Multivariate Algorithm Research Articles

Establishment and Evaluation of Artificial Intelligence-Based Prediction Models for Chronic Kidney Disease under the Background of Big Data.

Objective To establish a prediction model for the risk evaluation of chronic kidney disease (CKD) to guide the management and prevention of CKD. Methods A total of 1263 patients with CKD and 1948 patients without CKD admitted to the Tongde Hospital of the Zhejiang Province from January 1, 2008, to December 31, 2018, were retrospectively analyzed. Spearman's correlation was used to analyze the relationship between CKD and laboratory parameters. XGBoost, random forest, Naive Bayes, support vector machine, and multivariate logistic regression algorithms were employed to establish prediction models for the risk evaluation of CKD. The accuracy, precision, recall, F1 score, and area under the receiver operating curve (AUC) of each model were compared. The new bidirectional encoder representations from transformers with light gradient boosting machine (MD-BERT-LGBM) model was used to process the unstructured data and transform it into researchable unstructured vectors, and the AUC was compared before and after processing. Results Differences in laboratory parameters between CKD and non-CKD patients were observed. The neutrophil ratio and white blood cell count were significantly associated with the occurrence of CKD. The XGBoost model demonstrated the best prediction effect (accuracy = 0.9088, precision = 0.9175, recall = 0.8244, F1 score = 0.8868, AUC = 0.8244), followed by the random forest model (accuracy = 0.9020, precision = 0.9318, recall = 0.7905, F1 score = 0.581, AUC = 0.9519). Comparatively, the predictions of the Naive Bayes and support vector machine models were inferior to those of the logistic regression model. The AUC of all models was improved to some extent after processing using the new MD-BERT-LGBM model. Conclusion The new MD-BERT-LGBM model with the inclusion of unstructured data has contributed to the higher accuracy, sensitivity, and specificity of the prediction models. Clinical features such as age, gender, urinary white blood cells, urinary red blood cells, thrombin time, serum creatinine, and total cholesterol were associated with CKD incidence.

Using Nix color sensor and Munsell soil color variables to classify contrasting soil types and predict soil organic carbon in Eastern India

Optimal soil management depends on rapid and frequent monitoring of key soil properties, which are conventionally measured in the laboratory using laborious wet-chemistry protocols. The Nix color sensor has recently exhibited promise for predicting several soil properties using soil color. This study evaluated the relationship between the Munsell Soil Color Chart (MSCC) color values of dry and ground surface soil samples to those reported by the Nix color sensor with (NixSTD) and without MSCC standardization (NixNON-STD) to classify 371 samples collected from three contrasting soil types, collected from three agro-climatic zones (coastal saline zone, red and laterite zone, and Gangetic alluvial zone) and to predict soil organic carbon (OC) using different multivariate data mining algorithms. Comparing the CIEL*a*b* color values reported by the MSCC and the NixSTD, an acceptable mean color difference (ΔE*ab) value of 5.20 was obtained, indicating the potential accuracy of the Nix sensor. Principal component analysis efficiently clustered the soil types using the RGB variables extracted from the MSCC color chips in tandem with the NixSTD/NixNON-STD data. Both classification tree and linear support vector machine algorithms perfectly classified all three contrasting soil types using NixNON-STD data alone. Besides, the combination of the MSCC and the NixNON-STD datasets produced the best OC prediction (R2 = 0.66) via random forest (RF) algorithm and indicated the potential of Nix in digital soil morphometrics. In most of the RF models, redness (a*), yellowness (b*), and yellow (Y) variables appeared influential, presumably because of their negative correlation with OC in red and laterite soils. More research is warranted to measure the impacts of variable soil moisture and other confounding soil morphological features on the soil classification and OC prediction performance to extend the approach for classifying soil types and predicting OC in-situ.

Nanotube abundance from non-negative matrix factorization of Raman spectra as an example of chemical purity from open source machine learning

The chemical purity of materials is important for semiconductors, including the carbon nanotube material system, which is emerging in semiconductor applications. One approach to get statistically meaningful abundances and/or concentrations is to measure a large number of small samples. Automated multivariate classification algorithms can be used to draw conclusions from such large data sets. Here, we use spatially-mapped Raman spectra of mixtures of chirality-sorted single walled carbon nanotubes dispersed sparsely on flat silicon/silicon oxide substrates. We use non-negative matrix factorization (NMF) decomposition in scikit-learn, an open-source, python language “machine learning” package, to extract spectral components and derive weighting factors. We extract the abundance of minority species (7,5) nanotubes in mixtures by testing both synthetic data, and real samples prepared by dilution. We show how noise limits the purity level that can be evaluated. We determine real situations where this approach works well, and identify situations where it fails.

Radiomics model based on multi-sequence MR images for predicting preoperative immunoscore in rectal cancer.

To establish and validate a radiomics model based on multi-sequence magnetic resonance (MR) images for preoperative prediction of immunoscore in rectal cancer. This retrospective study included 133 patients with pathologically confirmed rectal cancer after surgical resection who underwent MR examination before treatment within two weeks. All patients were randomly divided into training cohort (n = 92) and validation (n = 41) cohort according to a ratio of 7:3. The volumes of interest were manually delineated in the T2-weighted images (T2WI) and apparent diffusion coefficient (ADC) images, from which a total of 804 radiomics features were extracted. Thereafter, we used Spearman correlation analysis and gradient boosting decision tree (GBDT) algorithm to select the strongest features, and the radiomics scores were established using multivariate logistic regression algorithm, including two single-mode models and two dual-mode models. The predictive performance and the clinical usefulness of the model were assessed by the receiver operating characteristic (ROC) curve, calibration curve and decision curve analysis (DCA). Integrated model A based on T2WI and ADC images showed a better predictive performance, which yielded an AUC of 0.770 (95% CI 0.673-0.867) in the training cohort and 0.768 (95% CI 0.619-0.917) in the validation cohort. Calibration curve showed good agreement between predicted results of the model and actual events, and DCA indicated good clinical usefulness. Moreover, stratification analysis proved that the integrated model A had strong robustness. Integrated model A based on T2WI and ADC images has the potential to be used as a non-invasive tool for preoperative prediction of immunoscore in rectal cancer. It may be useful in evaluating prognosis and guiding individualized immunotherapy of patients.

A comparison of online methods for change point detection in ion-mobility spectrometry data

When on-site classification of volatile organic compounds (VOCs) is required, a portable ion mobility spectrometer (IMS) is a suitable choice. However, the IMS readings often show transient phases before they stabilize. Even so the importance of transient phase and features extracted from it has been highlighted in the literature, it has not, to our knowledge, been used for IMS-based classification so far. This paper analyzes whether change point detection algorithms with low computational complexity can separate transient and stable phases in IMS readings. The algorithms were tested on IMS data from different types of mushrooms. All algorithms successfully detected switches from transient to stable phase. The most accurate results were provided by the previously proposed multivariate max-CUSUM algorithm and the matrix form CUSUM algorithm, which is developed in this paper. • A dataset of IMS measurements from 2 types of mushrooms is presented and shared. • A modification of the CUSUM change point detection algorithm is proposed. • The proposed algorithm is compared to existing change point detection algorithms.

Spectral analysis based robust multi-level intrusion detection in wide area frequency control

Frequency control is one of the critical systems responsible for maintaining grid stability, making it an attractive choice for cyber-attacks. The paper presents a singular spectrum analysis for extracting the system’s dynamics under regular operation and proposes a projection-based distance tracking method to detect Multi-area Load Frequency Control (MA-LFC) attacks. Two variations of the algorithm, namely single-variate and multi-variate algorithms, are proposed for detection at different levels of the power system. The proposed methodology is robust, adaptive, and computationally efficient, especially when system and measurement noises are considered. The method’s main advantage is that it does not require attack data and can be integrated with the existing grid control system. The algorithm can be implemented within the RTUs/IEDs without the need for additional processors. The detection algorithm successfully detects different types of attacks, including stealth attacks on multiple sensors. By selecting proper set of signals, it is possible to obtain 99.85% accuracy and 2 cycles deduction speed. The algorithm is tested on an IEEE 39-bus New England test system in the MATLAB environment, and its performance under noisy measurements is better than existing methods.

The Effect of Different Doses of 1-Methylcyclopropene on Postharvest Physiology and Predicting Ethylene Production through Multivariate Adaptive Regression Splines in Cocktail Tomato

Maintaining the postharvest quality of tomatoes, which are essential in the world vegetable trade, is very important; otherwise, storage may cause severe quality and economic losses. This study aimed to determine the effects of 1-MCP treatments on quality and storage time to prevent quality loss and deterioration in tomatoes due to high temperature and to predict ethylene production through the Multivariate Adaptive Regression Splines algorithm during long-term storage. For this purpose, same-sized fruits were divided into three different groups. Two groups were treated with 1-MCP (625 and 1250 ppb), and the untreated fruit was the control. Then, the tomatoes were stored for 39 days at a temperature of 20 °C and relative humidity of 85–90%. The results obtained from the present study showed that the 1-MCP treatments prolonged the life and maintained the quality of tomato fruit during storage. The ethylene production and respiration rate were significantly lower in 1-MCP-treated fruit than in the untreated fruit during the storage period (p < 0.05). According to the Multivariate Adaptive Regression Splines algorithm, it was observed that increasing doses of 1-MCP inhibited ethylene production. Furthermore, weight loss and respiratory rate were effective on ethylene production at 100 and 43.9%, respectively. In conclusion, treating the fruit with 1250 ppb of 1-MCP was determined to be the best practice for maintaining all quality criteria during storage.

Dissociation of early and late face-related processes in autism spectrum disorder and Williams syndrome

BackgroundWilliams syndrome (WS) and Autism Spectrum Disorders (ASD) are neurodevelopmental conditions associated with atypical but opposite face-to-face interactions patterns: WS patients overly stare at others, ASD individuals escape eye contact. Whether these behaviors result from dissociable visual processes within the occipito-temporal pathways is unknown.Using high-density electroencephalography, multivariate signal processing algorithms and a protocol designed to identify and extract evoked activities sensitive to facial cues, we investigated how WS (N = 14), ASD (N = 14) and neurotypical subjects (N = 14) decode the information content of a face stimulus.ResultsWe found two neural components in neurotypical participants, both strongest when the eye region was projected onto the subject's fovea, simulating a direct eye contact situation, and weakest over more distant regions, reaching a minimum when the focused region was outside the stimulus face. The first component peaks at 170 ms, an early signal known to be implicated in low-level face features. The second is identified later, 260 ms post-stimulus onset and is implicated in decoding salient face social cues.Remarkably, both components were found distinctly impaired and preserved in WS and ASD. In WS, we could weakly decode the 170 ms signal based on our regressor relative to facial features, probably due to their relatively poor ability to process faces’ morphology, while the late 260 ms component was highly significant. The reverse pattern was observed in ASD participants who showed neurotypical like early 170 ms evoked activity but impaired late evoked 260 ms signal.ConclusionsOur study reveals a dissociation between WS and ASD patients and points at different neural origins for their social impairments.

A New Multivariate Linear Regression MPPT Algorithm for Solar PV System with Boost Converter

Operating solar photovoltaic (PV) panels at the maximum power point (MPP) is considered to enrich energy conversion efficiency. Each MPP tracking technique (MPPT) has its conversion efficiency and methodology for tracking the MPP. This paper introduces a new method for operating the PV panel at MPP by implementing the multivariate linear regression (MLR) machine learning algorithm. The MLR machine learning model in this study is trained and tested using the data collected from the PV panel specifications. This MLR algorithm can predict the maximum power available at the panel, and the voltage corresponds to this maximum power for specific values of irradiance and temperature. These predicted values help in the calculation of the duty ratio for the boost converter. The MATLAB/SIMULINK results illustrate that, as time progresses, the PV panel is forced to operate at the MPP predicted by the MLR algorithm, yielding a mean efficiency of more than 96% in the steady-state operation of the PV system, even under variable irradiances and temperatures.

Multi-Scale Process Monitoring Based on Time-Frequency Analysis and Feature Fusion

Data-driven process monitoring is an important tool to ensure safe production and smooth operation. Generally, implicit information can be mined through data processing and analysis algorithms to detect process disturbances on the basis of historical production data. In industrial practice, signals with different sources of disturbance show different distribution patterns along with the time domain and frequency domain, that is, noise and pulse-type changes are usually contained in the high-frequency portion while most process dynamic is contained in the low-frequency portion. However, feature extraction is usually implemented at a single scale in traditional multivariate statistical algorithms. With this concern, a novel multi-scale process monitoring method is proposed in this work, by which wavelet packet decomposition is first employed for time-frequency analysis. After decomposition, multivariate statistical models are established for each scale to construct process statistics. For the high-frequency part, the classical principal component analysis (PCA) algorithm is adopted to construct squared prediction error (SPE) and Hotelling T2(T2) statistics. While for the low-frequency part, the slow feature analysis (SFA) algorithm is adopted to construct T2, Te2, S2 and Se2 statistics for the extraction of the long-term slowly changing trend. Then the monitoring statistics, obtained from each method at different scales, are integrated by a support vector data description (SVDD) method to give a final fault detection decision. The performance of the proposed method is verified on the benchmark Tennessee Eastman Process (TEP) and an industrial continuous catalytic reforming heat exchange unit by comparing with related multivariate statistical methods, which only focus on a single scale.

Abstract 2259: Predictive modeling of smoldering multiple myeloma progression to multiple myeloma by continuous variable analysis

Abstract Introduction: Multiple myeloma (MM) is consistently preceded by two precursor conditions, monoclonal gammopathy of undetermined significance (MGUS) and smoldering multiple myeloma (SMM). The distinctions between MGUS, SMM, and MM rely on clinical values with inherent variation relating to tumor burden quantified from bone marrow biopsies, a measure subject to inconsistencies in location, timing, and pathologist interpretation. These challenges limit the potential of current standardized risk criteria and advocate for models that examine precursor disease kinetics. We thus developed a risk model that leverages dynamic changes in markers of precursor disease to improve clinical ability to predict time to disease progression. Methods: To model the evolution of progression risk, we built PANGEA, an international retrospective cohort of precursor patients with baseline and serial time points of clinical and biological variables. This cohort comprises 1095 SMM patients, 254 (23%) of which progressed to MM. Using this cohort, we modeled progression to MM with Cox regression using time-dependent and continuous clinical variables. The model was trained on a subset of data restricted to patients of the Dana-Farber Cancer Institute (DFCI) and validated its performance by computing the c-statistic in a sub-cohort independent from the DFCI training cohort. Results: The PANGEA cohort was first used to validate current models of SMM disease progression. We validated the 20/2/20 International Myeloma Working Group criteria for SMM patients using binary cutoffs of initial measurements (baseline model), and then extended this model, allowing for re-stratification by the 20/2/20 criteria over time (dynamic model). We then assessed whether rates of change in a set of myeloma-specific clinical variables unrestricted to those of the 20/2/20 criteria improved the predictive ability of the model. This improved our progression prediction as indicated by a c-statistic increase of more than 10% with respect to both 20/2/20 models (baseline and dynamic). Specifically, changes in disease indicators such as age and creatinine are highly predictive of imminent disease progression (p-value < 0.01). Finally, we clustered patients based on latent trajectories of these time-varying clinical variables and included the trajectory classes in the Cox regression. The resulting multivariable, dynamic algorithm is a dramatic improvement over current clinical standards in predicting progression from SMM to MM disease. Conclusion: The PANGEA multivariable algorithm’s use of continuous clinical variables enhances progression risk predictions in SMM. These findings demonstrate that disease progression from SMM to MM, which likely occurs by the acquisition of sequential changes to the plasma cell clone, can be tracked by trends in clinical values, thus improving prognostication for precursor patients. Citation Format: Annie Cowan, Habib El-Khoury, Federico Ferrari, Samuel S. Freeman, Robert Redd, Jacqueline Perry, Vidhi Patel, Priya Kaur, Hadley Barr, Katelyn Downey, David Argyelan, Anna V. Justis, David J. Lee, Elizabeth D. Lightbody, Foteini Theodorakakou, Despina Fotiou, Nikolaos Kanellias, Christine Liacos, Gad Getz, Lorenzo Trippa, Catherine Marinac, Efstathios Kastritis, Dimopoulos Meletios, Irene Ghobrial. Predictive modeling of smoldering multiple myeloma progression to multiple myeloma by continuous variable analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2259.

Proteins associated with future suicide attempts in bipolar disorder: A large-scale biomarker discovery study

Suicide is a major cause of death worldwide. Several biological systems have been implicated in suicidal behavior but studies of candidate biomarkers have failed to produce clinically relevant biomarkers for suicide prediction. The objective of the present study was to identify novel candidate biomarkers for suicidal behavior. We used a nested case-control study design where a large cohort of patients with bipolar disorder (N = 5 110) were followed up to 8 years after blood sampling. We included patients that attempted suicide during follow-up (N = 348) and matched bipolar disorder patients from the same cohort who did not attempt suicide during the study period (N = 348) and analyzed a total of 92 proteins with a neuro exploratory multiplex panel. Using a multivariate classification algorithm devised to minimize bias in variable selection, we identified a parsimonious set of proteins that best discriminated bipolar disorder patients with and without prospective suicide attempts. The algorithm selected 16 proteins for the minimal-optimal classification model, which outperformed 500 models with permuted outcome (p = 0.0004) but had low sensitivity (53%) and specificity (64%). The candidate proteins were then entered in separate logistic regression models to calculate protein-specific associations with prospective suicide attempts. In individual analyses, three of these proteins were significantly associated with prospective suicide attempt (SCGB1A1, ANXA10, and CETN2). Most of the candidate proteins are novel to suicide research.

Mismatch Negativity in Response to Auditory Deviance and Risk for Future Psychosis in Youth at Clinical High Risk for Psychosis

Although clinical criteria for identifying youth at risk for psychosis have been validated, they are not sufficiently accurate for predicting outcomes to inform major treatment decisions. The identification of biomarkers may improve outcome prediction among individuals at clinical high risk for psychosis (CHR-P). To examine whether mismatch negativity (MMN) event-related potential amplitude, which is deficient in schizophrenia, is reduced in young people with the CHR-P syndrome and associated with outcomes, accounting for effects of antipsychotic medication use. MMN data were collected as part of the multisite case-control North American Prodrome Longitudinal Study (NAPLS-2) from 8 university-based outpatient research programs. Baseline MMN data were collected from June 2009 through April 2013. Clinical outcomes were assessed throughout 24 months. Participants were individuals with the CHR-P syndrome and healthy controls with MMN data. Participants with the CHR-P syndrome who developed psychosis (ie, converters) were compared with those who did not develop psychosis (ie, nonconverters) who were followed up for 24 months. Analysis took place between December 2019 and December 2021. Electroencephalography was recorded during a passive auditory oddball paradigm. MMN elicited by duration-, pitch-, and duration + pitch double-deviant tones was measured. The CHR-P group (n = 580; mean [SD] age, 19.24 [4.39] years) included 247 female individuals (42.6%) and the healthy control group (n = 241; mean age, 20.33 [4.74] years) included 114 female individuals (47.3%). In the CHR-P group, 450 (77.6%) were not taking antipsychotic medication at baseline. Baseline MMN amplitudes, irrespective of deviant type, were deficient in future CHR-P converters to psychosis (n = 77, unmedicated n = 54) compared with nonconverters (n = 238, unmedicated n = 190) in both the full sample (d = 0.27) and the unmedicated subsample (d = 0.33). In the full sample, baseline medication status interacted with group and deviant type indicating that double-deviant MMN, compared with single deviants, was reduced in unmedicated converters compared with nonconverters (d = 0.43). Further, within the unmedicated subsample, deficits in double-deviant MMN were most strongly associated with earlier conversion to psychosis (hazard ratio, 1.40 [95% CI, 1.03-1.90]; P = .03], which persisted over and above positive symptom severity. This study found that MMN amplitude deficits were sensitive to future psychosis conversion among individuals at risk of CHR-P, particularly those not taking antipsychotic medication at baseline, although associations were modest. While MMN shows limited promise as a biomarker of psychosis onset on its own, it may contribute novel risk information to multivariate prediction algorithms and serve as a translational neurophysiological target for novel treatment development in a subgroup of at-risk individuals.

Comparing Calibration Algorithms for the Rapid Characterization of Pretreated Corn Stover Using Near-Infrared Spectroscopy

Rapid characterization of biomass composition is a key enabling technology for biorefineries—the ability to measure the chemical composition of biomass materials entering the biorefinery as well as the composition of key process intermediate streams would allow real-time process control and the development of robust models to predict process performance. The utility of near-infrared (NIR) spectroscopy for rapid characterization requires multivariate algorithms for building calibration models. The most prevalent algorithm used for building calibration models using NIR spectra is the linear modeling algorithm Partial Least Squares Regression (PLS). Nonlinear regression algorithms (which are typically more computationally intensive than linear modeling approaches) have gained popularity in recent years due to their ability to solve a wide variety of classification and regression problems and the dramatic increase in available computational resources. In this work, we demonstrate that a calibration model can predict the composition of corn stover process intermediate samples pretreated with three different treatments—hot water (HW), dilute acid (DA), and deacetylation followed by dilute acid (DDA). We quantitatively compare three different algorithms for building prediction models based on near-infrared spectroscopy—partial least squares (PLS), support vector machines (SVM), and random forests (RF). We demonstrate the utility of improving model performance by accounting for instrument performance variability using repeated measurements of standard materials (e.g., the “repeatability file” strategy) and investigate its performance with nonlinear regression techniques, and we discuss methods for quantifying the uncertainties of specific predictions among the three methods.

A further study comparing forward search multivariate outlier methods including ATLA with an application to clustering

This paper makes comparisons of automated procedures for robust multivariate outlier detection through discussion and simulation. In particular, automated procedures that use the forward search along with Mahalanobis distances to identify and classify multivariate outliers subject to predefined criteria are examined. Procedures utilizing a parametric model criterion based on a chi ^2-distribution are among these, whereas the multivariate Adaptive Trimmed Likelihood Algorithm (ATLA) identifies outliers based on an objective function that is derived from the asymptotics of the location estimator assuming a multivariate normal distribution. Several criterion including size (false positive rate), sensitivity, and relative efficiency are canvassed. To illustrate relative efficiency in a multivariate setting in a new way, measures of variability of the multivariate location parameter when the underlying distribution is chosen from a multivariate generalization of the Tukey–Huber epsilon -contamination model are used. Mean slippage models are also entertained. The simulation results here are illuminating and demonstrate there is no broadly accepted procedure that outperforms in all situations, albeit one may ascertain circumstances for which a particular method may be best if implemented. Finally the paper explores graphical monitoring for existence of clusters and the potential of classification through occurrence of multiple minima in the objective function using ATLA.

Synbiotic-Derived Metabolites Reduce Neuroinflammatory Symptoms of Alzheimer’s Disease

ObjectivesSynbiotics, the combination of probiotics and prebiotics, can be designed to produce specific bioavailable metabolites that penetrate the blood-brain-barrier and reduce neuropathologies associated with Alzheimer's Disease (AD). We designed an in vitro model of the human gastrointestinal tract that can be used to optimize the production of metabolites from synbiotic formulations using machine learning algorithms that can then be tested in animal models of AD. Our objective is to optimize/characterize a synbiotic's metabolite production using an in vitro bioreactor as a therapeutic tool against AD. We hypothesize that a synbiotic designed towards the production of specific brain-bioactive metabolites will synergistically mitigate the neuro- and systemic-pathologies associated with AD.MethodsUsing the in vitro model of the human gastrointestinal tract, we have used this optimized combination of polyphenols and tested 15 combinations of 6 probiotic bacteria as a synbiotic to optimize metabolite production using a multivariate regression algorithm (MARS) and found that three probiotic bacteria, Lactobacillus plantarum, Bifidobacterium infantis and L. salivarius together synergistically improve the production of brain bioavailable metabolites including quercetin, kamferin, 4-hydroxyphenylpyruvic acid and 4-hydroxyphenylacetic acid.ResultsWe have found that two polyphenolic metabolites 3-hydroxybenzoic acid and 3-(3′-hydroxyphenyl) propionic acid derived from a grapeseed polyphenolic extract (GSPE) can penetrate the blood-brain-barrier and in vitro, inhibit the aggregation of amyloid plaques and tau fibrils. By increasing the diversity of the polyphenolic pool by adding a concord grape extract and resveratrol to the GSPE, we provided greater protection against cognitive impairment and amyloid aggregation in an AD mouse model than the components alone.Conclusions In vitro studies have confirmed that these metabolites have potent anti-inflammatory activity. This synbiotic combining potent grape-derived polyphenolic precursors with bioactive probiotic bacteria has the potential to slow the progression and treat AD by synergistically targeting multiple of its neuropathologies including inflammation, amyloid aggregation and tau fibril formation.Funding SourcesThese studies were funded by the NCCIH.

A multivariate adaptive gradient algorithm with reduced tuning efforts

Large neural networks usually perform well for executing machine learning tasks. However, models that achieve state-of-the-art performance involve arbitrarily large number of parameters and therefore their training is very expensive. It is thus desired to implement methods with small per-iteration costs, fast convergence rates, and reduced tuning. This paper proposes a multivariate adaptive gradient descent method that meets the above attributes. The proposed method updates every element of the model parameters separately in a computationally efficient manner using an adaptive vector-form learning rate, resulting in low per-iteration cost. The adaptive learning rate computes the absolute difference of current and previous model parameters over the difference in subgradients of current and previous state estimates. In the deterministic setting, we show that the cost function value converges at a linear rate for smooth and strongly convex cost functions. Whereas in both the deterministic and stochastic setting, we show that the gradient converges in expectation at the order of O(1/k) for a non-convex cost function with Lipschitz continuous gradient. In addition, we show that after T iterates, the cost function of the last iterate scales as O(log(T)/T) for non-smooth strongly convex cost functions. Effectiveness of the proposed method is validated on convex functions, smooth non-convex function, non-smooth convex function, and four image classification data sets, whilst showing that its execution requires hardly any tuning unlike existing popular optimizers that entail relatively large tuning efforts. Our empirical results show that our proposed algorithm provides the best overall performance when comparing it to tuned state-of-the-art optimizers.

S-ICD in heart failure patients: real-world data from a multicenter, european analysis

Abstract Funding Acknowledgements Type of funding sources: None. Background Data on patients with heart failure (HF) and subcutaneous implantable cardioverter defibrillator (S-ICD) are very scarce and limited to a single prospective analysis from the UNTOUCHED trial. Purpose Aim of this study was to assess clinical outcomes of the S-ICD in HF patients, comparing them with a no-HF population, in a real-world analysis from the largest European retrospective S-ICD registry (ELISIR registry). Methods All consecutive patients undergoing S-ICD implantation at 20 European institutions enrolled in the ELISIR registry were used for the current analysis. According to European Guidelines, the registry population was classified into two groups: the HF cohort (further classified as HF with reduced and mid-range ejection fraction – HFrEF and HFmrEF) vs the no-HF group. The primary outcome of the study was the inappropriate shock (IS) rate across the two cohorts. As secondary outcomes, appropriate shocks, cardiovascular mortality and device-related complications during follow-up were assessed. Results A total of 1409 patients from the ELISIR registry were included in this analysis; HF patients represented 57.3% of the entire cohort (n=701, 86.9% HFrEF; n=106,13.1% HFmrEF). As expected, the HF cohort showed significantly higher rates of cardiovascular risk factors and comorbidities when compared to the no-HF cohort. Over a median follow-up of approximately 2 years, a total of 133 inappropriate shocks were observed in the entire cohort, without significant differences among the two groups (9.2% vs 9.8%, p=0.689). 133 complex ventricular arrhythmias were adequately recognized and treated in the overall cohort, showing similar rates of appropriate shocks (9.2% vs 9.8%, p=0.689). Inappropriate and effective shocks-free survival has been represented in Figure 1, showing Kaplan-Meier estimates comparing HF vs no-HF patients, also stratified by left ventricular ejection fraction (LVEF). The impact of baseline and procedural characteristics on the primary outcome was tested through univariable and multivariable Cox regression analysis in HF patients; at multivariate analysis, only age (HR=0.974 [0.955–0.992], p=0.005), LVEF (HR=0.954 [0.926-0.984], p=0.003), ARVC (HR=3.364 [1.206-9.384], p=0.020) and smart pass algorithm "on" (HR=0.321 [0.184-0.560], p<0.001) remained associated with inappropriate shocks (Figure 2). A low number of patients (n=76) experienced device-related complications, more frequently in the HF cohort (6.2% vs 3.8%, p=0.031) with no significant differences regarding any specific outcome of interest: lead infection (1.1% vs 0.7%, p=0.381), pocket infection (1.9% vs 0.8%, p=0.107), pocket hematoma (3.2% vs 2.8%, p=0.668). Conclusion The rate of inappropriate shocks seems to be comparable in both HF and non-HF patients implanted with S-ICD. However, the rate of device-related complications was slightly more frequent in HF patients.

Computed Tomography-Based Radiomics for Differentiation of Thymic Epithelial Tumors and Lymphomas in Anterior Mediastinum.

ObjectiveTo investigate the differential diagnostic performance of computed tomography (CT)-based radiomics in thymic epithelial tumors (TETs) and lymphomas in anterior mediastinum.MethodsThere were 149 patients with TETs and 93 patients with lymphomas enrolled. These patients were assigned to a training set (n = 171) and an external validation set (n = 71). Dedicated radiomics prototype software was used to segment lesions on preoperative chest enhanced CT images and extract features. The multivariable logistic regression algorithm was used to construct three models according to clinico-radiologic features, radiomics features, and combined features, respectively. Performance of the three models was compared by using the area under the receiver operating characteristic curves (AUCs). Decision curve analysis was used to evaluate clinical utility of the three models.ResultsFor clinico-radiologic model, radiomics signature model, and combined model, the AUCs were 0.860, 0.965, 0.975 and 0.843, 0.961, 0.955 in the training cohort and the test cohort, respectively (all P<0.05). The accuracies of each model were 0.836, 0.895, 0.918 and 0.845, 0.901, 0.859 in the two cohorts, respectively (all P<0.05). Compared with the clinico-radiologic model, better diagnostic performances were found in the radiomics signature model and the combined model.ConclusionsRadiomics signature model and combined model exhibit outstanding and comparable differential diagnostic performances between TETs and lymphomas. The CT-based radiomics analysis might serve as an effective tool for accurately differentiating TETs from lymphomas before treatment.

Three-dimensional instantaneous orbit map for rotor-bearing system based on a novel multivariate complex variational mode decomposition algorithm

Full spectrum and holospectrum are homogeneous information fusion technology developed for the fault diagnosis of rotating machinery, which is extensively exploited in the analysis of the orbits of rotor-bearing systems. However, they are not adapted for non-stationary signals, nor can they be used for fusion analysis of vibrations of multiple bearing sections. By drawing inspiration from the multivariate variational mode decomposition (MVMD) and the complex-valued signal decomposition, we propose a method called multivariate complex variational mode decomposition (MCVMD). It can simultaneously extract the forward and backward components of multiple bearing sections and realize non-stationary complex signal decomposition of multiple bearing sections of the rotor. To achieve the visualization goal of condition monitoring, we propose the three-dimensional instantaneous orbit map (3D-IOM). It enables more features of shaft vibration of a rotor system to be displayed and offers a new way for the fusion analysis of vibration signals of multiple bearing sections of rotating machinery. Furthermore, making the most of the joint information, we also provide a high-resolution time-full spectrum (Time-FS) to display the forward and backward frequency components of multiple bearing sections. The effectiveness of the proposed method through both the simulated experiment and the real-life complex-valued signals is demonstrated in this paper.