3D Input Research Articles

Subtask-Based Usability Evaluation of Control Interfaces for Teleoperated Excavation Tasks

This study aims to experimentally determine the most suitable control interface for different subtasks in the teleoperation of construction robots in a simulation environment. We compare a conventional lever-based rate control interface (“Rate-lever”) with two alternative methods: rate control (“Rate-3D”) and position control (“Position-3D”), both using a 3D positional input device. In the experiments, participants operated a construction machine in a virtual environment and evaluated the control interfaces across three tasks: sagittal plane excavation, turning, and continuous operation. The results revealed that “Position-3D” outperformed others for sagittal excavation, while both “Rate-lever” and “Rate-3D” were more effective for turning. Notably, “Position-3D” and “Rate-3D” can be implemented on the same input device and are easily integrated. This feature offers the possibility of a hybrid-type interface suitable for operators to obtain optimized performance in sagittal and horizontal tasks.

Voxel-wise segmentation for porosity investigation of additive manufactured parts with 3D unsupervised and (deeply) supervised neural networks

AbstractAdditive Manufacturing (AM) has emerged as a manufacturing process that allows the direct production of samples from digital models. To ensure that quality standards are met in all samples of a batch, X-ray computed tomography (X-CT) is often used in combination with automated anomaly detection. For the latter, deep learning (DL) anomaly detection techniques are increasingly used, as they can be trained to be robust to the material being analysed and resilient to poor image quality. Unfortunately, most recent and popular DL models have been developed for 2D image processing, thereby disregarding valuable volumetric information. Additionally, there is a notable absence of comparisons between supervised and unsupervised models for voxel-wise pore segmentation tasks. This study revisits recent supervised (UNet, UNet++, UNet 3+, MSS-UNet, ACC-UNet) and unsupervised (VAE, ceVAE, gmVAE, vqVAE, RV-VAE) DL models for porosity analysis of AM samples from X-CT images and extends them to accept 3D input data with a 3D-patch approach for lower computational requirements, improved efficiency and generalisability. The supervised models were trained using the Focal Tversky loss to address class imbalance that arises from the low porosity in the training datasets. The output of the unsupervised models was post-processed to reduce misclassifications caused by their inability to adequately represent the object surface. The findings were cross-validated in a 5-fold fashion and include: a performance benchmark of the DL models, an evaluation of the post-processing algorithm, an evaluation of the effect of training supervised models with the output of unsupervised models. In a final performance benchmark on a test set with poor image quality, the best performing supervised model was UNet++ with an average precision of 0.751 ± 0.030, while the best unsupervised model was the post-processed ceVAE with 0.830 ± 0.003. Notably, the ceVAE model, with its post-processing technique, exhibited superior capabilities, endorsing unsupervised learning as the preferred approach for the voxel-wise pore segmentation task.

Speeding up echocardiographic examination by automated generation of 2D imaging views from 3D volumetric data using machine learning

Abstract Background 3D echocardiographic data contain infinite 2D echocardiographic planes generated by multiplanar reconstruction, but the process is labor-intensive and time-consuming. Artificial intelligence (AI) utilizing cutting-edge machine learning techniques may improve the accuracy and efficiency of the process. In the study, we developed and validated an AI software that automatically reconstructs 7 2D views recommended by the American Society of Echocardiography from 3D echocardiographic data. Method Convolutional neural network (CNN) architecture utilizing Spatial Configuration Net was used as the machine learning model. We identify a minimum of 3 key feature landmarks on each 2D cross-sectional plane, with the 3D coordinates representing 2D cross-sectional planes accordingly. Subsequently, these landmarks are accurately labeled within a 3D space. 31 landmarks were identified across 7 2D cross-sectional planes and a total of 472 3D volume data were labelled. The performance of the model is evaluated by Root Mean Square Error (RMSE) of the predicted landmark coordinates. Results The mean RMSE over 31 predicted landmarks evaluated in testing is 2.70% for the input 3D volume dimensions. The average time used to generate the 7 2D imaging views was 19.91±3.19 seconds (figure). Conclusion The AI algorithm can efficiently convert 3D echocardiographic data into guideline-recommended 2D images, which shortens the length of echo examination but still preserves the accuracy.AI generation of 2D views from 3D echo

Dynamic MRI interpolation in temporal direction using an unsupervised generative model

PurposeCardiac cine magnetic resonance imaging (MRI) is an important tool in assessing dynamic heart function. However, this technique requires long acquisition time and long breath holds, which presents difficulties. The aim of this study is to propose an unsupervised neural network framework that can perform cardiac cine interpolation in time, so that we can increase the temporal resolution of cardiac cine without increasing acquisition time. MethodsIn this study, a subject-specific unsupervised generative neural network is designed to perform temporal interpolation for cardiac cine MRI. The network takes in a 2D latent vector in which each element corresponds to one cardiac phase in the cardiac cycle and then the network outputs the cardiac cine images which are acquired on the scanner. After the training of the generative network, we can interpolate the 2D latent vector and input the interpolated latent vector into the network and the network will output the frame-interpolated cine images. The results of the proposed cine interpolation neural network (CINN) framework are compared quantitatively and qualitatively with other state-of-the-art methods, the ground truth training cine frames, and the ground truth frames removed from the original acquisition. Signal-to-noise ratio (SNR), structural similarity index measures (SSIM), peak signal-to-noise ratio (PSNR), strain analysis, as well as the sharpness calculated using the Tenengrad algorithm were used for image quality assessment. ResultsAs shown quantitatively and qualitatively, the proposed framework learns the generative task well and hence performs the temporal interpolation task well. Furthermore, both quantitative and qualitative comparison studies show the effectiveness of the proposed framework in cardiac cine interpolation in time. ConclusionThe proposed generative model can effectively learn the generative task and perform high quality cardiac cine interpolation in time.

Schwarzschild modelling of barred s0 galaxy NGC 4371

ABSTRACT We apply the barred Schwarzschild method developed by Tahmasebzadeh et al. (2022) to a barred S0 galaxy, NGC 4371, observed by IFU instruments from the TIMER and ATLAS3D projects. We construct the gravitational potential by combining a fixed black hole mass, a spherical dark matter halo, and stellar mass distribution deprojected from 3.6 μm S$^4$G image considering an axisymmetric disc and a triaxial bar. We independently modelled kinematic data from TIMER and ATLAS3D. Both models fit the data remarkably well. We find a consistent bar pattern speed from the two sets of models with $\Omega _{\rm p} = 23.6 \pm 2.8 \, \mathrm{km \, s^{-1} \, kpc^{-1} }$ and $\Omega _{\rm p} = 22.4 \pm 3.5 \, \mathrm{km \, s^{-1} \, kpc^{-1} }$, respectively. The dimensionless bar rotation parameter is determined to be $\mathcal {R} \equiv R_{\rm cor}/R_{\rm bar}=1.88 \pm 0.37$, indicating a likely slow bar in NGC 4371. Additionally, our model predicts a high amount of dark matter within the bar region ($M_{\rm DM}/ M_{\rm total}$$\sim 0.51 \pm 0.06$), which, aligned with the predictions of cosmological simulations, indicates that fast bars are generally found in baryon-dominated discs. Based on the best-fitting model, we further decompose the galaxy into multiple 3D orbital structures, including a BP/X bar, a classical bulge, a nuclear disc, and a main disc. The BP/X bar is not perfectly included in the input 3D density model, but BP/X-supporting orbits are picked through the fitting to the kinematic data. This is the first time a real barred galaxy has been modelled utilizing the Schwarzschild method including a 3D bar.

ESMformer: Error-aware self-supervised transformer for multi-view 3D human pose estimation

Multi-view 3D human pose estimation (HPE) currently faces several key challenges. Information from different viewpoints exhibits high variability due to complex environmental factors, posing difficulties in cross-view feature extraction and fusion. Additionally, multi-view 3D labeled pose data is rather scarce, and the impact of input 2D poses on 3D HPE accuracy has received little attention. To address these issues, we propose an Error-aware Self-supervised transformer framework for Multi-view 3D HPE (ESMformer). Firstly, we introduce a single-view multi-level feature extraction module to enhance pose features in individual viewpoints, which incorporates a novel relative attention mechanism for representative feature extraction at different levels. Subsequently, we develop multi-view intra-level and cross-level fusion modules to exploit spatio-temporal feature dependencies among human joints, and progressively fuse pose information from all views and levels. Furthermore, we explore an error-aware self-supervised learning strategy to reduce the model’s reliance on 3D pose annotations and mitigate the impact of incorrect 2D poses. This strategy adaptively selects reliable input 2D poses based on 3D pose prediction errors. Experiments on three popular benchmarks show that ESMformer achieves state-of-the-art results and maintains cost-effective computational complexity. Notably, ESMformer does not rely on any 3D pose annotations or prior human body knowledge, making it highly versatile and adaptable in practical applications.11The code and models are available at https://github.com/z0911k/ESMformer.

Recognition of Forward Head Posture Through 3D Human Pose Estimation With a Graph Convolutional Network: Development and Feasibility Study.

Prolonged improper posture can lead to forward head posture (FHP), causing headaches, impaired respiratory function, and fatigue. This is especially relevant in sedentary scenarios, where individuals often maintain static postures for extended periods-a significant part of daily life for many. The development of a system capable of detecting FHP is crucial, as it would not only alert users to correct their posture but also serve the broader goal of contributing to public health by preventing the progression of chronic injuries associated with this condition. However, despite significant advancements in estimating human poses from standard 2D images, most computational pose models do not include measurements of the craniovertebral angle, which involves the C7 vertebra, crucial for diagnosing FHP. Accurate diagnosis of FHP typically requires dedicated devices, such as clinical postural assessments or specialized imaging equipment, but their use is impractical for continuous, real-time monitoring in everyday settings. Therefore, developing an accessible, efficient method for regular posture assessment that can be easily integrated into daily activities, providing real-time feedback, and promoting corrective action, is necessary. The system sequentially estimates 2D and 3D human anatomical key points from a provided 2D image, using the Detectron2D and VideoPose3D algorithms, respectively. It then uses a graph convolutional network (GCN), explicitly crafted to analyze the spatial configuration and alignment of the upper body's anatomical key points in 3D space. This GCN aims to implicitly learn the intricate relationship between the estimated 3D key points and the correct posture, specifically to identify FHP. The test accuracy was 78.27% when inputs included all joints corresponding to the upper body key points. The GCN model demonstrated slightly superior balanced performance across classes with an F1-score (macro) of 77.54%, compared to the baseline feedforward neural network (FFNN) model's 75.88%. Specifically, the GCN model showed a more balanced precision and recall between the classes, suggesting its potential for better generalization in FHP detection across diverse postures. Meanwhile, the baseline FFNN model demonstrates a higher precision for FHP cases but at the cost of lower recall, indicating that while it is more accurate in confirming FHP when detected, it misses a significant number of actual FHP instances. This assertion is further substantiated by the examination of the latent feature space using t-distributed stochastic neighbor embedding, where the GCN model presented an isotropic distribution, unlike the FFNN model, which showed an anisotropic distribution. Based on 2D image input using 3D human pose estimation joint inputs, it was found that it is possible to learn FHP-related features using the proposed GCN-based network to develop a posture correction system. We conclude the paper by addressing the limitations of our current system and proposing potential avenues for future work in this area.

Bilinear Distance Feature Network for Semantic Segmentation in PowerLine Corridor Point Clouds.

Semantic segmentation of target objects in power transmission line corridor point cloud scenes is a crucial step in powerline tree barrier detection. The massive quantity, disordered distribution, and non-uniformity of point clouds in power transmission line corridor scenes pose significant challenges for feature extraction. Previous studies have often overlooked the core utilization of spatial information, limiting the network's ability to understand complex geometric shapes. To overcome this limitation, this paper focuses on enhancing the deep expression of spatial geometric information in segmentation networks and proposes a method called BDF-Net to improve RandLA-Net. For each input 3D point cloud data, BDF-Net first encodes the relative coordinates and relative distance information into spatial geometric feature representations through the Spatial Information Encoding block to capture the local spatial structure of the point cloud data. Subsequently, the Bilinear Pooling block effectively combines the feature information of the point cloud with the spatial geometric representation by leveraging its bilinear interaction capability thus learning more discriminative local feature descriptors. The Global Feature Extraction block captures the global structure information in the point cloud data by using the ratio between the point position and the relative position, so as to enhance the semantic understanding ability of the network. In order to verify the performance of BDF-Net, this paper constructs a dataset, PPCD, for the point cloud scenario of transmission line corridors and conducts detailed experiments on it. The experimental results show that BDF-Net achieves significant performance improvements in various evaluation metrics, specifically achieving an OA of 97.16%, a mIoU of 77.48%, and a mAcc of 87.6%, which are 3.03%, 16.23%, and 18.44% higher than RandLA-Net, respectively. Moreover, comparisons with other state-of-the-art methods also verify the superiority of BDF-Net in point cloud semantic segmentation tasks.

Entropy‐driven Progressive Compression of 3D Point Clouds

Abstract3D point clouds stand as one of the prevalent representations for 3D data, offering the advantage of closely aligning with sensing technologies and providing an unbiased representation of a measured physical scene. Progressive compression is required for real‐world applications operating on networked infrastructures with restricted or variable bandwidth. We contribute a novel approach that leverages a recursive binary space partition, where the partitioning planes are not necessarily axis‐aligned and optimized via an entropy criterion. The planes are encoded via a novel adaptive quantization method combined with prediction. The input 3D point cloud is encoded as an interlaced stream of partitioning planes and number of points in the cells of the partition. Compared to previous work, the added value is an improved rate‐distortion performance, especially for very low bitrates. The latter are critical for interactive navigation of large 3D point clouds on heterogeneous networked infrastructures.

RetroSphere: Enabling Low-power and Affordable 3D Input for Lightweight Augmented Reality Devices

RetroSphere introduces a self-contained 6-degree-of-freedom (6DoF) controller tracker, designed for seamless integration with resourceconstrained Augmented Reality (AR) devices such as AR glasses and smartphones. This system employs a tracking mechanism using one to three retro-reflective spheres on a passive controller, thus eliminating the need for any onboard electronics or batteries. Detection is enabled by a stereo pair of mass-produced infrared blob trackers, each outfitted with infrared LED emitters. The system utilizes a lightweight stereo depth-estimation algorithm on an Arduino-compatible ESP32 microcontroller to achieve 6DoF tracking. RetroSphere achieves a tracking accuracy of approximately 96.5%, with positional errors as low as 3.5 cm across a tracking range of 100 cm, while maintaining a power consumption of just about 400 mW. The complete hardware and firmware details for RetroSphere are available as open source on GitHub: https://github.com/AnantaBalaji/RetroSphere

Benchmarking the Performance of the Actuator-Disk Method for Low-Pressure Axial Flow Fan Simulation

Abstract The actuator-disk method is a cost-effective simulation tool that implicitly represents the rotor of a turbomachine using a blade-element approach combined with two-dimensional (2D) airfoil coefficient input data. Actuator-disk models further rely upon empirical coefficient corrections to modify the 2D input data to better mimic physical blade aerodynamic characteristics. However, the fabrication of high-fidelity, general-purpose corrections remains a formidable challenge, so the limits of actuator-disk model accuracy have never been rigorously tested and remain uncertain. This is especially the case for low-pressure axial flow fan models, given the relative lack of empirical corrections conceived specifically for fan rotor simulation. In this study, benchmark performance limits of the actuator-disk method for axial flow fan analysis are explored. The limits of the modeling approach are interrogated by simulating actuator-disk fan models embedded with accurate physical blade data derived from explicit three-dimensional (3D) fan model computations. It is subsequently shown that even with a near-precise coefficient description, the method remains unreliable. However, using an unconventional actuator-disk model formulation that is based directly on 3D blade force inputs, it is demonstrated that the accuracy of conventional models can be noticeably enhanced if the input coefficient data is artificially manipulated. This nonphysical tuning of the input coefficient data is required to compensate for the simplified flow fields produced by the reduced-order modeling approach. The needed coefficient adjustments, referred to as modeling corrections, are subsequently defined and explained alongside the presentation of new realistic performance targets for future actuator-disk fan model variants.

New 2D to 3D Reconstruction of Heterogeneous Porous Media via Deep Generative Adversarial Networks (GANs)

AbstractAccurately characterizing rock microstructures in three dimensions (3D) is crucial for modeling various physical phenomena and estimating rock properties. Despite advancements in 3D imaging, limitations arise from the trade‐off between sample size and resolution, particularly in heterogeneous rocks with multi‐scale features where both high resolution and a large field of view are essential. These challenges have prompted interest in accurate 3D reconstructions from high‐resolution two‐dimensional (2D) images using advanced generative models like generative adversarial networks (GANs). In this study, using scanning electron microscopy and optical microscopy, we acquired 2D images from three orthogonal sections of a Berea sandstone sample. These images were employed to train a modified SliceGAN model, a variant of GANs, for 3D reconstruction. Unlike previous studies utilizing methods for 2D to 3D reconstructions that typically incorporated 3D images in their training, our approach relies exclusively on 2D images. We propose a systematic workflow which enables us to produce 3D reconstructions that closely mirror the original 2D inputs and a 3D X‐ray tomography in terms of structural and morphological characteristics. Additionally, we highlight the importance of input data size and training our model with representative images which enable us to generate diverse reconstructions with transport properties that align with previous studies on Berea sandstone. This underscores the potential of 2D to 3D reconstructions as an effective alternative to multiple X‐ray tomographies, integral for assessing variability in heterogeneous rocks.

Deep learning techniques to process 3D chest CT

The idea of using X–rays and Computed Tomography (CT) images as diagnostic method has been explored in several studies. Most of these studies work with slices of CT image in 2D, requiring less computational capacity and less time to process them than 3D. The processing of volumetric data (the complete CT images in 3D) adds an extra dimension of information. However, the magnitude of the data is considerably larger than working with slices in 2D, so extra computational processing is required. In this study a model capable of performing a classification of a 3D input that represents the volume of the CT scan is proposed. The model is able to classify the 3D input between COVID–19 and Non–COVID–19, but reducing the use of resources when performing the classification. The proposed model is the ResNet–50 model with a new dimension of information added, which is a simple autoencoder. This autoencoder is trained on the same dataset, and a vector representation of each exam is generated and used together with the exams to feed the ResNet–50. To validate the proposal, the same proposed model is compared with and without the autoencoder module that provides more information to the proposed model. The proposed model obtains better metrics than the same model without the autoencoder, confirming that extracting relevant features from the dataset helps improve the performance of the model.

Improving Soil Quality Index Prediction by Fusion of Vis-NIR and pXRF spectral data

Soil quality assessment, as a means to assess the impact of human activities on soil, is of great significance to achieve sustainable development. Proximal sensing offers a rapid and cost-effective means for soil spectral acquisition, which resolves the shortcomings of traditional laboratory analysis and provides a new way for soil quality assessment. However, using a singular spectroscopic technique may not encompass adequate information needed for soil quality assessment. Therefore, in this research, the visible-near-infrared (vis-NIR) and portable X-ray fluorescence (pXRF) spectral data were fused, and the continuous wavelet transform (CWT) was combined with the parallel input 2D convolutional neural network (PI-2D-CNN) algorithm, to enhance the prediction accuracy of soil quality index (SQI). A total of 287 soil samples were collected from the Yarkand-Kashgar River Basin in Xinjiang, China for proximal sensing in the laboratory, and 21 soil physicochemical properties were measured for SQI calculation simultaneously. After preprocessing the vis-NIR and pXRF spectral data with the CWT, partial least squares regression (PLSR) and CNN models were constructed to predict SQI. The results showed that the SQI prediction accuracy based on vis-NIR spectra (coefficient of determination for validation set (R2) = 0.32–0.54) was higher than that based on pXRF spectra (R2 = 0.17–0.36). The prediction accuracy of both vis-NIR and pXRF spectra were improved by CWT combined with CNN. Compared with the traditional spectral data fusion methods (concatenation, sequential and orthogonalised-partial least squares (SO-PLS)), the PI-CNN algorithm demonstrated superior SQI prediction accuracy, and the constructed PI-2D-CNN model obtained the highest prediction accuracy (R2 = 0.67). Overall, this study proved that the combination of CWT and PI-CNN could increase the SQI prediction accuracy. This research will provide a rapid and low-cost approach for large-scale soil quality assessment.

Unsupervised EEG-Based Seizure Anomaly Detection with Denoising Diffusion Probabilistic Models.

While many seizure detection methods have demonstrated great accuracy, their training necessitates a substantial volume of labeled data. To address this issue, we propose a novel method for unsupervised seizure anomaly detection called SAnoDDPM, which uses denoising diffusion probabilistic models (DDPM). We designed a novel pipeline that uses a variable lower bound on Markov chains to identify potential values that are unlikely to occur in anomalous data. The model is first trained on normal data, then anomalous data is input to the trained model. The model resamples the anomalous data and converts it to normal data. Finally, the presence of seizures can be determined by comparing the before and after data. Moreover, the input 2D spectrograms are encoded into vector-quantized representations, which enables powerful and efficient DDPM while maintaining its quality. Experimental comparisons on the publicly available datasets, CHB-MIT and TUH, show that our method delivers better results, significantly reduces inference time, and is suitable for deployment in a clinical environments. As far as we are aware, this is the first DDPM-based method for seizure anomaly detection. This novel approach significantly contributes to the progression of seizure detection algorithms, thereby augmenting their practicality in clinical settings.

Structure tensor-based Gaussian kernel edge-adaptive depth map refinement with triangular point view in images

<span lang="EN-US">Image reconstruction is the process of restoring the image resolution. In 3D image reconstruction, the objects in different viewpoints are processed with the triangular point view (TPV) method to estimate object geometry structure for 3D model. This work proposes a depth refinement methodology in preserving the geometric structure of objects using the structure tensor method with a Gaussian filter by transforming a series of 2D input images into a 3D model. The computation of depth map errors can be found by comparing the masked area/patch with the distribution of the original image's greyscale levels using the error pixel-based patch extraction algorithm. The presence of errors in the depth estimation could seriously deteriorate the quality of the 3D effect. The depth maps were iteratively refined based on histogram bins number to improve the accuracy of initial depth maps reconstructed from rigid objects. The existing datasets such as the dataset tanks and unit (DTU) and Middlebury datasets, were used to build the model out of the object scene structure. The results of this work have demonstrated that the proposed patch analysis outperformed the existing state of the art models depth refinement methods in terms of accuracy.</span>

User-centric design of a 3D search interface for protein-ligand complexes

In this work, we present the frontend of GeoMine and showcase its application, focusing on the new features of its latest version. GeoMine is a search engine for ligand-bound and predicted empty binding sites in the Protein Data Bank. In addition to its basic text-based search functionalities, GeoMine offers a geometric query type for searching binding sites with a specific relative spatial arrangement of chemical features such as heavy atoms and intermolecular interactions. In contrast to a text search that requires simple and easy-to-formulate user input, a 3D input is more complex, and its specification can be challenging for users. GeoMine’s new version aims to address this issue from the graphical user interface perspective by introducing an additional visualization concept and a new query template type. In its latest version, GeoMine extends its query-building capabilities primarily through input formulation in 2D. The 2D editor is fully synchronized with GeoMine’s 3D editor and provides the same functionality. It enables template-free query generation and template-based query selection directly in 2D pose diagrams. In addition, the query generation with the 3D editor now supports predicted empty binding sites for AlphaFold structures as query templates. GeoMine is freely accessible on the ProteinsPlus web server (https://proteins.plus).

Cervic cancer classification using quantum fuzzy set

In this sophisticated world living with CIN cervic cancer is much jeopardy. Cancer is a stochastic (random) process. So, in that CIN in initial stage is not jeopardy. It is totally because of hazardous malign cells. Using its modality in image is selected in existing system only automated classification depends on input image. But in proposed methodology the innovative alludes the jeopardy of CIN cancer is found using the size of /area of nucleus or cytoplasm. This proposed methodology was developed with an algorithm to find CIN area/size. This research work establishes a Cervic Cancer Classification Using Contour Based on Area of Nucleolus and Cytoplasm in Cells (CBANC) which classifies noise spread images into any one of five phases. A similarity measure produces 90% efficiency in proposed system as par with inefficient existing system which fetches us 50%. By pragmatic application it is proved that CBANC with fuzzy is better than Baye’s. This can be accomplished by removing well distinct consistency features and choosing preeminent classifier. Proposed work can extend with 3D input images for future research. It produces mightiest parameter shape and intensity which is very essential for 3D approach. The inference of proposed system can extend the latest classifier engines for more accuracy. It can easily predict more than 90% accuracy will be there. And also derive cancer growing and after therapy for cancer shrinking algorithm will be used for 2D or 3D CIN cancer classification. The outcomes of the proposed methodology CBANC shows that better when compared to the existing methodology like Bayes. It can be implemented in the real world environments of the medical field.

Adaptive design of experiments to fit surrogate Gaussian process regression models allows fast sensitivity analysis of the input waveform for patient-specific 3D CFD models of liver radioembolization

Background and objectivePatient-specific 3D computational fluid dynamics (CFD) models are increasingly being used to understand and predict transarterial radioembolization procedures used for hepatocellular carcinoma treatment. While sensitivity analyses of these CFD models can help to determine the most impactful input parameters, such analyses are computationally costly. Therefore, we aim to use surrogate modelling to allow relatively cheap sensitivity analysis. As an example, we compute Sobol's sensitivity indices for three input waveform shape parameters. MethodsWe extracted three characteristic shape parameters from our input mass flow rate waveform (peak systolic mass flow rate, heart rate, systolic duration) and defined our 3D input parameter space by varying these parameters within 75 %-125 % of their nominal values. To fit our surrogate model with a minimal number of costly CFD simulations, we developed an adaptive design of experiments (ADOE) algorithm. The ADOE uses 100 Latin hypercube sampled points in 3D input space to define the initial design of experiments (DOE). Subsequently, we re-sample input space with 10,000 Latin Hypercube sampled points and cheaply estimate the outputs using the surrogate model. In each of 27 equivolume bins which divide our input space, we determine the most uncertain prediction of the 10,000 points, compute the true outputs using CFD, and add these points to the DOE. For each ADOE iteration, we calculate Sobol's sensitivity indices, and we continue to add batches of 27 samples to the DOE until the Sobol indices have stabilized. ResultsWe tested our ADOE algorithm on the Ishigami function and showed that we can reliably obtain Sobol's indices with an absolute error <0.1. Applying ADOE to our waveform sensitivity problem, we found that the first-order sensitivity indices were 0.0550, 0.0191 and 0.407 for the peak systolic mass flow rate, heart rate, and the systolic duration, respectively. ConclusionsAlthough the current study was an illustrative case, the ADOE allows reliable sensitivity analysis with a limited number of complex model evaluations, and performs well even when the optimal DOE size is a priori unknown. This enables us to identify the highest-impact input parameters of our model, and other novel, costly models in the future.

Epileptic Seizure Detection from Decomposed EEG Signal through 1D and 2D Feature Representation and Convolutional Neural Network

Electroencephalogram (EEG) has emerged as the most favorable source for recognizing brain disorders like epileptic seizure (ES) using deep learning (DL) methods. This study investigated the well-performed EEG-based ES detection method by decomposing EEG signals. Specifically, empirical mode decomposition (EMD) decomposes EEG signals into six intrinsic mode functions (IMFs). Three distinct features, namely, fluctuation index, variance, and ellipse area of the second order difference plot (SODP), were extracted from each of the IMFs. The feature values from all EEG channels were arranged in two composite feature forms: a 1D (i.e., unidimensional) form and a 2D image-like form. For ES recognition, the convolutional neural network (CNN), the most prominent DL model for 2D input, was considered for the 2D feature form, and a 1D version of CNN was employed for the 1D feature form. The experiment was conducted on a benchmark CHB-MIT dataset as well as a dataset prepared from the EEG signals of ES patients from Prince Hospital Khulna (PHK), Bangladesh. The 2D feature-based CNN model outperformed the other 1D feature-based models, showing an accuracy of 99.78% for CHB-MIT and 95.26% for PHK. Furthermore, the cross-dataset evaluations also showed favorable outcomes. Therefore, the proposed method with 2D composite feature form can be a promising ES detection method.