mitigation methods implemented in removing bias and artifacts. For comparison, MRF data were acquired using current state-of-the-art method but limited to typical whole-body gradient specifications to demonstrate that the proposed developments resulted in ∼3× shorter scan time while producing more accurate parameter maps. Data were also acquired at ∼3.8× smaller voxel size, 360-μm isotropic, using the developed technique, to achieve mesoscale multi-parameter quantitative mapping in vivo. Tailored 3D MRF acquisition and reconstruction were developed to enable fast and accurate T1 and T2 mapping across the whole-brain at mesoscale resolution on the NexGen 7T scanner.

This paper investigates the application of machine learning (ML) models, specifically artificial neural networks (ANN) and XGBoost, for real-time motor control, focusing on switched reluctance motors (SRM) and brushless DC motors (BLDC). Traditional inverse dynamics mapping for motor control is compared with ML approaches to highlight advantages in speed, accuracy, and deployment efficiency. Datasets simulating the input-output behavior of both motor types are used to train and test the models. Key performance metrics such as mean squared error (MSE), R² score, training time, and latency are evaluated, with the goal of replacing traditional control methods in real-time applications. Results indicate that ML models outperform traditional methods in terms of prediction accuracy and deployment speed, suggesting a promising path toward more efficient and adaptive motor control systems. The novelty of this work lies in applying supervised learning directly for inverse motor control mapping, thereby eliminating the need for explicit analytical models and enabling a unified, data-driven benchmarking framework across SRM and BLDC.

A weakly-supervised deep learning model for fast localisation and delineation of the skeleton, internal organs, and spinal canal on whole-body diffusion-weighted MRI (WB-DWI).

Quantitative MRI (qMRI) offers tissue-specific biomarkers that can be tracked over time or compared across populations; however, its adoption in clinical research is hindered by significant computational demands of parameter estimation. Images acquired at high spatial resolution or requiring fitting multiple parameters often require lengthy processing time, constraining their use in routine pipelines and slowing methodological innovation and clinical translation.We present GACELLE, an open source, GPU-accelerated framework for high-throughput qMRI analysis. GACELLE provides a stochastic gradient descent optimiser and a stochastic sampler in MATLAB, enabling fast parameter mapping, improved estimation robustness via spatial regularisation, and uncertainty quantification. GACELLE prioritises accessibility: users only need to provide a forward signal model, while GACELLE's backend manages computational parallelisation, automatic parameter updates, and memory-batching. The stochastic solver performs fully vectorised Markov chain Monte Carlo with identical likelihood on CPU and GPU, ensuring reproducibility across hardware.Benchmarking demonstrates up to 451-fold acceleration for the stochastic gradient descent solver and 14,380-fold acceleration for stochastic sampling compared to CPU-based estimation, without compromising accuracy. We demonstrated GACELLE's versatility on three representative qMRI models and on an image reconstruction task. Across these applications, GACELLE improves parameter precision, enhances test-retest reproducibility, and reduces noise in quantitative maps.By combining speed, usability and flexibility, GACELLE provides a generalisable optimisation framework for medical image analysis. It lowers the computational barrier for qMRI, paving the way for reproducible biomarker development, large-scale imaging studies, and clinical translation.

Polyethylene glycol (PEO) is a commonly used polymer additive for Zn dendrite suppression, to achieve longer-lasting and higher-performance zinc aqueous batteries. However, the precise mechanism by which PEO suppresses interfacial roughness remains unclear. In this study, we addressed these knowledge gaps by using in situ electrochemical atomic force microscopy (EC-AFM) to observe the nucleation and growth of Zn metal plates on copper (Cu) substrates in the presence of different concentrations of ZnSO4 and PEO. Here the ZnSO4 solution provided the electrolyte and the Cu substrates served as the electrodes, both of which are widely utilized in Zn batteries. Our results show that PEO biases the crystallographic orientation of the initially deposited Zn metal nuclei, but does not have an obvious influence on subsequent growth of the resulting Zn platelets. The consistent aspect ratio of the Zn plates combined with the lack of an effect on growth rates suggests that PEO does not interact significantly with the surface of the newly formed Zn plates. High-speed and high-resolution in-situ AFM, along with in-situ attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) reveal that Zn metal undergoes rapid surface reorganization in a mildly acidic aqueous solution due to oxidation, which is not affected by PEO adsorption. Adhesion force maps, obtained through contact fast force mapping under flowing AFM, demonstrate real-time PEO adsorption and distribution on Cu and Zn surfaces, confirming a strong Cu-PEO interaction and a weak PEO-Zn interaction after oxidation. Based on these findings, we hypothesize that PEO primarily interacts with the Cu substrate to adjust the interfacial structure and energy of the Cu-electrolyte interface. To test this hypothesis, we conducted molecular dynamics (MD) simulations to simulate the electric double-layer structure in the presence and absence of PEO. We also calculated the Cu-Zn and Zn-solvent interfacial energies under both conditions. Our findings provide a clear picture of how PEO flattens the electrode and suggest key design and engineering principles for flat electrode synthesis in energy-related applications, emphasizing the use of polymer additives that exhibit appropriately strong binding to the substrate metal and weak binding to the deposited metal.

Summary Recently, semi-airborne transient electromagnetic (TEM) systems have gained attention in geophysical investigations due to their ability for fast mapping and minimal ground access requirement. These systems consist of a ground-based transmitter source and an inductive receiver coil, carried by an uncrewed aerial vehicle. This study investigates how transmitter source selection in field-based semi-airborne TEM systems affects model parameter uncertainty, using synthetic subsurface models. The simulated dB/dt responses highlight distinct signal characteristics between the galvanic-based system (herein galvanic source) and the inductive-based system (herein inductive source), with differences observed across varying subsurface conditions. An analysis of four synthetic 3-layer models highlights that the inductive-based system resolves shallow conductors better at short offsets, whereas the galvanic-based system is better at resolving parameters for deeper targets at large offsets. Both systems, however, face challenges in accurately resolving resistive targets embedded between conductors, highlighting the need for strategic selection of the transmitter source. The galvanic-based system consistently achieves a higher signal-to-noise ratio (SNR), particularly at large offsets, making it better suited for deep exploration. In contrast, the inductive-based system exhibits lower SNR, higher noise susceptibility, and sign-changing dB/dt responses at increasing offsets adding complexity to data processing and interpretation. Despite these limitations, inductive-based systems enable earlier time measurements with signal magnitudes at short offsets comparable to galvanic-based, due to shorter current turn-off times. In this analysis we have used two system setups utilizing inductive and galvanic sources that reflect commonly used systems, but obviously, assumptions regarding transmitter characteristics such as type, size, waveform, and current amplitude, will influence the results when examining details more closely.

Abstract We present MomentEmu, a general-purpose polynomial emulator for fast and interpretable mappings between theoretical parameters and observational features. The method constructs moment matrices to project simulation data onto polynomial bases, yielding symbolic expressions that approximate the target mapping. Compared to neural-network-based emulators, MomentEmu offers negligible training cost, millisecond-level evaluation, and transparent functional forms. As a proof-of-concept demonstration, we develop two emulators: PolyCAMB-Dℓ, which maps six cosmological parameters to the CMB power spectra (TT, EE, BB, TE), and PolyCAMB-peak, which enables a bidirectional mapping between the cosmological parameters and the acoustic peak features of $D_\ell ^{\rm TT}$. PolyCAMB-Dℓ achieves sub-percent accuracy over multipoles ℓ ≤ 4050, while PolyCAMB-peak also attains comparable precision and produces symbolic forms consistent with known analytical approximations. The method is well suited for forward modelling, parameter inference, and uncertainty propagation, particularly when the parameter space is moderate in dimensionality and the mapping is smooth. MomentEmu offers a lightweight and portable alternative to regression-based or black-box emulators in cosmological analysis.

Blood T2 is sensitive to blood oxygenation level, and right ventricular (RV) oxygenation abnormalities may be detected by the relative difference between T2 values of the left ventricular (LV) and RV blood pools. To investigate the prognostic value of right-to-left ventricular blood pool T2 ratio (RV/LV T2 ratio) in patients with dilated cardiomyopathy (DCM). Retrospective. Three hundred and eleven patients with DCM (mean age: 52 ± 14.5 years; 72% male) and 30 age- and sex-matched healthy controls. 1.5 or 3.0 T, steady-state free precession cine sequence, T2-prepared fast low-angle shot T2 mapping sequence, and phase-sensitive inversion recovery late gadolinium enhancement (LGE) sequence. Clinical characteristics, conventional cardiac MRI parameters (ventricular volumes, function, mass), LGE extent, native myocardial T2, LV and RV strain, and RV/LV T2 ratio were assessed. Patients were followed up for a median duration of 34 months (interquartile range 21-45 months). The primary outcome, major adverse cardiac events (MACE), includes all-cause mortality, heart failure-related hospitalization, heart transplantation, or aborted sudden cardiac death. The incremental prognostic value of RV/LV T2 ratio for MACE was assessed beyond traditional risk factors. The prognostic value of RV/LV T2 ratio was evaluated using multivariable Cox regression analysis and Kaplan-Meier curves. Incremental prognostic value was evaluated using C indices and likelihood ratio tests. A p value < 0.05 was considered significant. One hundred and twenty-four patients experienced MACE. RV/LV T2 ratio was significantly lower in participants with MACE, and was associated with MACE in patients with DCM, irrespective of LVEF and LGE extent (above or below median value). In Cox analysis, RV/LV T2 ratio was independently associated with MACE (hazard ratio: 1.32 per 0.1 decrease), and provided significant incremental prognostic value beyond traditional risk factors in patients with DCM. RV/LV T2 ratio was independently associated with MACE among patients with DCM, providing incremental prognostic value when combined in a model with clinical and conventional MRI risk factors. 3. 5.

This study explored the possible influence of phonotactic and orthotactic probability on fast mapping in school-age children with and without a developmental disorder. Participants were exposed to pictures of novel objects paired with spoken and written novel words, which varied in phonotactic and orthotactic probability. Fast mapping was examined with gaze behavior (proportion of looking time) and accuracy (picture identification, orthographic identification, picture naming). Participants were highly accurate during picture identification and orthographic identification, and less accurate during picture naming. Fast mapping was somewhat influenced by phonotactic and orthotactic probability. During picture identification, children were less likely to be accurate in trials with novel words with low orthotactic probability. Gaze behavior during exposure and response accuracy varied in terms of language ability and age.

Cardiac quantitative MRI (qMRI) is a powerful imaging technique for diagnosing pathologies such as diffuse myocardial fibrosis. One main challenge is cardiac motion, which requires synchronization of data acquisition with the heartbeat, leading to long scan times. We present a novel deep learning-based image registration method for cardiac qMRI that enables non-rigid motion correction of data acquired continuously over multiple cardiac cycles, thereby reducing scan times. Our method is a zero-shot approach that utilizes the physical qMRI signal model for accurate motion estimation. Non-rigid motion of dynamic images is estimated with a U-Net-based architecture. This exploits the intrinsic smoothness of cardiac motion, allowing sharing information between neighboring images. The approach is robust to undersampling artifacts, enabling motion estimation from dynamic images reconstructed from very few k-space data even without advanced image reconstruction methods. We evaluated the method for fast cardiac T1 mapping using a Golden radial sampling scheme on numerical simulations and in-vivo acquisitions. On numerical simulations, our method achieved a 61.64% improvement in T1 accuracy. On in-vivo data, our approach yielded a 45.13% improvement in sharpness of T1 maps, and temporal image alignment of motion-corrected dynamics improved on average by 11.78%. Our method enables accurate non-rigid motion correction of highly undersampled cardiac qMRI data obtained from continuously acquired data. As our method is individually optimized for each scan without the need for training on large datasets, it can easily be adapted to other cardiac qMRI approaches.

Abstract Advances in artificial neural networks (ANNs) have revolutionized the way we work, learn, and acquire information, achieving human-level capabilities. Yet, ANNs differ fundamentally from the human brain in how symbolic knowledge is acquired, typically requiring extensive training to form stable internal representations. In contrast, the human brain exhibits exceptional ability to instantaneously map new words to their referents, a process known as “fast mapping”, considered a fundamental mechanism underlying symbol acquisition in early ontogeny. This review provides an overview of neurocognitive research on rapid symbolic learning and examines recent advances in computational modeling approaches aimed at replicating this capability. Models constrained by neurobiological principles known to exist in the human brain are discussed, providing a first step toward neural- and cortical-level explanations of rapid symbolic learning and opening new venues for identifying the neural mechanisms underpinning rapid word acquisition. Archiving these advances may be particularly relevant for guiding the development of sustainable, energy-efficient architectures. A major desideratum from a linguistic and pragmatic perspective involves investigating the neural basis of fast mapping across diverse communicative and pragmatic contexts, an area where current models still fall short.

ScalableAligner: a fast NGS mapping tool for shared-memory system

Metric learning aims to learn a discriminative metric space, where samples of the same class stay close, and those of different classes far apart. Existing classical metric learning methods based on linear transformation have limited learning performance due to the low representation capability. Although deep metric learning learns nonlinear mappings, the training may come across convergence issues and be unstable. Additionally, many classical metric learning algorithms suffer from long computational time for iterative optimization especially when data dimension is high. Deep metric learning also requires high training cost. To learn a metric space more efficiently and effectively, this article proposes a novel broad metric learning (BML) model, which learns the data transformation by training a broad network. BML maps input data to a broad feature space by fast and convenient nonlinear feature mapping based on random weights, and learns a linear transformation to a discriminative output space. Intraclass distance is reduced by minimizing the distance between data and their class-specific reference points in the target space. The hard-triplet distance learning (HDL) is proposed to learn the distance of hard positive and negative sample pairs, which enhances the intraclass compactness and interclass separation. Closed-form solutions are adopted to solve the optimization problems efficiently when learning the linear transformation. Experiments are conducted on nine datasets to verify the efficiency and effectiveness of BML. BML learns fast and achieves high classification and clustering accuracies in the learned data space.

Momentum microscopy with combined hemispherical and time-of-flight electron analyzers at the soft X-ray beamline I09 of the diamond light source.

Successful word learning requires children to pay attention to corresponding auditory and visual input during naming events. However, differences in autistic children's visual attention that restrict their intake of information may impact encoding of novel word-referent associations in memory. This study investigated differences in autistic and neurotypical children's visual attention to stimuli, and whether these differences predicted referent selection and retention accuracy. Fifteen autistic (Mage = 91.87 months) and sixteen neurotypical (Mage = 52.31 months) children matched on receptive vocabulary (Mage autistic children = 53.27 months; Mage neurotypical children = 60.31) used a touch-screen computer to fast map novel words associated with animals (high-interest stimuli) and objects (neutral-interest stimuli). Retention was assessed after 5minutes and 24hours. Children's frequency and duration of looking towards targets was recorded directly via multiple cameras. Neurotypical children spent longer looking at targets during referent selection than autistic children. Autistic children looked at targets significantly more frequently than neurotypical children across word learning stages, and more frequently at targets in the animal condition at 5-minute retention. In-trial visual attention predicted response accuracy across word learning stages for both groups. Visual attention at referent selection also predicted 5-minute and 24-hour retention accuracy for both groups. Visual input during initial encoding influences children's likelihood of successfully forming long-term word-referent representations, indicating strong relationships between attention and learning accuracy. Moreover, population differences in visual attention may not have a detrimental impact on autistic children's word learning under experimental conditions when expectations are based on receptive vocabulary.

The mechanical properties of biological fluids serve as early indicators of disease, offering valuable insights into complex physiological and pathological processes. However, the existing technologies struggle to achieve high-throughput measurement, limiting their widespread applications in disease diagnosis. Here, we propose laser-emission vibrational microscopy of microdroplets for high-throughput measurement of the intrinsic mechanical properties of fluids. The microdroplet array supporting high Q-factor (104) whispering gallery modes (WGM) lasing was massively fabricated on a superhydrophobic surface with inkjet printing. Ultrasound was employed to actuate the mechanical vibrations of the microdroplets, and the vibration amplitude was quantified using time-resolved laser spectra. We found that the stimulus-response of the laser emission is strongly dependent on the liquid viscosity. Fast mapping of the microdroplets’ viscosities was achieved by stage scanning. High-throughput screening of hyperlipidemia disease was also demonstrated by performing over 2000 measurements within 25 min. Thanks to the small volume of the microdroplets, a single drop of blood can support over seven million measurements. The high-throughput ability and small sample consumption make it a promising tool for clinical diagnoses based on mechanical properties.

Abstract Our goal is to show that both the fast and slow versions of the triangle map (a type of multi-dimensional continued fraction algorithm) in dimension n are ergodic, resolving a conjecture of Messaoudi, Noguiera, and Schweiger [Ergodic properties of triangle partitions. Monatsh. Math.157 (2009), 283–299]. This particular type of higher dimensional multi-dimensional continued fraction algorithm has recently been linked to the study of partition numbers, with the result that the underlying dynamics has combinatorial implications.

Generative models cover various application areas, including image and video synthesis, natural language processing and molecular design, among many others1-11. As digital generative models become larger, scalable inference in a fast and energy-efficient manner becomes a challenge12-14. Here we present optical generative models inspired by diffusion models4, where a shallow and fast digital encoder first maps random noise into phase patterns that serve as optical generative seeds for a desired data distribution; a jointly trained free-space-based reconfigurable decoder all-optically processes these generative seeds to create images never seen before following the target data distribution. Except for the illumination power and the random seed generation through a shallow encoder, these optical generative models do not consume computing power during the synthesis of the images. We report the optical generation of monochrome and multicolour images of handwritten digits, fashion products, butterflies, human faces and artworks, following the data distributions of MNIST15, Fashion-MNIST16, Butterflies-10017, Celeb-A datasets18, and Van Gogh's paintings and drawings19, respectively, achieving an overall performance comparable to digital neural-network-based generative models. To experimentally demonstrate optical generative models, we used visible light to generate images of handwritten digits and fashion products. In addition, we generated Van Gogh-style artworks using both monochrome and multiwavelength illumination. These optical generative models might pave the way for energy-efficient and scalable inference tasks, further exploiting the potentials of optics and photonics for artificial-intelligence-generated content.

Comparison of Deep Learning Models for fast and accurate dose map prediction in Microbeam Radiation Therapy.

Rapid acquisition of new words and construction of large vocabularies is a unique capacity of developing human brain. This process is to a large degree mediated by a neurocognitive mechanism known as «fast mapping» (FM) which allows the child to quickly map new words onto neural representations after even a single exposure to them, using context-driven inference. However, the neurophysiological bases of this mechanism are still poorly understood. To address this open question, we used event-related potentials (ERPs) to investigate brain dynamics elicited by novel words following a single-shot audiovisual semantic learning task and to estimate cortical underpinnings of this process in healthy preschool children. We found that a single presentation of novel words in association with novel objects leads to a decrease in the brain's activation, registered as an early N400 effect for newly learnt word forms, indicating rapid lexicosemantic memory trace formation in the developing brain. Interestingly, source analysis indicated this effect to be chiefly underpinned by activity modulations in the right-hemispheric temporal cortices, indicating their involvement in speech processing at an early age (known to be diminished later in life). Overall, current findings provide the electrophysiological evidence of the specific mechanism in the developing brain that promotes rapid integration of novel word representations into neocortical lexicosemantic networks after a single exposure, subserving efficient native word acquisition and mastering the mother tongue.