Nested Cross Validation Research Articles

Voting-based 1D CNN model for human lower limb activity recognition using sEMG signal.

Surface electromyography (sEMG) signal classification has many applications such as human-machine interaction, diagnosis of kinesiological studies, and neuromuscular diseases. However, these signals are complicated because of different artifacts added to the sEMG signal during recording. In this study, a multi-stage classification technique is proposed for the identification of distinct movements of the lower limbs using sEMG signals acquired from leg muscles of healthy knee and abnormal knee subjects. This investigation involves 11 subjects with a knee abnormality and 11 without knee abnormality for three distinct activities viz. walking, leg extension from sitting position (sitting), and flexion of the leg (standing). Discrete wavelet denoising to fourth level decomposition has been implemented for the artifact reduction and the signal has been segmented using overlapping windowing technique. A study of four different architectures of 1D convolutional neural network models is undertaken for the prediction of lower limb activities and the final prediction is achieved via a voting mechanism of all four model results. The performance parameters of CNN models have been calculated for three different cases: (1) healthy subjects (2) subjects with knee abnormality (3) Pooled data (combination of abnormal knee and healthy knee subjects) using nested threefold cross-validation. It has been found that the voting mechanism yields an average classification accuracy as 99.35%, 97.63%, and 97.14% for healthy subjects, knee abnormal subjects, and pooled data, respectively. The result validates that the proposed voting-based 1D CNN model is efficient and useful in lower limb activity recognition using the sEMG signal.

The Safety INdEx of Prehospital On Scene Triage (SINEPOST) study: the development and validation of a risk prediction model to support ambulance clinical transport decisions on-scene\u2014a protocol

BackgroundDemand for both the ambulance service and the emergency department (ED) is rising every year and when this demand is excessive in both systems, ambulance crews queue at the ED waiting to hand patients over. Some transported ambulance patients are ‘low-acuity’ and do not require the treatment of the ED. However, paramedics can find it challenging to identify these patients accurately. Decision support tools have been developed using expert opinion to help identify these low acuity patients but have failed to show a benefit beyond regular decision-making. Predictive algorithms may be able to build accurate models, which can be used in the field to support the decision not to take a low-acuity patient to an ED.Methods and analysisAll patients in Yorkshire who were transported to the ED by ambulance between July 2019 and February 2020 will be included. Ambulance electronic patient care record (ePCR) clinical data will be used as candidate predictors for the model. These will then be linked to the corresponding ED record, which holds the outcome of a ‘non-urgent attendance’. The estimated sample size is 52,958, with 4767 events and an EPP of 7.48. An XGBoost algorithm will be used for model development. Initially, a model will be derived using all the data and the apparent performance will be assessed. Then internal-external validation will use non-random nested cross-validation (CV) with test sets held out for each ED (spatial validation). After all models are created, a random-effects meta-analysis will be undertaken. This will pool performance measures such as goodness of fit, discrimination and calibration. It will also generate a prediction interval and measure heterogeneity between clusters. The performance of the full model will be updated with the pooled results.DiscussionCreating a risk prediction model in this area will lead to further development of a clinical decision support tool that ensures every ambulance patient can get to the right place of care, first time. If this study is successful, it could help paramedics evaluate the benefit of transporting a patient to the ED before they leave the scene. It could also reduce congestion in the urgent and emergency care system.Trial RegistrationThis study was retrospectively registered with the ISRCTN: 12121281

Feasibility of pocket-sized near-infrared spectrometer for the prediction of cheese quality traits

This study aimed to evaluate the feasibility of a low-cost pocket-sized near-infrared spectrometer (NIR; 740–1070 nm) to predict cheese total nitrogen (%), soluble nitrogen (%), ripening index (%), major minerals (g/100 g of cheese), and fatty acids (g/100 of cheese). A total of 104 ground cheeses samples were scanned 5 times with a pocket-sized NIR and matched with the reference data to develop the prediction models. Modified partial least squares regressions were developed and parameters were optimized through a leave one out cross-validation. Models were validated using 13-fold nested cross validation. The best models were obtained for total fatty acids (ratio of prediction to deviation, RPD = 6.61) and total nitrogen (RPD = 5.95) content. Regarding mineral content, best prediction models were obtained for P (RPD = 4.04), Na (RPD = 3.12), and Ca (RPD = 2.77). Main individual fatty acids and fatty acid groups were poorly predicted, except for SFA (RPD = 2.92), C16:0 (RPD = 2.83), and C4:0 (RPD = 2.51). However, the accuracy of the models was similar to those obtained with benchtop infrared devices. Thus, these results demonstrated that miniaturization of the NIR instruments represents an opportunity in the dairy industry.

Predicting CRT outcome by combining markers from personalized ventricular models of electrophysiology and clinical data

Abstract Introduction Up to 30% of chronic heart failure (CHF) patients undergoing cardiac resynchronization therapy (CRT) do not respond to the treatment. Low CRT response can be associated with incorrect pacing lead positioning in ventricles and not accounting for individual structural changes in myocardium. Therefore, patient stratification for CRT, operation planning and optimization of CRT device settings remain a challenge. Purpose In this study, we develop a predictive model of CRT outcomes based on a combination of clinical data recorded in patients before CRT and on simulations of the response to biventricular pacing in personalized computational models of the ventricles. Methods We used retrospective data from 27 CHF patients who underwent CRT device implantation. CRT responders were defined using clinical data in a year after therapy as showing a decrease in the end-systolic volume of the left ventricle (LV) and an increase in the ejection fraction. For each patient, an anatomical model of the heart and torso was reconstructed from CT images and used to compute ventricular activation and ECGs without pacing and under bi-ventricular pacing with various locations of stimulating leads. For building a predictive model of patient response to CRT, we used clinical data before CRT (12 features in total per patient) together with model-derived biomarkers under biventricular pacing with clinical lead position (16 features in total per a model). Using nested stratified cross-validation, logistic regression models were fitted to various combinations of datasets. Results The best classifier using a complete set of clinical and simulation data showed an average accuracy = 0.78, ROC AUC = 0.78, sensitivity = 0.75, specificity = 0.8, that were much higher than in the model built on the clinical data only. When the best model was fitted to all data on biventricular pacing with clinical lead position, all ten clinical non-responders were truly classified by the model (Fig. 1, left). Using simulations with optimal lead location as input data for this model, seven out of the ten patients were predicted as responders (Fig. 1, right). Blue and orange colours indicate clinical responders and non-responders. Red circles and numbers mark some clinical non-responding patients that are predicted as responders using optimal TAT values. Conclusion Our pilot results show that combination of clinical and simulation data significantly increases the accuracy of classification models for CRT outcomes. Our results suggest that model predictions on the optimal ventricular lead location may essentially increase the probability of CRT success. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Russian Science Foundation Figure 1

Automatic cough classification for tuberculosis screening in a real-world environment

Objective. The automatic discrimination between the coughing sounds produced by patients with tuberculosis (TB) and those produced by patients with other lung ailments. Approach. We present experiments based on a dataset of 1358 forced cough recordings obtained in a developing-world clinic from 16 patients with confirmed active pulmonary TB and 35 patients suffering from respiratory conditions suggestive of TB but confirmed to be TB negative. Using nested cross-validation, we have trained and evaluated five machine learning classifiers: logistic regression (LR), support vector machines, k-nearest neighbour, multilayer perceptrons and convolutional neural networks. Main Results. Although classification is possible in all cases, the best performance is achieved using LR. In combination with feature selection by sequential forward selection, our best LR system achieves an area under the ROC curve (AUC) of 0.94 using 23 features selected from a set of 78 high-resolution mel-frequency cepstral coefficients. This system achieves a sensitivity of 93% at a specificity of 95% and thus exceeds the 90% sensitivity at 70% specificity specification considered by the World Health Organisation (WHO) as a minimal requirement for a community-based TB triage test. Significance. The automatic classification of cough audio sounds, when applied to symptomatic patients requiring investigation for TB, can meet the WHO triage specifications for the identification of patients who should undergo expensive molecular downstream testing. This makes it a promising and viable means of low cost, easily deployable frontline screening for TB, which can benefit especially developing countries with a heavy TB burden.

Rheumatic Heart Disease Screening Based on Phonocardiogram.

Rheumatic heart disease (RHD) is one of the most common causes of cardiovascular complications in developing countries. It is a heart valve disease that typically affects children. Impaired heart valves stop functioning properly, resulting in a turbulent blood flow within the heart known as a murmur. This murmur can be detected by cardiac auscultation. However, the specificity and sensitivity of manual auscultation were reported to be low. The other alternative is echocardiography, which is costly and requires a highly qualified physician. Given the disease’s current high prevalence rate (the latest reported rate in the study area (Ethiopia) was 5.65%), there is a pressing need for early detection of the disease through mass screening programs. This paper proposes an automated RHD screening approach using machine learning that can be used by non-medically trained persons outside of a clinical setting. Heart sound data was collected from 124 persons with RHD (PwRHD) and 46 healthy controls (HC) in Ethiopia with an additional 81 HC records from an open-access dataset. Thirty-one distinct features were extracted to correctly represent RHD. A support vector machine (SVM) classifier was evaluated using two nested cross-validation approaches to quantitatively assess the generalization of the system to previously unseen subjects. For regular nested 10-fold cross-validation, an f1-score of 96.0 ± 0.9%, recall 95.8 ± 1.5%, precision 96.2 ± 0.6% and a specificity of 96.0 ± 0.6% were achieved. In the imbalanced nested cross-validation at a prevalence rate of 5%, it achieved an f1-score of 72.2 ± 0.8%, recall 92.3 ± 0.4%, precision 59.2 ± 3.6%, and a specificity of 94.8 ± 0.6%. In screening tasks where the prevalence of the disease is small, recall is more important than precision. The findings are encouraging, and the proposed screening tool can be inexpensive, easy to deploy, and has an excellent detection rate. As a result, it has the potential for mass screening and early detection of RHD in developing countries.

Molecular dynamic characteristic temperatures for predicting metallic glass forming ability

We explore the use of characteristic temperatures derived from molecular dynamics to predict aspects of metallic Glass Forming Ability (GFA). Temperatures derived from cooling curves of self-diffusion, viscosity, and energy were used as features for machine learning models of GFA. Multiple target and model combinations with these features were explored. First, we use the logarithm of critical casting thickness, log10(Dmax), as the target and trained regression models on 21 compositions. Application of 3-fold cross-validation on the 21 log10(Dmax) alloys showed only weak correlation between the model predictions and the target values. Second, the GFA of alloys were quantified by melt-spinning or suction casting amorphization behavior, with alloys that showed crystalline phases after synthesis classified as Poor GFA and those with pure amorphous phases as Good GFA. Binary GFA classification was then modeled using decision tree-based methods (random forest and gradient boosting models) and were assessed with nested-cross validation. The maximum F1 score for the precision–recall with Good Glass Forming Ability as the positive class was 0.82±0.01 for the best model type. We also compared using simple functions of characteristic temperatures as features in place of the temperatures themselves and found no statistically significant difference in predictive abilities. Although the predictive ability of the models developed here are modest, this work demonstrates clearly that one can use molecular dynamics simulations and machine learning to predict metal glass forming ability.

Predicting local failure of brain metastases after stereotactic radiosurgery with radiomics on planning MR images and dose maps.

Stereotactic radiosurgery (SRS) has become an important modality in the treatment of brain metastases. The purpose of this study is to investigate the potential of radiomic features from planning magnetic resonance (MR) images and dose maps to predict local failure after SRS for brain metastases. Twenty-eight patients who received Gamma Knife (GK) radiosurgery for brain metastases were retrospectively reviewed in this IRB-approved study. 179 irradiated tumors included 42 that locally failed within one-year follow-up. Using SRS tumor volumes, radiomic features were calculated on T1-weighted contrast-enhanced MR images acquired for treatment planning and planned dose maps. 125 radiomic features regarding tumor shape, dose distribution, MR intensities and textures were extracted for each tumor. Logistic regression with automatic feature selection was built to predict tumor progression from local control after SRS. Feature selection and model evaluation using receiver operating characteristic (ROC) curves were performed in a nested cross validation (CV) scheme. The associations between selected radiomic features and treatment outcomes were statistically assessed by univariate analysis. The logistic model with feature selection achieved ROC AUC of 0.82±0.09 on 5-fold CV, providing 83% sensitivity and 70% specificity for predicting local failure. A total of 10 radiomic features including 1 shape feature, 6 MR images and 3 dose distribution features were selected. These features were significantly associated with treatment outcomes (p<0.05). The model was validated on independent holdout data with an AUC of 0.78. Radiomic features from planning MR images and dose maps provided prognostic information in SRS for brain metastases. A model built on the radiomic features shows promise for early prediction of tumor local failure after treatment, potentially aiding in personalized care for brain metastases.

Driver fatigue detection based on prefrontal EEG using multi-entropy measures and hybrid model

The development of a real-time monitoring and warning platform for driver fatigue detection is of great importance to avoid traffic accidents and fatalities. Due to the explosive growth of biosignal data, especially electroencephalogram (EEG) signals, generated in the wearable-IoT age, the practical application and recognition stability of these signals should be considered and further studied. The application of EEG signals, especially prefrontal EEG, is a potential key method for effectively determining when a driver is experiencing fatigue, which would promote the development of wearable devices for fatigue warning. In this work, we reported a rapid and effective forehead EEG-based multientropy feature-driven fatigue identification framework using a robust hybrid model. This method investigated and utilized several common entropies to enhance the characteristic quality of EEG signal, and yielded a better expected outcome than those of several classical algorithms due to the hybrid model. The experimental results using prefrontal EEG for assessing driver fatigue showed that the proposed method is convenient and effective, as validated by 32 healthy participants through double-layer nested cross validation. In addition, multi-entropy measures were evaluated for single-channel EEG fatigue detection, and the results showed that wavelet log-energy entropy (WLE) outperformed the other entropy indices with a better recognition rate and higher computational efficiency. We demonstrated that multientropy fusion was able to significantly increase detection quality. This study provided a new strategy of using prefrontal EEG signals to construct a high-efficiency method for detecting driver fatigue, which could possess great potential for real-time single-channel signal analysis.

Identification of voxel-based texture abnormalities as new biomarkers for schizophrenia and major depressive patients using layer-wise relevance propagation on deep learning decisions

Non-segmented MRI brain images are used for the identification of new Magnetic Resonance Imaging (MRI) biomarkers able to differentiate between schizophrenic patients (SCZ), major depressive patients (MD) and healthy controls (HC). Brain texture measures such as entropy and contrast, capturing the neighboring variation of MRI voxel intensities, were computed and fed into deep learning technique for group classification. Layer-wise relevance was applied for the localization of the classification results. Texture feature map of non-segmented brain MRI scans were extracted from 141 SCZ, 103 MD and 238 HC. The gray level co-occurrence matrix (GLCM) was calculated on a voxel-by-voxel basis in a cube of voxels. Deep learning tested if texture feature map could predict diagnostic group membership of three classes under a binary classification (SCZ vs. HC, MD vs. HC, SCZ vs. MD). The method was applied in a repeated nested cross-validation scheme and cross-validated feature selection. The regions with the highest relevance (positive/negative) are presented. The method was applied on non-segmented images reducing the computation complexity and the error associated with segmentation process.

Classification of first-episode psychosis using cortical thickness: A large multicenter MRI study

Machine learning classifications of first-episode psychosis (FEP) using neuroimaging have predominantly analyzed brain volumes. Some studies examined cortical thickness, but most of them have used parcellation approaches with data from single sites, which limits claims of generalizability. To address these limitations, we conducted a large-scale, multi-site analysis of cortical thickness comparing parcellations and vertex-wise approaches. By leveraging the multi-site nature of the study, we further investigated how different demographical and site-dependent variables affected predictions. Finally, we assessed relationships between predictions and clinical variables. 428 subjects (147 females, mean age 27.14) with FEP and 448 (230 females, mean age 27.06) healthy controls were enrolled in 8 centers by the ClassiFEP group. All subjects underwent a structural MRI and were clinically assessed. Cortical thickness parcellation (68 areas) and full cortical maps (20,484 vertices) were extracted. Linear Support Vector Machine was used for classification within a repeated nested cross-validation framework. Vertex-wise thickness maps outperformed parcellation-based methods with a balanced accuracy of 66.2% and an Area Under the Curve of 72%. By stratifying our sample for MRI scanner, we increased generalizability across sites. Temporal brain areas resulted as the most influential in the classification. The predictive decision scores significantly correlated with age at onset, duration of treatment, and positive symptoms. In conclusion, although far from the threshold of clinical relevance, temporal cortical thickness proved to classify between FEP subjects and healthy individuals. The assessment of site-dependent variables permitted an increase in the across-site generalizability, thus attempting to address an important machine learning limitation.

Use of Empirical Mode Decomposition in ERP Analysis to Classify Familial Risk and Diagnostic Outcomes for Autism Spectrum Disorder.

Event-related potentials (ERPs) activated by faces and gaze processing are found in individuals with autism spectrum disorder (ASD) in the early stages of their development and may serve as a putative biomarker to supplement behavioral diagnosis. We present a novel approach to the classification of visual ERPs collected from 6-month-old infants using intrinsic mode functions (IMFs) derived from empirical mode decomposition (EMD). Selected features were used as inputs to two machine learning methods (support vector machines and k-nearest neighbors (k-NN)) using nested cross validation. Different runs were executed for the modelling and classification of the participants in the control and high-risk (HR) groups and the classification of diagnosis outcome within the high-risk group: HR-ASD and HR-noASD. The highest accuracy in the classification of familial risk was 88.44%, achieved using a support vector machine (SVM). A maximum accuracy of 74.00% for classifying infants at risk who go on to develop ASD vs. those who do not was achieved through k-NN. IMF-based extracted features were highly effective in classifying infants by risk status, but less effective by diagnostic outcome. Advanced signal analysis of ERPs integrated with machine learning may be considered a first step toward the development of an early biomarker for ASD.

Functional Network Alterations as Markers for Predicting the Treatment Outcome of Cathodal Transcranial Direct Current Stimulation in Focal Epilepsy.

Background and PurposeTranscranial direct current stimulation (tDCS) is an emerging non-invasive neuromodulation technique for focal epilepsy. Because epilepsy is a disease affecting the brain network, our study was aimed to evaluate and predict the treatment outcome of cathodal tDCS (ctDCS) by analyzing the ctDCS-induced functional network alterations.MethodsEither the active 5-day, −1.0 mA, 20-min ctDCS or sham ctDCS targeting at the most active interictal epileptiform discharge regions was applied to 27 subjects suffering from focal epilepsy. The functional networks before and after ctDCS were compared employing graph theoretical analysis based on the functional magnetic resonance imaging (fMRI) data. A support vector machine (SVM) prediction model was built to predict the treatment outcome of ctDCS using the graph theoretical measures as markers.ResultsOur results revealed that the mean clustering coefficient and the global efficiency decreased significantly, as well as the characteristic path length and the mean shortest path length at the stimulation sites in the fMRI functional networks increased significantly after ctDCS only for the patients with response to the active ctDCS (at least 20% reduction rate of seizure frequency). Our prediction model achieved the mean prediction accuracy of 68.3% (mean sensitivity: 70.0%; mean specificity: 67.5%) after the nested cross validation. The mean area under the receiver operating curve was 0.75, which showed good prediction performance.ConclusionThe study demonstrated that the response to ctDCS was related to the topological alterations in the functional networks of epilepsy patients detected by fMRI. The graph theoretical measures were promising for clinical prediction of ctDCS treatment outcome.

MRI-based clinical-radiomics model predicts tumor response before treatment in locally advanced rectal cancer

Neoadjuvant chemo-radiotherapy (CRT) followed by total mesorectal excision (TME) represents the standard treatment for patients with locally advanced (≥ T3 or N+) rectal cancer (LARC). Approximately 15% of patients with LARC shows a complete response after CRT. The use of pre-treatment MRI as predictive biomarker could help to increase the chance of organ preservation by tailoring the neoadjuvant treatment. We present a novel machine learning model combining pre-treatment MRI-based clinical and radiomic features for the early prediction of treatment response in LARC patients. MRI scans (3.0 T, T2-weighted) of 72 patients with LARC were included. Two readers independently segmented each tumor. Radiomic features were extracted from both the “tumor core” (TC) and the “tumor border” (TB). Partial least square (PLS) regression was used as the multivariate, machine learning, algorithm of choice and leave-one-out nested cross-validation was used to optimize hyperparameters of the PLS. The MRI-Based “clinical-radiomic” machine learning model properly predicted the treatment response (AUC = 0.793, p = 5.6 × 10–5). Importantly, the prediction improved when combining MRI-based clinical features and radiomic features, the latter extracted from both TC and TB. Prospective validation studies in randomized clinical trials are warranted to better define the role of radiomics in the development of rectal cancer precision medicine.

A deep survival interpretable radiomics model of hepatocellular carcinoma patients.

This work aims to identify a new radiomics signature using imaging phenotypes and clinical variables for risk prediction of overall survival (OS) in hepatocellular carcinoma (HCC) patients treated with stereotactic body radiation therapy (SBRT). 167 patients were retrospectively analyzed with repeated nested cross-validation to mitigate overfitting issues. 56 radiomic features were extracted from pre-treatment contrast-enhanced (CE) CT images. 37 clinical factors were obtained from patients' electronic records. Variational autoencoders (VAE) based survival models were designed for radiomics and clinical features and a convolutional neural network (CNN) survival model was used for the CECT. Finally, radiomics, clinical and raw image deep learning network (DNN) models were combined to predict the risk probability for OS. The final models yielded c-indices of 0.579 (95%CI: 0.544-0.621), 0.629 (95%CI: 0.601-0.643), 0.581 (95%CI: 0.553-0.613) and 0.650 (95%CI: 0.635-0.683) for radiomics, clinical, image input and combined models on nested cross validation scheme, respectively. Integrated gradients method was used to interpret the trained models. Our interpretability analysis of the DNN showed that the top ranked features were clinical liver function and liver exclusive of tumor radiomics features, which suggests a prominent role of side effects and toxicities in liver outside the tumor region in determining the survival rate of these patients. In summary, novel deep radiomic analysis provides improved performance for risk assessment of HCC prognosis compared with Cox survival models and may facilitate stratification of HCC patients and personalization of their treatment strategies. Liver function was found to contribute most to the OS for these HCC patients and radiomics can aid in their management.

Inferred retinal sensitivity in recessive Stargardt disease using machine learning

Spatially-resolved retinal function can be measured by psychophysical testing like fundus-controlled perimetry (FCP or ‘microperimetry’). It may serve as a performance outcome measure in emerging interventional clinical trials for macular diseases as requested by regulatory agencies. As FCP constitute laborious examinations, we have evaluated a machine-learning-based approach to predict spatially-resolved retinal function (’inferred sensitivity’) based on microstructural imaging (obtained by spectral domain optical coherence tomography) and patient data in recessive Stargardt disease. Using nested cross-validation, prediction accuracies of (mean absolute error, MAE [95% CI]) 4.74 dB [4.48–4.99] were achieved. After additional inclusion of limited FCP data, the latter reached 3.89 dB [3.67–4.10] comparable to the test–retest MAE estimate of 3.51 dB [3.11–3.91]. Analysis of the permutation importance revealed, that the IS&OS and RPE thickness were the most important features for the prediction of retinal sensitivity. ’Inferred sensitivity’, herein, enables to accurately estimate differential effects of retinal microstructure on spatially-resolved function in Stargardt disease, and might be used as quasi-functional surrogate marker for a refined and time-efficient investigation of possible functionally relevant treatment effects or disease progression.

Optimization strategies of neural networks for impact damage classification of RC panels in a small dataset

Predicting damage modes of reinforced concrete (RC) panels subjected to impact loading is a difficult task and often involves considerable effort in doing experiments or simulations. The development of missiles in terms of strength and destructive power requires an accurate estimation of future damage levels. In addition, data collected from the experiment are often small. Therefore, this study aims to build an artificial neural network (ANN) model to classify the damage modes of RC panel under impact loading and enhance its performance by optimizing the model’s hyperparameters when learning a small dataset (254 observes for four classes in this study). To address this a novel optimization strategy was proposed and two metaheuristic optimization algorithms, i.e., genetic algorithm (GA) and particle swarm optimization (PSO) were presented for automatic selection of hyperparameters to increase the accuracy of the ANN model. The proposed optimization strategy was developed based on the incorporation of a stepwise gridsearch (SG) method into a nested cross-validation (NCV) process to find the optimal parameters for the ANN model, named SG-NCV-ANN. The efficiency of the proposed SG-NCV-ANN model and two hybrid models including ANN-GA, ANN-PSO are evaluated by comparing to each other and other machine-learning-based classification methods including ANN using a randomized cross-validation search (RCV-ANN), oblique random forest (oRF), support vector machine (SVM) and k-nearest neighbors (k-NN). Accuracy, micro f1 score, receiver operating characteristic (ROC) curve, and area under the ROC curve (AUC) were employed to thoroughly assess the obtained results from the model. The experimental results indicated that the ANN-GA model achieves the highest AUC and f1 score compared to other state-of-the-art methods, following by the ANN-PSO model. While the proposed SG-NCV-ANN model obtained the best generalization performance on the present small dataset.

Classification and feature analysis of the Human Connectome Project dataset for differentiating between males and females

We analysed features relevant for differentiation between males and females based on the data available from the Human Connectome Project (HCP) S1200 dataset. We used 354 features containing cognitive and emotional measures as well as measures derived from task functional magnetic resonance imaging (MRI) and structural brain MRI. The paper presents a thorough analysis of this extensive set of features using a machine learning approach with a goal of identifying features that have the ability to differentiate between males and females. We used two state of the art classification algorithms with different properties: support vector machine (SVM) and random forest classifier (RFC). For each classifier the hyperparameters were obtained and classifiers were optimized using nested cross validation and grid search. This resulted in the classification performance of 91% and 89% accuracy using SVM and RFC, respectively. Using SHAP (SHapley Additive exPlanations) method we obtained relevance of features as indicators of sex differences and identified features with high discriminative power for sex classification. The majority of top features were brain morphological measures, and only a small proportion were features related to cognitive performance. Our results demonstrate the importance and advantages of using a machine learning approach when analysing sex differences.

Temporal detection of sharp landslide deformation with ensemble-based LSTM-RNNs and Hurst exponent

The sharp slope deformation which often contains seasonal patterns is the major source of the landslide hazard with respect to the local community, which it is a serious geological environment problem. In this paper, a long short-term memory-based deep learning framework has been proposed to model the deformation behaviors especially the sharp deformation of the landslide. First, the Box–Cox transformation is applied to normalize the dataset that includes time-series deformation, precipitation, and reservoir water level. Then, an elastic net (EN)-based ensemble of long short-term memory recurrent neural networks (LSTM-RNNs) is applied to forecast landslide deformation by using month forward-chaining nested cross-validation. This method is performed on the time-series data as our training strategy. Last, the Hurst exponent is formulated to identify incoming sharp deformation. The computational results demonstrated that this approach can accurately identify future sharp deformation. The Hurst exponent illustrates the abnormal patterns in the prediction errors which indicate sharp deformation. As a result, the proposed framework would assist the on-site risk analysis and decision-making process for geological engineers to prevent the landslide hazard in the future.

Origin Determination of Walnuts (Juglans regia L.) on a Worldwide and Regional Level by Inductively Coupled Plasma Mass Spectrometry and Chemometrics.

To counteract food fraud, this study aimed at the differentiation of walnuts on a global and regional level using an isotopolomics approach. Thus, the multi-elemental profiles of 237 walnut samples from ten countries and three years of harvest were analyzed with inductively coupled plasma mass spectrometry (ICP-MS), and the resulting element profiles were evaluated with chemometrics. Using support vector machine (SVM) for classification, validated by stratified nested cross validation, a prediction accuracy of 73% could be achieved. Leave-one-out cross validation was also applied for comparison and led to less satisfactory results because of the higher variations in sensitivity for distinct classes. Prediction was still possible using only elemental ratios instead of the absolute element concentrations; consequently, a drying step is not mandatory. In addition, the isotopolomics approach provided the classification of walnut samples on a regional level in France, Germany, and Italy, with accuracies of 91%, 77%, and 94%, respectively. The ratio of the model’s accuracy to a random sample distribution was calculated, providing a new parameter with which to evaluate and compare the performance of classification models. The walnut cultivar and harvest year had no observable influence on the origin differentiation. Our results show the high potential of element profiling for the origin authentication of walnuts.