SEARCH · Engineering Papers

Results for “feature normalization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5

All sources NASA NTRS DOE OSTI

DOE OSTI2022

Interaction of molecular nitrogen with vanadium oxide in the absence and presence of water vapor at room temperature: Near-ambient pressure XPS

Interactions of N2 at oxide surfaces are important for understanding electrocatalytic nitrogen reduction reaction (NRR) mechanisms. Interactions of N2 at the polycrystalline vanadium oxide/vapor interface were monitored at room temperature and total pressures up to 10−1 Torr using Near-Ambient Pressure X-ray Photoelectron Spectroscopy (NAP-XPS). The oxide film was predominantly V(IV), with V(III) and V(V) components. XPS spectra were acquired in environments of both pure N2 and equal pressures of N2 and H2O vapor. In pure N2, broad, partially resolved N1s features were observed at binding energies of 401.0 and 398.7 eV, with a relative intensity of ∼3:1, respectively. These features remained upon subsequent pumpdown to 10−9 Torr. The observed maximum N surface coverage was ∼1.5 × 1013 cm−2—a fraction of a monolayer. In the presence of equal pressures of H2O, the adsorbed N intensity at 10−1 Torr is ∼25% of that observed in the absence of H2O. The formation of molecularly adsorbed H2O was also observed. Density functional theory-based calculations suggest favorable absorption energies for N2 bonding to both V(IV) and V(III) cation sites but less so for V(V) sites. Hartree–Fock-based cluster calculations for N2–V end-on adsorption show that experimental XPS doublet features are consistent with the calculated shake-up and normal, final ionic configurations for N2 end-on bonding to V(III) sites but not V(IV) sites. The XPS spectra of vanadium oxide transferred in situ between electrochemical and UHV environments indicate that the oxide surfaces studied here are stable upon exposure to the electrolyte under NRR-relevant conditions.

Balogun, K.↗

DOE OSTI2020

Learning method and system for separating independent and dependent attacks

Streams of monitoring node signal values over time, representing a current operation of the industrial asset, are used to generate current monitoring node feature vectors. Each feature vector is compared with a corresponding decision boundary separating normal from abnormal states. When a first monitoring node passes a corresponding decision boundary, an attack is detected and classified as an independent attack. When a second monitoring node passes a decision boundary, an attack is detected and a first decision is generated based on a first set of inputs indicating if the attack is independent/dependent. From the beginning of the attack on the second monitoring node until a final time, the first decision is updated as new signal values are received for the second monitoring node. When the final time is reached, a second decision is generated based on a second set of inputs indicating if the attack is independent/dependent.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2020

Ionization induced dynamic alignment of water

Two-body dissociation resulting from strong-field double ionization of water is investigated. Two distinct features are seen in the alignment of the fragment momenta with respect to the laser polarization. One feature shows alignment of the H–OH axis with the laser polarization, while the other indicates polarization alignment normal to the H–OH axis. By analyzing kinematic differences between the OH + /D + and OD + /H + channels of HOD, these two alignment features are shown to result from dissociation from different states in the dication. Only dissociation from one of these states has an alignment dependence consistent with predictions of sequential strong-field tunneling ionization models. Finally, the alignment dependence of dissociation from the other state can only be explained by dynamic alignment launched by the unbending of the molecule during ionization.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2022

Bragg edge imaging (BEI) of B12W and M8N socket sections of the Arecibo telescope

From neutron user principal investigator: We kindly request the public release of three neutron imaging datasets through ONCat. All datasets were collected from two forensic specimens, B12W and M8N, sectioned from zinc-filled steel-wire sockets recovered from the collapsed Arecibo Telescope. The dataset titled “Neutron radiographs of B12W and M8N socket sections of the Arecibo telescope” contains normalized two-dimensional (2D) neutron radiographs of the specimens, showing the geometry and spatial distribution of the steel wires embedded within the zinc matrix, as well as internal features such as voids and cracks. The dataset titled “Neutron computed tomography of B12W and M8N socket sections of the Arecibo telescope” contains normalized 2D neutron projection images acquired over a range of specimen rotation angles for one selected region of each specimen. These projection images were used to reconstruct three-dimensional (3D) tomographic volumes that reveal the embedded-wire geometry and internal defects. The dataset titled “Bragg edge imaging (BEI) of B12W and M8N socket sections of the Arecibo telescope” contains six time-of-flight (TOF) neutron imaging datasets, three from each specimen, acquired at regions of interest selected based on the radiographs. The spatially resolved 2D TOF images show the zinc matrix and embedded steel wires, and the wavelength-dependent neutron transmission data were used to characterize crystallographic texture within the zinc. All components and their condition are in the public domain as they are the property of the National Science Foundation (NSF). The neutron imaging data, part geometries, and detailed forensic information have been widely published in the Arecibo Telescope Collapse Forensic Report by Thornton Tomasetti Engineers and others (NASA report and NASEM report).

Bilheux, Hassina↗

DOE OSTI2026

Identification of mechanisms driving heterogeneous void growth in ductile aluminum

Void growth plays a central role in ductile fracture, yet the specific mechanisms that control this remain obscure. Classical models, such as those proposed by Rice and Tracey in 1969, are able to capture average rates of void growth, but cannot capture the heterogeneity of individual void growth. Building on recent work, the present study employs laboratory-based diffraction contrast tomography and in-situ x-ray computed tomography to investigate the effect of grain structure and other microstructural factors on void growth in an Al-2219 alloy. Crystal plasticity finite element (CP-FE) modeling is used alongside experimental data to evaluate the contributions of local mechanical states, grain orientation, grain size, and neighboring microstructural features. No strong linear relationships are found with any of the considered descriptors and void growth rate. Potential complex nonlinear relationships are explored with the use of a random forest regression model, which identifies initial void volume, void aspect ratio, local normal stress state, local shear stress state, and local equivalent plastic strain (EQPS) as features that most improve void growth rate predictions. The combination of these analyses suggests that these features should be prioritized to improve models of void growth.

Diffraction contrast tomography (DCT)↗

DOE OSTI2021

Process Anomaly Detection for Sparsely Labeled Events in Nuclear Power Plants

An essential aspect of online monitoring, subtle anomaly detection increases the detection lead time for equipment failure and enables a nuclear power plant (NPP) to mitigate unexpected partial or full outages, resulting in significant cost saving to the plant. Once an anomaly is detected by plant staff, its cause and severity are investigated. Because the vast majority of anomalies require some level of investigation, including some that require time-consuming examination, before they are passed over to the engineering organization for further analysis, plants are often equipped with tools to assist the staff in performing anomaly detection. Those tools operate as a black box and are often based on statistical methods that establish sensor correlations using preconfigured mathematical models and flag correlation deviations as anomalies. Due to the number of anomalies detected at a given NPP on a daily basis, a significant number of flagged anomalies usually await examination for days or weeks. A primary cause of this backlog is that the methods used by the tools generate many false positives. Though this is usually attributed to oversensitive model settings due to very narrow normal operation bands, it can also be associated with the model development being inadequate for the process being monitored, or with missing model inputs that could have explained misclassified positives. The performance of anomaly detection tools impacts their plant acceptance and utilization, especially when the effort to address false positives generated by the tool depletes the value or cost saved by using that tool. Thus, means to advance anomaly detection performance have been investigated by the Department of Energy’s Light Water Reactor Sustainability program. Previous and ongoing efforts have targeted unsupervised machine-learning (ML) methods, which do not require the labeling of any data fed into the ML model. By contrast, in supervised anomaly detection methods, every data point is labeled as either a normal or abnormal process condition, and the model is trained to replicate the classification process. Supervised methods usually outperform unsupervised methods, due to the added value in differentiating normal from anomalous states of the monitored process. An NPP’s corrective action program requires it to track and document, via a dedicated report, the resolution of any issues that occur within the plant. Once created, each report is reviewed by a plant screening committee, and several classifications and decisions are made. Recently, a collaborating NPP developed an artificial intelligence and ML-based classifier to categorize a condition report (CR) into classes that can serve to label the data as normal or anomalous. Applying CRs as labels represents a semi-supervised use case. Semi-supervised ML assumes that labels exist for some data points (i.e., labeled anomalies, in this case) but not for the rest. In this effort, semi-supervised ML methods were used to fuse data from CRs with anomaly detection methods in order to test the hypothesis that partially labeled anomalies would improve the accuracy of the anomaly detection methods. Specifically, two methods were used. The first is the deep Semi-supervised Anomaly Detection (deep SAD) method, which can handle labels ranging from fully unsupervised to fully supervised cases. The second is a newly designed ML method developed specifically for this effort and referred to as the high-order feature (HOF)-based method. To evaluate these two methods in controlled environments, synthetic data generators were developed and used. The first datasets used a spring-mass-damper (SMD) system simulator commonly found in mechanical engineering references. This was used to create two use cases: a one- and a three-mass system. Anomalies were introduced by changing the spring and damper coefficients while the system was actuated by random forces. The second datasets used the commercial Dymola-Modelica software to build a simplified nuclear reactor model. Anomalies were added in the form of corrupted sensor readings and/or control commands. The deep SAD method was tested using the SMD system, while the HOF method was tested using both datasets. Application of the deep SAD semi-supervised ML method demonstrated that labels can generate increased confidence in detecting true anomalies. This helped increase the number of true positives and decrease the number of false negatives—something that would aid in addressing the backlog of possible anomalies. Application of the HOF method demonstrated that labels can aid in down selecting from a candidate set of features to a more optimal subset in order to better differentiate between normal and anomalous conditions.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE OSTI2024

Modeling inter- and intra-granular dislocation transport using crystal plasticity

Here, this work presents the development of a crystal plasticity material model that incorporates both dislocation transport within grains and dislocation transfer across grain boundaries. This model has been implemented in the open-source finite element code MOOSE. In addition, a novel geometry-based criterion is developed to determine the direction of dislocation transfer across grain boundaries. The transfer criterion incorporates the geometric features of the grain boundary, such as the grain boundary plane normal, and its misorientation, which is accounted for through the orientation of the incoming and outgoing slip systems. The model is tested with several cases, including a copper single crystal, bi-crystal, and polycrystal. The development of the transfer criterion, implementation of the model, and its application to these test cases are discussed in detail.

36 MATERIALS SCIENCE↗

DOE OSTI2026

Mapping the Perseus galaxy cluster with XRISM: Gas kinematic features and their implications for turbulence

We present extended gas kinematic maps of the Perseus cluster based on a combination of five new XRISM/Resolve pointings observed in 2025 with four performance verification datasets from 2024, totaling a net exposure of 745 ks. To date, Perseus remains the only cluster that has been extensively mapped out to ≃0.7 r 2500 by XRISM/Resolve, while simultaneously offering sufficient spatial resolution to resolve gaseous substructures driven by mergers and active galactic nucleus (AGN) feedback. Our observations cover multiple radial directions and a broad range of dynamical scales, enabling us to characterize the kinematic properties of the intracluster medium up to a scale of ∼500 kpc. In the measurements, we detected high-velocity dispersions (≃300km s −1 ) in the eastern region of the cluster that are spatially coincident with the extended X-ray surface brightness excess and correspond to a nonthermal pressure fraction of ≃7 − 13%. The velocity field outside the AGN-dominant region can be effectively described by a single, large-scale kinematic driver based on the velocity structure function, which statistically favors an energy injection scale of at least a few hundred kpc. The estimated turbulent dissipation energy is comparable to the gravitational potential energy released by a recent merger, implying a significant role of turbulent cascade in the merger energy conversion. In the bulk velocity field, we observed a dipole-like pattern along the east-west direction with an amplitude of ≃ ± 200 − 300 km s −1 , indicating rotational motions induced by the recent merger event. This feature constrains the viewing direction to ≃30° −50° relative to the normal of the merger plane. Our hydrodynamic simulations suggest that Perseus has experienced at least two energetic mergers since redshift z ∼ 1, the most recent of which is associated with the radio galaxy IC310, in agreement with recent SRG/eROSITA findings. This study showcases exciting scientific opportunities for future missions with high-resolution spectroscopic capabilities (e.g., HUBS, LEM, and NewAthena).

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2020

Ultrafast dynamics in the high-symmetry and in the charge density wave phase of 2 H –NbSe 2

Here we investigate carrier and collective mode dynamics in 2H-NbSe 2 using time-resolved optical pump-probe spectroscopy and compare the results with first-principles calculations. Broadband ultrafast reflectivity studies of 2H-NbSe 2 in a wide temperature interval covering the normal, charge density wave (CDW) and superconducting phase were performed. Spectral features observed in the transient reflectivity experiment were associated with specific optical transitions obtained from band structure calculations. Displacive excitation of coherent phonons showed CDW-associated coherent oscillations of the soft phonon mode across the whole spectral range. Temperature evolution of this coherent phonon mode in the low-excitation linear regime shows softening of the mode down to the CDW transition temperature T CDW with subsequent hardening below T CDW . The global fit of the broadband probe data reveals four different relaxation times associated with characteristic electron-electron, electron-phonon, and phonon-phonon relaxation processes. From first-principles calculations of electron-phonon coupling we associate the few picosecond electron-phonon relaxation time $\tau_2$ with a specific group of phonons with frequencies around 20 meV. On the other hand, the anomalously long relaxation time of $\tau_3 \sim$ to 100 ps is associated with anharmonicity-driven phonon-phonon scattering. All relaxation processes result from anomalies near the second order CDW phase transition that are reflected in the temperature dependencies of the characteristic relaxation times and amplitudes of coherent oscillations. At the highest fluences we observe electronic melting of the CDW and disappearance of the mode hardening below T CDW .

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

DOE OSTI2020

Identifying the nature of the QCD transition in relativistic collision of heavy nuclei with deep learning

Using deep convolutional neural network (CNN), the nature of the QCD transition can be identified from the final-state pion spectra from hybrid model simulations of heavy-ion collisions that combines a viscous hydrodynamic model with a hadronic cascade “after-burner”. Two different types of equations of state (EoS) of the medium are used in the hydrodynamic evolution. The resulting spectra in transverse momentum and azimuthal angle are used as the input data to train the neural network to distinguish different EoS. Different scenarios for the input data are studied and compared in a systematic way. A clear hierarchy is observed in the prediction accuracy when using the event-by-event, cascade-coarse-grained and event-fine-averaged spectra as input for the network, which are about 80%, 90% and 99%, respectively. A comparison with the prediction performance by deep neural network (DNN) with only the normalized pion transverse momentum spectra is also made. High-level features of pion spectra captured by a carefully-trained neural network were found to be able to distinguish the nature of the QCD transition even in a simulation scenario which is close to the experiments.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2020

X-Ray Afterglows from the Gamma-Ray Burst “Large-angle” Emission

In this work, we derive basic analytical results for the timing and decay of the gamma-ray burst (GRB) counterpart and delayed afterglow light curves for a brief emission episode from a relativistic surface endowed with angular structure, consisting of a uniform core of size ${\theta }_{c}$ (Lorentz factor ${{\rm{\Gamma }}}_{c}$ and surface emissivity ${i}_{\nu ^{\prime} }^{{\prime} }$ are angle independent) and an axially symmetric power-law envelope (${\rm{\Gamma }}\sim {\theta }^{-g}$). In this "large-angle emission" model, radiation produced during the prompt emission phase (GRB) at angles $\theta \gt {\theta }_{c}$ arrives at the observer well after the burst (delayed emission). The dynamical time range of the very fast decaying GRB "tail" and of the flat afterglow "plateau" and the morphology of the GRB counterpart/afterglow are all determined by two parameters: the core's parameter ${{\rm{\Gamma }}}_{c}{\theta }_{c}$ and the envelope's Lorentz factor index g, leading to three types of light curves that display three post-GRB phases (type 1: tail, plateau/slow decay, post-plateau/normal decay), two post-GRB phases (type 2: tail and fast decay), or just one (type 3: normal decay). We show how X-ray light-curve features can be used to determine core and envelope dynamical and spectral parameters. Testing of the large-angle emission model is done using the Swift/XRT X-ray emission of two afterglows of type 1 (GRB 060607A, GRB 061121), one of type 2 (GRB 061110A), and one of type 3 (GRB 061007). We find that the X-ray afterglows with plateaus require an envelope Lorentz factor ${\rm{\Gamma }}\sim {\theta }^{-2}$ and a comoving-frame emissivity ${i}_{\nu ^{\prime} }^{{\prime} }\sim {\theta }^{2}$; thus, for a typical afterglow spectrum ${F}_{\nu }\sim {\nu }^{-1}$, the lab-frame energy release is uniform over the emitting surface.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2023

Accelerating Nuclear Fuels and Materials Qualification by Multi-Level Irradiation Experiment Campaign

The advanced reactor technologies feature fuels, coolants, and materials that promise safer operating conditions under normal and accident scenarios. However, the nuclear fuels and materials qualifications require several decades (for example, new reactor fuel qualification from conceptualization requires about 20 years). Therefore, accelerating nuclear fuels and materials qualification is essential, and it can be achieved by combining high through-put materials irradiation and testing, advanced post-irradiation examinations, and Multiphysics modeling. This paper addressed the associated challenges in accelerating nuclear fuels and material qualification for new and advanced reactor designs, which differ based on fuel, coolant, operating conditions, and structural materials. These challenges vary for radiation level, operating conditions (e.g., temperature and pressure), and coolant type (e.g., corrosion environment). In addition, the challenges and limitations in modeling tools, experimental facilities, and licensing guidelines are also discussed, and a general solution path forward is recommended.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2023

Presentation: Accelerating Nuclear Fuels and Materials Qualification by Multi-Level Irradiation Experiment Campaign

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2022

PRODeepSyn: predicting anticancer synergistic drug combinations by embedding cell lines with protein–protein interaction network

Abstract Although drug combinations in cancer treatment appear to be a promising therapeutic strategy with respect to monotherapy, it is arduous to discover new synergistic drug combinations due to the combinatorial explosion. Deep learning technology holds immense promise for better prediction of in vitro synergistic drug combinations for certain cell lines. In methods applying such technology, omics data are widely adopted to construct cell line features. However, biological network data are rarely considered yet, which is worthy of in-depth study. In this study, we propose a novel deep learning method, termed PRODeepSyn, for predicting anticancer synergistic drug combinations. By leveraging the Graph Convolutional Network, PRODeepSyn integrates the protein–protein interaction (PPI) network with omics data to construct low-dimensional dense embeddings for cell lines. PRODeepSyn then builds a deep neural network with the Batch Normalization mechanism to predict synergy scores using the cell line embeddings and drug features. PRODeepSyn achieves the lowest root mean square error of 15.08 and the highest Pearson correlation coefficient of 0.75, outperforming two deep learning methods and four machine learning methods. On the classification task, PRODeepSyn achieves an area under the receiver operator characteristics curve of 0.90, an area under the precision–recall curve of 0.63 and a Cohen’s Kappa of 0.53. In the ablation study, we find that using the multi-omics data and the integrated PPI network’s information both can improve the prediction results. Additionally, the case study demonstrates the consistency between PRODeepSyn and previous studies.

Wang, Xiaowen↗

DOE OSTI2022

Lower‐Hybrid Wave Structures and Interactions With Electrons Observed in Magnetotail Reconnection Diffusion Regions

Abstract We investigate waves close to the lower‐hybrid frequency in 12 magnetotail reconnection electron diffusion region (EDR) events with guide field levels of near‐zero to 30%. In about half of the events, the wave vector has a small component along the current sheet normal, consistent with known lower‐hybrid drift wave properties, but the perpendicular magnetic field fluctuations can be comparable or greater than the parallel component, a feature unique to the waves inside and adjacent to EDRs. Another new wave property is that the wave vector has a significant component along the current sheet normal in some events and completely along the normal for one event. In 1/4 of the events, the term has a significant contribution to the wave electric field, possibly a feature of lower‐hybrid waves more likely to exist in the diffusion region than further away from the X‐line. Electron temperature variations are correlated with the wave potential, due to wave electric field acceleration and crossings at the corrugated separatrix region with different amounts of mixing between reconnection inflowing and outflowing populations. The latter also leads to the anti‐correlation between parallel and perpendicular temperature components. Using four‐spacecraft measurements, the magnetic field line twisting is demonstrated by the correlated fluctuations in and . The lower‐hybrid wave in the EDR of weak guide field reconnection may be generated near separatrices and penetrate to the mid‐plane or locally generated, and the latter possibility is beyond the prediction of previous reconnection simulations.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2021

Light Element Abundances and Multiple Populations in M53

We present results from a study of 94 red giant stars in the globular cluster M53. We use low-resolution spectra to measure the strength of CN and CH features at ∼3800 and 4300 Å, respectively. The strengths of these features are used to classify stars into a CN-enhanced and CN-normal population and to measure C and N abundances in all 94 stars. We find the red giant branch stars to be evenly split between the two populations identified, and observe the presence of CN-enhanced stars on the asymptotic giant branch. In addition, we identify five CH star candidates based on the strength of their CN and CH band features, and the presence of a P-branch in their CH band. We compare our identification of multiple populations to those based on the Na–O anticorrelation and pseudo-color indices in Hubble Space Telescope UV photometry, and find general agreement between all three methods. Our large sample size also allows us to study the radial distribution of each population, and we find that the CN-enhanced population is more centrally concentrated. We use our C and N measurements to compare the evolutionary changes in these elements as a function of magnitude between the two populations, and show that both populations experience similar evolutionary changes to the surface abundances of C and N. Finally, we calculate C+N+O abundances for each population and compare them to similar measurements made in M10; we find that in both clusters, CN-enhanced stars have a slightly enhanced C+N+O (Δ(C+N+O) ∼ 0.2 dex).

47 OTHER INSTRUMENTATION↗

DOE OSTI2022

Predicting defects in laser powder bed fusion using in-situ thermal imaging data and machine learning

Variation in the local thermal history during the Laser Powder Bed Fusion (LPBF) process in Additive Manufacturing (AM) can cause micropore defects, which add to the uncertainty of the mechanical properties (e.g., fatigue life, tensile strength) of the built materials. In-situ sensing has been proposed for monitoring the AM process to minimize defects, but successful minimization requires establishing a quantitative relationship between the sensing data and the porosity, which is particularly challenging with a large number of variables (e.g., laser speed, power, scan path, powder property). Physics-based modeling can simulate such an in-situ sensing-porosity relationship, but it is computationally costly. In this work, we develop Machine Learning (ML) models that can use in-situ thermographic data to predict the micropore of LPBF stainless steel materials. This work considers two identified key features from the thermal histories: the time above the apparent melting threshold ($\tau$) and the maximum radiance ( T max ). These features are computed, stored for each voxel in the built material, and then used as inputs. The binary state of each voxel, either defective or normal, is the output. Different ML models are trained and tested for the binary classification task. In addition to using the thermal features of each voxel to predict its own state, the thermal features of neighboring voxels are also included as inputs. This is shown to improve the prediction accuracy, which is consistent with thermal transport physics around each voxel contributing to its final state. Among the models trained, the F1 scores on test sets reach above 0.96 for Random Forests. Feature importance analysis based on the ML models shows that T max is more important to the voxel state than $\tau$ . The analysis also finds that the thermal history of the voxels above the present voxel is more influential than those beneath it. Our study significantly extends the capability of using in-situ thermographic data to predict porosity in LPBF materials. Finally, since ML models are fast, they may play integral roles in the optimization and control of such AM technologies.

36 MATERIALS SCIENCE↗

DOE OSTI2023

Discriminative analysis of schizophrenia patients using graph convolutional networks: A combined multimodal MRI and connectomics analysis

Introduction Recent studies in human brain connectomics with multimodal magnetic resonance imaging (MRI) data have widely reported abnormalities in brain structure, function and connectivity associated with schizophrenia (SZ). However, most previous discriminative studies of SZ patients were based on MRI features of brain regions, ignoring the complex relationships within brain networks. Methods We applied a graph convolutional network (GCN) to discriminating SZ patients using the features of brain region and connectivity derived from a combined multimodal MRI and connectomics analysis. Structural magnetic resonance imaging (sMRI) and resting-state functional magnetic resonance imaging (rs-fMRI) data were acquired from 140 SZ patients and 205 normal controls. Eighteen types of brain graphs were constructed for each subject using 3 types of node features, 3 types of edge features, and 2 brain atlases. We investigated the performance of 18 brain graphs and used the TopK pooling layers to highlight salient brain regions (nodes in the graph). Results The GCN model, which used functional connectivity as edge features and multimodal features (sMRI + fMRI) of brain regions as node features, obtained the highest average accuracy of 95.8%, and outperformed other existing classification studies in SZ patients. In the explainability analysis, we reported that the top 10 salient brain regions, predominantly distributed in the prefrontal and occipital cortices, were mainly involved in the systems of emotion and visual processing. Discussion Our findings demonstrated that GCN with a combined multimodal MRI and connectomics analysis can effectively improve the classification of SZ at an individual level, indicating a promising direction for the diagnosis of SZ patients. The code is available at https://github.com/CXY-scut/GCN-SZ.git .

Chen, Xiaoyi↗