搜索结果：IEEE access : practical innovations, open solutions

Large-scale functional magnetic resonance imaging (fMRI) datasets provide exciting opportunities for understanding and improving brain health. Data-driven techniques such as independent component analysis (ICA) and independent vector analysis (IVA) have been attractive solutions for multi-subject fMRI analysis, as the extraction of functional connectivity networks is the key step in many studies. Constrained versions of ICA and IVA help significantly improve performance and interpretability, but their comparative advantages and the practical impact of their different formulations remain unclear. This work addresses this gap by conducting a comprehensive comparison of three state-of-the-art constrained algorithms: threshold-free constrained IVA (tf-cIVA), adaptive-reverse constrained IVA (ar-cIVA), and adaptive-reverse constrained ICA (ar-cEBM). These methods differ significantly in how they leverage the cross-subject information (joint processing of IVA versus the subject-wise approach of constrained ICA) and in their definitions of the closeness with the references (Lagrangian-based adaptive thresholding versus a threshold-free regularization term). We perform a multidimensional comparison among these methods using multiple metrics such as reproducibility, scalability, alignment with references, connectivity, and consistency on a multi-site fMRI dataset of 429 subjects. Our results reveal replicability across the three methods regarding their spatial correlation with the references and identification of biomarkers, as well as distinct trade-offs in other aspects: tf-cIVA excels in reproducibility and produces highly structured temporal functional network connectivity (FNC), making it a strong candidate for dynamic or connectivity-based analyses. Meanwhile, ar-cIVA demonstrates the greatest sensitivity to group differences in spatial FNC, suggesting its utility for identifying spatial biomarkers. Finally, ar-cEBM, via its subject-wise approach, offers superior computational scalability for large datasets. Surprisingly, despite not jointly modeling cross-subject information, ar-cEBM produces more stable spatial maps across subjects, suggesting its flexible density matching may be more critical for group consistency than the joint-processing framework itself. Therefore, besides providing a complete picture, the work provides practical guidance, indicating that the algorithm choice might depend on the specific research question.

Thin-film polyimide neural probes have shown great promise in neuroscience but remain difficult to clinically translate due to the unavailability and lack of customizability of commercially available medical-grade polyamic acids. We present an open-source, end-to-end platform for synthesizing BPDA-pPDA-based polyimide from a custom polyamic acid and translating it into microfabricated thin-film neural interfaces. The approach combines accessible polymer chemistry with a streamlined MEMS-compatible fabrication process to produce flexible, biocompatible depth and surface electrode arrays with high thermal stability, chemical inertness, and low moisture uptake. Devices were validated through benchtop characterization, ISO 10993-11 systemic toxicity testing, and in vivo electrophysiology, both acute and semi-chronic, in wild-type and laforin knockout epileptic mice. The arrays reliably captured high-quality multi- and single-unit activity, as well as spontaneous epileptiform discharges, over implantation periods of up to 12 days. By demonstrating a customizable, end-to-end platform for synthesizing and fabricating thin-film polyimide neural electrodes, and by mimicking human neurosurgical workflows through depth, surface, and semi-chronic studies in mice, this work underscores the translational potential of polyimide-based neural microelectrodes and provides a practical pathway to accelerate clinical adoption.

Falls represent a critical health hazard for elderly individuals, often leading to severe injuries and decreased quality of life. While existing fall detection systems predominantly rely on threshold-based algorithms or conventional machine learning approaches like CNNs and RNNs, they frequently exhibit limitations in capturing the complex temporal dependencies and spatial features inherent in fall data. This paper presents a novel multi-head attention transformer architecture specifically designed for wearable in situ fall detection. Our approach leverages the transformer's self-attention mechanism to effectively model long-range dependencies in inertial measurement unit (IMU) sensor data collected from a shinbone-mounted device. The proposed architecture introduces several key innovations: 1) a position-aware embedding layer that preserves the temporal structure of IMU signals, 2) multi-head self-attention blocks that simultaneously attend to different signal features across varying time scales, and 3) a hybrid feature fusion module that combines global context with local patterns. Experimental results on our human subjects dataset demonstrate the critical importance of addressing dataset imbalance. A baseline model trained on imbalanced data achieved only 48.0% fall sensitivity, rendering it unsafe for practical use. By training our proposed transformer architecture on a large, balanced dataset with a specialized Focal Loss function, we increased fall sensitivity to 90.5% on the test set while maintaining an overall accuracy of 93.1% (validation best epoch: sensitivity 92.5% and specificity 95.8%, see Fig. 3). Furthermore, we present a complete hardware/software co-design, including a custom-designed wearable PCB with optimized power management and a comprehensive strategy for deploying our model on an ultra-low-power Lattice Semiconductor iCE40UP5K FPGA, achieving efficient operation with only 22 mA current consumption. We frame this study as technical feasibility and edge deployment validation; direct elderly clinical validation remains future work.

Large Language Models (LLMs) are being widely adopted in different domains including education, healthcare, and finance. In healthcare domain, LLMs are used in disease diagnosis, abnormality classification, remedy suggestions etc. Multi-abnormality classification of radiology reports is essential in healthcare, medical decision-making, and drug discovery. LLMs are increasingly utilized for such tasks because of their remarkable Natural Language Processing (NLP) capabilities, which streamline medical report processing and reduce administrative burden. To enhance the predictive accuracy, LLMs are often fine-tuned on private, locally available datasets, such as medical reports. However, this practice raises significant privacy concerns, as LLMs are prone to memorizing training data, making them susceptible to data extraction attacks even through query-based access. Additionally, sharing fine-tuned models and weights poses adversarial risks, because they may inadvertently reveal sensitive information about the training data. Despite the growing application of LLMs to medical text classification, privacy-preserving fine-tuning for multi-abnormality classification remains underexplored. To bridge this gap, we propose a differentially private (DP) fine-tuning approach that preserves privacy while enabling multi-abnormality classification from text radiology reports through Low Rank Adaptation (LoRA). Our framework leverages DP optimization techniques to fine-tune LLMs on local patient data while mitigating data leakage risks. To our knowledge, this is the first study to incorporate DP fine-tuning of LLMs for multi-abnormality classification using text-based radiology reports. We used labels generated by a larger LLM to fine-tune a smaller LLM, accelerating inference while maintaining privacy constraints. We conducted extensive experiments on the MIMIC-CXR, and CT-RATE datasets to evaluate DP fine-tuning method across varying privacy regimes, analyzing the privacy-utility trade-off and demonstrating the efficacy of our approach. For instance, on the MIMIC-CXR dataset, our proposed DP-LoRA framework achieves weighted F1-scores of up to 0.89 under moderate privacy budgets ( ϵ = 10 ), approaching the performance of non-private LoRA (0.90) and full fine-tuning (0.96). These results demonstrate that strong privacy protection can be achieved with only moderate performance degradation.

Here, we introduce the ID-GBA (Information Distance Guilt By Association) method to expand highly connected sets of nodes by deploying a novel algorithm for subgraph extension based on the guilt-by-association principle and information distance. In this study, ID-GBA was utilized to expand disease clusters, and identify novel disease genes. We first validate its ability to expand related disease sets from disease/disease graphs built using Open Targets' gene association scores. We then analyze disease/control gene expression networks and show that ID-GBA recaptures known disease genes in nine disease/control graphs. Compared to existing methods such as Random Walk with Restarts and Personalized PageRank, ID-GBA achieves significantly higher Normalized Discounted Cumulative Gain scores, which indicates superior predictive performance at capturing known disease genes. Additionally, unlike other approaches that require users to specify either a threshold parameter or a fixed number of nodes to include in the extended subgraph, ID-GBA includes a built-in, automated, and data-driven thresholding mechanism. These results establish ID-GBA as a novel open-source tool to uncover hidden relationships in gene/gene, disease/disease, and other complex networks.

Using temporal analysis of fMRI (functional Magnetic Resonance Imaging) data, we can characterize dynamic changes in brain connectivity over time. However, dynamic temporal analysis of fMRI data is challenging due to the high dimensionality of the datasets. Another fundamental challenge of dynamic temporal analysis of fMRI is the presence of non-neural artifacts that add sources of variation in the data that are not directly related to brain activity. For example, when data are acquired at different scanners at different temporal sampling rates and later analyzed as a single dataset, we have to contend with different number of image snapshots for different subjects. Also, high-frequency scans lead to more fine-grained temporal snapshotting than low-frequency scans. These factors can obscure true neural signals and lead to inconsistent characterization of dynamic brain connectivity across scans. Existing graph-based solutions often struggle with parameter sensitivity, since their outcomes depend heavily on selecting an arbitrary correlation threshold for defining network edges. In contrast, topological data analysis (TDA) sweeps across all threshold values to track the persistence of connectivity features, making it more robust for capturing fine-grained temporal dynamics. Clustering methods become imperative in this context as they offer a powerful means to uncover underlying structures within the high-dimensional temporal data. We address these challenges by developing a topological data analysis based temporal clustering pipeline targeted for dynamic functional connectivity derived from fMRI datasets that can preserve the dynamics of the temporal datasets and mask out the non-neural variability induced by varying sampling rates. The TDA-based pipeline extracts robust features that are invariant to non-neural noise and uses them to perform temporal clustering. We evaluate our framework by performing temporal clustering of resting-state fMRI-derived dynamic functional connectivity brain networks obtained from 316 subjects, each of whom was scanned thrice using different temporal sampling periods. The efficacy of our TDA-based pipeline is compared against three alternative approaches: direct time-series clustering, PCA-based dimensionality reduction and clustering, and a traditional fully connected network analysis pipeline with MDS-based dimensionality reduction. Additionally, we demonstrate that for a majority of cases, the number of clusters remains consistent for the same subjects scanned at different temporal sampling rates- showcasing the greater robustness of our TDA-based pipeline compared to other pipelines. The TDA pipeline achieved higher overlaps (59 %) in optimal cluster numbers across sampling cohorts, as well as higher pairwise similarity (74-77 %) between subjects' cluster solutions. This indicates that incorporating network topology via TDA enables more robust clustering of temporal fMRI datasets despite changes in sampling rates.Furthermore, we validate our method on a clinical dataset (ADHD-200). The TDA-based pipeline successfully captures consistent clustering patterns across different sites and scanning protocols, with higher stability of cluster assignments (> 80% similarity) and better separation of subject-level dynamics compared to existing approaches. This reinforces the method's robustness in multisite, multi-condition settings. Our results demonstrate that incorporating network topology via TDA significantly enhances the reliability of temporal clustering in fMRI studies, offering a robust framework for studying brain dynamics across heterogeneous acquisition settings.

The intrinsic magnetic susceptibility of noninteracting magnetic nanoparticles is calculated from the measured complex radio-frequency susceptibility of dipolar-coupled particles in colloidal suspension. The susceptibility, derived from the impedance of toroidal specimens in a shorted coaxial transmission line, requires no adjustment for demagnetizing factors. Three established models for interacting superparamagnetic particles - the Weiss molecular-field model, the Onsager model, and the modified mean-field model - are extended to magnetically viscous particles at frequencies above their Néel relaxation frequency. Magnetic figures of merit for nanoparticles are the real and imaginary parts of complex intrinsic susceptibility scaled by particle concentration. The work seeks to promote standard characterization methods for interlaboratory comparison of realistic distributions of magnetic nanoparticles at radio frequencies.

Petri nets are an increasingly used modeling framework for the spread of disease across populations or within an individual. For example, the Susceptible-Infectious-Recovered (SIR) compartment model is foundational for population epidemiological modeling and has been implemented in several prior Petri net studies. While the SIR model is typically expressed as Ordinary Differential Equations (ODEs), with continuous time and variables, Petri nets operate as discrete event simulations with deterministic or stochastic timings. We present the first systematic study of the numerical convergence of two distinct Petri net implementations of the SIRS compartment model relative to the standard ODE. In particular, we introduce a novel deterministic implementation of the SIRS model using variable transition weights in the GPenSIM package and stochastic Petri net models using Spike. We show how rescaling and rounding procedures are critical for the numerical convergence of Petri net SIR models relative to the ODEs, and we achieve a relative root mean squared error of less than 1% compared to ODE simulations for biologically relevant parameter ranges. Our findings confirm that both stochastic and deterministic discrete time Petri nets are valid for modeling SIR-type dynamics with appropriate numerical procedures, laying the foundations for larger-scale use of Petri net models.

Accurate labeling of training data is essential for reliable supervised machine learning, particularly in sensitive applications such as virus classification, autonomous driving, precision manufacturing, and medical diagnostics. However, the labeling process is labor-intensive and error-prone. Even widely used datasets such as MNIST and ImageNet contain numerous mislabeled samples. To address this challenge, we developed a transfer learning-based ensemble method that identifies mislabeled data through majority filtering and consensus filtering using fine-tuned pretrained deep neural networks, including ResNet-50, ResNet-101, VGG-16, EfficientNet, MobileNet, and Inception. Our approach was first validated on the MNIST dataset, where the ensemble detected approximately 751 label inconsistencies, which closely aligns with previously reported estimates of mislabeled samples. Additional experiments with synthetically injected mislabels demonstrated that the method could recover up to 100% of known corrupted labels using majority and consensus voting strategies. The method was then applied to a highly pure adeno-associated virus (AAV) nanopore dataset, where artificial mislabels were introduced for evaluation; the ensemble successfully identified most mislabeled samples and correctly recovered their true labels. Experiments on balanced and unbalanced AAV datasets further showed improved performance on the balanced subset, where all injected mislabels were detected. Compared to classical filtering techniques such as KNN, k-means clustering, and advanced machine learning-based mislabel detection (e.g., DivideMix), the proposed ensemble method demonstrated superior accuracy, stability, and true-label recovery, establishing it as a strong mislabel detection framework-well-suited for complex, fine-grained datasets such as nanopore signals and other biological measurement data.

Neurostimulation therapies are often applied as an alternative method to pharmaceutical treatment for pain relief. This paper demonstrates a programmable stimulator for analgesic nerve stimulation through invasive electrodes. It provides two modes of operation: constant-current biphasic and capacitor-coupled biphasic. The stimulator is integrated on a 50 × 30 mm Printed Circuit Board (PCB) and can generate up to ±1.6 mA current pulses in steps of 6.3 μ A . A custom STM32 microcontroller library is used to tune the pulse duration range from 1 μ s to 10 s and stimulation frequency range between 0.1 Hz to 100 kHz. An Active Impedance Monitoring (AIM) unit in the PCB measures the frequency response from 1 Hz to 10 kHz with 8 samples per decade every 30 minutes. The electrode resistance, tissue resistance and RC time constant of the electrode-tissue system could be calculated from the frequency response curve, generated by an input square wave stimulation current at different frequencies. In different measurements with discrete RC circuit components, the PCB could calculate the impedance values with >90% precision. A novel charge-balancing algorithm implemented on the STM32 microcontroller could match the positive and negative charge within a 5% error. Through online adjustment of stimulation parameters using active impedance monitoring to achieve charge balancing, the proposed programmable stimulation system extends the electrode lifetime for more than 288 million continuous pulses in vitro studies. The in vitro studies clearly demonstrated the efficacy of charge balanced stimulation that enable the electrodes to last around 3 times longer with negligible degradation.

Accurate liver tumor segmentation is a critical component of clinical assessment, forming the basis for treatment planning, therapy response monitoring, prognostic assessment, and the delivery of precision medicine. However, in real-world clinical practice, this task remains particularly challenging. The intrinsic diversity of liver tumors -manifested in variations of shape, texture, size, and location-combined with the similarity of neighboring organs, indistinct tumor boundaries, and inconsistencies in image acquisition conditions, makes accurate liver lesion segmentation particularly difficult. In clinical practice, traditional segmentation methods are often used due to their interpretability and lower computational requirements. These approaches are labor-intensive and time-consuming, especially when dealing with 3D medical images, making them impractical for large-scale or real-time applications. In recent years, deep learning (DL) models have gained considerable attention for automating liver lesion segmentation. In this comprehensive review, we analyze over 100 research papers focused on DL-based segmentation of liver tumors from computed tomography (CT) images. This survey examines these studies across multiple dimensions, including input data, model architecture, and evaluation metrics. By exploring both pioneering contributions and emerging trends, we highlight the impact of various methodological choices and address the associated limitations of current approaches.

Alcohol intoxication increases Blood Alcohol Content (BAC) and impairs cognitive, motor, and psychomotor functions, and contributed to over 30% of motor vehicle fatalities in 2017. Traditional methods for detecting Driving Under Influence (DUI), such as breathalyzers and blood tests, are invasive, require external hardware, and are unsuitable for continuous monitoring. Passive approaches such as gait analysis from smartphone sensor data offer a non-invasive solution for detecting impairment and enhancing road safety. Prior work has explored traditional machine learning (ML) and some Deep Learning (DL) approaches but have limitations such as analyzing on handcrafted features and facing challenges such as class imbalance and gait variability. This paper proposes a novel DL framework for passive alcohol intoxication detection from smartphone accelerometer data. A subject-level pre-processing pipeline was employed to address inter- and intra-subject variability, including stratified splitting, low-pass filtering, sliding window segmentation, and random oversampling to mitigate class imbalance. We propose the Multichannel Hybrid 1D-CNN-Attention-BiLSTM (MC-Hybrid) model, which extracts short-term features via parallel 1D-CNNs, uses a self-attention mechanism to increase weights on predictive patterns, and utilizes a bidirectional LSTM to model temporal dependencies. Rigorous evaluation includes comparison to a comprehensive set of ML and DL baselines, investigation of multiple sensor data window sizes, attention types, and ablation studies. MC-Hybrid achieved 93% accuracy and an F1-score of 0.8653, outperforming the state-of-the-art by 9.5% and all baselines by 9.0%. Self-attention resulted in a 2% performance gain over other attention mechanisms, demonstrating its effectiveness in DUI detection. The proposed method could be a practical, non-invasive approach to detect alcohol impairment.

Integration of multi-task brain imaging data is crucial for furthering our understanding of neurodevelopment. By utilizing functional magnetic resonance imaging (fMRI) data and machine learning methods such as graph neural networks to measure brain activity, one can efficiently capture complex brain network interactions. In this work, we introduce the Multi-Task Joint Non-Negative Matrix Factorization Graph Attention Network (MJNMF-GAT) to investigate fMRI data of multiple tasks with the goal of determining age-related differences in adolescence. Our framework integrates joint non-negative matrix factorization for extracting shared latent features across tasks with graph attention networks to model brain connectivity patterns. The model is capable of integrating data from different fMRI tasks to enhance both predictive performance and network interpretation. We applied GNNExplainer for model interpretability and identified key subnetworks that contribute to prediction tasks. The proposed MJNMF-GAT model demonstrates superior results over existing methods, achieving a root mean square error (RMSE) of 1.9212 ± 0.1742, mean absolute error (MAE) of 1.5368 ± 0.1469, and correlation of 0.8032 ± 0.0469 in an age prediction task on the Philadelphia Neurodevelomental Cohort (PNC). The model also identifies the critical functional connections that change with the different stages of adolescent neurodevelopment. Our approach improves the predictive accuracy and, more importantly, interprets informative brain connectivity patterns, especially for detecting those important sub-networks responsible in age prediction. The MJNMF-GAT framework opens a new avenue for analyzing multi-task neuroimaging studies, which helps identify significant brain networks and develop a deeper understanding of functional connectivity changes in adolescence.

Three-dimensional (3D) ultrasound (US) imaging is widely used for real-time, non-ionizing, and cost-effective medical diagnostics. However, using a one-dimensional (1D) transducer often results in limited elevational resolution due to the inherent beam thickness. In this paper, we introduce an elevational Synthetic Aperture Focusing (SAF) algorithm specifically designed for rotational 3D US imaging. Unlike previous methods requiring channel data, our approach operates on in-plane beamformed radio-frequency (RF) data, making it more accessible on many commercial scanners. Through simulations and experiments, we demonstrate significant improvements in elevational resolution (up to 96.4%) and contrast (up to 274.7%). These findings highlight the potential of the proposed algorithm to enhance both research and clinical applications of rotational 3D US imaging.

Histopathological image analysis presents unique challenges due to subtle inter-class variations, complex tissue structures, and high intra-class heterogeneity. Convolutional neural networks (CNNs) have traditionally dominated this domain by leveraging strong inductive biases toward locality, while vision transformers (ViTs) have recently emerged as a promising alternative due to their ability to model long-range dependencies. However, ViTs often struggle to capture fine-grained spatial patterns critical to histopathology, particularly when trained on limited data. In this work, we propose a knowledge distillation (KD) framework that transfers spatial and semantic knowledge from a CNN teacher to a ViT student while explicitly addressing representation and spatial misalignment between heterogeneous architectures. We analyze the limitations of conventional feature based distillation methods, which commonly rely on naive layer matching or pixel wise feature alignment, leading to suboptimal knowledge transfer. To overcome these issues, we introduce a principled layer alignment strategy based on representation similarity analysis using centered kernel alignment (CKA) and kernel canonical correlation analysis (KCCA). These measures enable architecture agnostic comparison of internal representations and facilitate the identification of semantically aligned stages between teacher and student models without enforcing strict spatial correspondence. Furthermore, we propose a stage-level representation alignment mechanism that preserves semantic consistency across representations while accommodating architectural differences between CNNs and ViTs, ensuring effective knowledge transfer at both representation and spatial levels. Extensive experiments on the BreakHis histopathological dataset demonstrate consistent performance improvements, particularly across fine-grained subtypes. Notably, our method achieves 96.87% using a ViT-Large student with a ResNet152 teacher on the patient-wise split setting and 98.82% accuracy on the image-wise subtype split, outperforming the best-performing state-of-the-art KD-based methods by 4% showing the promise of cross architecture KD in computational pathology.

Spiking neural networks (SNNs) hold significant promise as energy-efficient alternatives to conventional artificial neural networks (ANNs). However, SNNs require computations across multiple timesteps, resulting in increased latency, heightened energy consumption, and additional memory access overhead. Techniques to reduce SNN latency down to a unit timestep have emerged to realize true superior energy efficiency over ANNs. Nonetheless, this latency reduction often comes at the expense of noticeable accuracy degradation. Therefore, achieving an optimal balance in the tradeoff between accuracy and energy consumption by adjusting the latency of multiple timesteps remains a significant challenge. This work leverages an additional dimension to enhance the accuracy-energy tradeoff space using a novel one-hot multi-level leaky integrate-and-fire (M-LIF) neuron model. The proposed one-hot M-LIF model represents the inputs and outputs of hidden layers as a set of one-hot binary-weighted spike lanes to find better tradeoff points while still being able to model conventional SNNs. For image classification on static datasets, we demonstrate one-hot M-LIF SNNs outperform iso-architecture conventional LIF SNNs in terms of accuracy (2% higher than VGG16 SNN on ImageNet) while still being energy-efficient (20&#xd7; lower energy than VGG16 ANN on ImageNet). For dynamic vision datasets, we demonstrate the ability of M-LIF SNNs to reduce latency by 3&#xd7; compared to conventional LIF SNNs while limiting accuracy degradation (< 1%).

Patient-reported outcomes (PROs), directly captured from cancer patients undergoing radiation therapy, play a crucial role in guiding clinicians' counseling on treatment-related toxicities. Accurate prediction and assessment of symptoms and health status linked to PROs are essential for improving clinical decision-making and planning post-treatment support as patients transition into survivorship. However, raw PRO data collected in clinical settings presents two inherent challenges, including data sparsity (due to incomplete item responses) and imbalanced toxicity distributions. These factors complicate predictive modeling. This study investigates advanced machine learning techniques to address these challenges by predicting outcomes such as pain and sleep disturbances using PRO datasets from a cancer therapy center. We implement advanced classifiers (i.e., RF, XGBoost, GB, SVM, MLP-Bagging, and LR) for multi-class imbalance tasks across three cancers. To address the minority cases, we apply oversampling while preserving class ratios. Experimental results demonstrate RF and XGBoost's strong generalization, highlighting their utility in categorizing post-therapy severity levels for clinical decision support.

Transcranial focused ultrasound (tFUS) has emerged as a promising non-invasive modality for precision neuromodulation. However, the heterogeneous acoustic properties of the skull often induce phase aberrations that shift the ultrasound focus and compromise energy delivery. In this study, we developed and validated a phase-reversal based aberration correction method to enhance the targeting specificity of tFUS using a 128-element random array ultrasound transducer. Individual head models were constructed from T1-weighted magnetic resonance (MR) images and corresponding pseudo-computed tomography (pCT) data to accurately represent subject-specific skull geometries and the targeted left V5 (V5L) region. Acoustic simulations were conducted with the k-Wave toolbox by first acquiring free-field pressure waveforms and then recording the aberrated waveforms in the presence of the skull. The phase differences between these conditions were used to compute corrective delays for each transducer element. Quantitative evaluation using metrics such as focal overlap with the target region, axial focal positioning, and the delivered ultrasound energy demonstrated significant improvements: the overlap volume increased by 98.70%, mean axial positioning errors were reduced by up to 14.36%, and energy delivery to the target improved by 17.58%. We further demonstrated that the proposed approach outperforms the conventional ray-tracing methods. The results show that phase-reversal based aberration correction markedly increases the spatial targeting accuracy of tFUS and enhances the efficiency of focused ultrasound energy deposition for the customized random array transducer, paving a way for effective and personalized non-invasive neuromodulation therapies.

Given GPS points on a transportation network, the goal of the Quad-tree Based Driver Classification (QBDC) problem is to identify whether drivers have Mild Cognitive Impairment (MCI). The QBDC problem is challenging due to the large volume and complexity of the data. This paper proposes a quad-tree based approach to the QBDC problem by analyzing driving patterns using a real-world dataset. We propose a geo-regional quad-tree structure to capture the spatial hierarchy of driving trajectories and introduce new driving features representation for input into a convolutional neural network (CNN) for driver classification. The experimental results demonstrate the effectiveness of the proposed algorithm, achieving an F1 score of 95% that significantly outperforms the baseline models. These results highlight the potential of geo-regional quad-tree structures to extract interpretable features and describe complex driving patterns. This approach offers significant implications for driver classification, with the potential to improve road safety and cognitive health monitoring.