Search for: All records

Background: Traumatic brain injury (TBI) remains a major public health concern, with over 69 million cases annually worldwide. Accurate patient-specific biomechanical modeling is critical for injury risk assessment, but it requires integrating heterogeneous data sources such as volumetric neuroimaging, scalar demographic parameters, and acquisition metadata. Conventional finite element solvers can perform this modeling, yet they remain too computationally expensive for time-sensitive clinical settings. Neural operators have emerged as a promising alternative by learning resolution-invariant mappings between function spaces at orders-of-magnitude faster inference. However, while recent work has introduced parameter-conditioned and multi-input operator architectures for incorporating auxiliary scalar or geometric inputs, the systematic integration of volumetric medical imaging with heterogeneous scalar metadata within an operator learning framework remains underexplored, particularly for biomechanical prediction tasks involving patient-specific anatomical and demographic variability. Objective: This study presents a systematic investigation of multimodal neural operator architectures for brain biomechanics, evaluating strategies for fusing heterogeneous input modalities such as volumetric anatomical imaging, scalar demographic features, and acquisition parameters to predict full-field brain displacement fields from MRE data. Methods: We reformulated TBI modeling as a multimodal operator learning problem and proposed two fusion strategies: field projection for Fourier Neural Operator (FNO) based architectures (broadcasting scalars onto spatial grids) and branch decomposition for Deep Operator Networks (DeepONet) (separate encoding with multiplicative fusion). Four architectures (FNO, Factorized FNO (F-FNO), Multi-Grid FNO (MG-FNO), DeepONet) were extended with multimodal fusion mechanisms and evaluated on 249 in vivo Magnetic Resonance Elastography (MRE) datasets across physiologically relevant frequencies (20 to 90 Hz). Results: DeepONet achieved the highest accuracy on real displacement fields (MSE = 0.0039, 90.0% accuracy) with the fastest inference (3.83 it/s) and fewest parameters (2.09M), while MG-FNO achieved the best performance on imaginary fields (MSE = 0.0058, 88.3% accuracy) with the lowest GPU memory among FNO variants (7.12 GB). No single architecture dominated across all criteria, revealing distinct trade-offs between accuracy, spatial fidelity, and computational cost. Conclusion: The results demonstrate that neural operators augmented with multimodal fusion mechanisms can accurately predict full-field brain displacement from heterogeneous biomedical inputs, with inference times orders of magnitude faster than finite element solvers. The systematic comparison of fusion strategies and architectures provides practical guidance for selecting operator learning approaches in biomedical settings where heterogeneous data integration is required. 

Comprehensive monitoring of cerebral hemodynamics has led to significant advances in Functional Near-Infrared Systems (fNIRS), particularly in terms of hardware design and development of wearable platforms. These advancements have established fNIRS devices as valuable tools in research and clinical practices; however, most existing literature focuses predominantly on clinical applications or high-level system performance. This review provides a rigorous, bottom-up analysis of bio-instrumentation architectures, evaluating the low-level trade-offs in component selection and circuit design that define modern wearable fNIRS performance. In this paper, we have identified and compared key hardware components of modern fNIRS technologies, including optical sensors, signal conditioning elements, control units, power systems, and communication modules. Significant progress has been made in terms of optical tomography, head coverage and conformity, multimodal integration, hyperscanning, motion tolerance, user comfort, and miniaturization. The paper underscores how systems may have unique architectures although they follow the same foundational principle. It also aims to identify the trade-offs existing in current fNIRS devices. Overall, this paper presents an overview of where we stand in terms of fNIRS development and attempts to trace an outline of the next generation of devices. 

Harmful algal blooms (HABs) are increasing in frequency in the Chesapeake Bay, posing risks to marine ecosystems, water quality, and public health. Chlorophyll-a (Chl-a) is a widely used indicator of algal biomass, and satellite observations such as Sentinel-3 Ocean and Land Color Instrument (OLCI) enable large-scale monitoring of bloom dynamics. However, cloud cover and atmospheric interference frequently introduce missing pixels in daily satellite products, reducing temporal continuity and limiting monitoring reliability. Satellite-derived chlorophyll-a (Chl-a) data exhibit substantial missingness, with daily pixel gaps ranging from approximately 52.30% to 100% (mean ≈ 88.95%). This study evaluates spatial interpolation, EOF-based, supervised machine-learning, deep-learning, and convolutional autoencoder approaches for reconstructing missing Chl-a values. Sentinel-3 OLCI Chl-a data from 2023–2024 were used for model training, while data from 2025 served as a temporally independent test set to avoid spatiotemporal leakage. To simulate cloud-induced data gaps, artificial missingness scenarios ranging from 50% to 90% were applied for the Inverse Distance Weighting (IDW) and Data Interpolating Empirical Orthogonal Functions (DINEOF) baseline approaches, while machine-learning, deep-learning, and convolutional autoencoder models were evaluated using real satellite-derived missing observations. The evaluated models include IDW, DINEOF, K-Nearest Neighbors (KNN), Random Forest (RF), Extra Trees (ET), XGBoost, a Long Short-Term Memory (LSTM) network, and a Temporal Data Interpolating Convolutional Autoencoder (Temporal DINCAE). Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), prediction bias, and the coefficient of determination (R2). Results indicate that tree-based ensemble models outperform spatial interpolation and EOF-based methods, with XGBoost achieving the best overall performance (R2 ≈ 0.86; RMSE ≈ 9.61 mg m−3). The LSTM model achieved lower prediction errors (RMSE ≈ 5.87 mg m−3; MAE ≈ 2.16 mg m−3), highlighting the benefit of incorporating temporal dependencies, although with slightly reduced variance capture. The convolutional autoencoder-based Temporal DINCAE model achieved strong reconstruction performance (R2 ≈ 0.84; RMSE ≈ 11.15 mg m−3). Uncertainty quantification shows that Extra Trees tends to underestimate uncertainty with narrower prediction intervals, whereas XGBoost provides better-calibrated but wider intervals. 

Other: Related Article: Anusha Biswas, Jeanette A. Krause, Hairong Guan|2026|CSD Communication||| Abstract: An entry from the Cambridge Structural Database, the world’s repository for small molecule crystal structures. The entry contains experimental data from a crystal diffraction study. The deposited dataset for this entry is freely available from the CCDC and typically includes 3D coordinates, cell parameters, space group, experimental conditions and quality measures. 

Abstract Bayesian optimization (BO) holds promise for accelerating materials science research; however, it faces challenges with high-dimensional inputs and experimental noise in real-world problems. This study addresses these issues by benchmarking batch BO on two synthetic six-variable optimization tasks at varying noise levels: a needle-in-a-haystack task (Ackley function) representing rare materials properties, and a smooth landscape (Hartmann function) simulating process optimization. We evaluate key BO strategies, including acquisition functions, batch-picking methods, and exploration hyperparameter tuning, while presenting a framework for tracking high-dimensional optimization progress. Results show optimization outcomes are highly sensitive to noise levels and landscape shapes. This information enables the design of robust materials optimization campaigns with pre-planned experimental budgets that account for real-world uncertainties. Our methodology facilitates greater BO utilization in experimental materials research, particularly for multi-variable optimization problems, by providing practical guidance for configuring BO campaigns in challenging scientific applications. Graphical abstract 

We present clear evidence that orthophosphate (PO₄^3-), commonly used by drinking water treatment plant (DWTP) operators to reduce lead pipe corrosion, can move beyond its intended domain and enter urban stream networks. This migration influences stream biogeochemistry in ways previously undocumented. In collaboration with a DWTP in Pittsburgh, Pennsylvania, we conducted a study across five urban streams, capturing pre- and post-implementation phases of PO₄^3--based corrosion control. Through comprehensive chemical analyses of nutrients, anions, metals, and nitrate isotopes, as well as nutrient limitation bioassays withCylindrospermopsissp. andRaphidocelis subcapitata, we demonstrate statistically robust increases in streamwater phosphorous concentrations (total dissolved phosphorus: p < 0.0001; total reactive phosphorus and total phosphorus: p < 0.05). These shifts coincide with elevated dissolved metal concentrations, implicating corrosion control byproducts as co-transported constituents. Principal Component Analysis reveals urban stream chemistry is governed by a complex interplay of solutes derived from PO₄^3--treated drinking water, pipe corrosion, mineral weathering, and wastewater. Bioassays confirm that nutrient additions, particularly P and NP, significantly stimulate algal biomass (p < 0.05) withCylindrospermopsissp. showing heightened responsiveness to N additions shortly after PO₄^3-deployment. These findings expose a critical, underrecognized pathway by which drinking water infrastructure contributes to nutrient enrichment and eutrophication in urban aquatic systems. The implications are clear: subsurface infrastructure is not a closed system, and its chemical footprint extends into the urban hydroscape with ecological consequences that warrant closer attention. 

pHigment explores the dual design challenge of creating on-body interfaces that are both environmentally sustainable and capable of on-body chemical sensing. We present a set of single-use cosmetic prototypes—eyeliner, temporary tattoos, and press-on nails—that respond to chemical changes in body and external fluids while being compostable after use. Each prototype is fabricated with skin-safe materials such as using alginate as a biodegradable substrate, and red cabbage powder as the anthocyanin colorimetric biosensor. This project identifies 10 expected affordances that support its use as a single-use biosensing interface. We describe the fabrication process and introduce a system that includes a custom-designed case and a computer vision–based mobile application. The app interprets color changes and outputs corresponding pH values. We report technical evaluations of the materials, system functionality, and compostability. This work contributes to the design of biocosmetic interfaces by demonstrating how ephemeral body-worn form factors can simultaneously serve as both responsive chemical sensors and components of a circular materials strategy.