Recursive Feature Elimination Approach Research Articles

SRFE: A stepwise recursive feature elimination approach for network intrusion detection systems

SRFE: A stepwise recursive feature elimination approach for network intrusion detection systems

RDR100: A Robust Computational Method for Identification of Krüppel-like Factors

Background: Krüppel-like factors (KLFs) are a family of transcription factors containing zinc fingers that regulate various cellular processes. KLF proteins are associated with human diseases, such as cancer, cardiovascular diseases, and metabolic disorders. The KLF family consists of 18 members with diverse expression profiles across numerous tissues. Accurate identification and annotation of KLF proteins is crucial, given their involvement in important biological functions. Although experimental approaches can identify KLF proteins precisely, large-scale identification is complicated, slow, and expensive. Methods: In this study, we developed RDR100, a novel random forest (RF)-based framework for predicting KLF proteins based on their primary sequences. First, we identified the optimal encodings for ten different features using a recursive feature elimination approach, and then trained their respective model using five distinct machine learning (ML) classifiers. Results: The performance of all models was assessed using independent datasets, and RDR100 was selected as the final model based on its consistent performance in cross-validation and independent evaluation. Conclusion: Our results demonstrate that RDR100 is a robust predictor of KLF proteins. RDR100 web server is available at https://procarb.org/RDR100/.

Development and evaluation of machine learning models for predicting large-for-gestational-age newborns in women exposed to radiation prior to pregnancy

IntroductionThe correlation between radiation exposure before pregnancy and abnormal birth weight has been previously proven. However, for large-for-gestational-age (LGA) babies in women exposed to radiation before becoming pregnant, there is no prediction model yet.Material and methodsThe data were collected from the National Free Preconception Health Examination Project in China. A sum of 455 neonates (42 SGA births and 423 non-LGA births) were included. A training set (n = 319) and a test set (n = 136) were created from the dataset at random. To develop prediction models for LGA neonates, conventional logistic regression (LR) method and six machine learning methods were used in this study. Recursive feature elimination approach was performed by choosing 10 features which made a big contribution to the prediction models. And the Shapley Additive Explanation model was applied to interpret the most important characteristics that affected forecast outputs.ResultsThe random forest (RF) model had the highest average area under the receiver-operating-characteristic curve (AUC) for predicting LGA in the test set (0.843, 95% confidence interval [CI]: 0.714–0.974). Except for the logistic regression model (AUC: 0.603, 95%CI: 0.440–0.767), other models’ AUCs displayed well. Thereinto, the RF algorithm’s final prediction model using 10 characteristics achieved an average AUC of 0.821 (95% CI: 0.693–0.949).ConclusionThe prediction model based on machine learning might be a promising tool for the prenatal prediction of LGA births in women with radiation exposure before pregnancy.

ADHD classification with cross-dataset feature selection for biomarker consistency detection

Objective. Attention deficit hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder in children. While numerous intelligent methods are applied for its subjective diagnosis, they seldom consider the consistency problem of ADHD biomarkers. In practice, these data-driven approaches lead to varying learned features for ADHD classification across diverse ADHD datasets. This phenomenon significantly undermines the reliability of identified biomarkers and hampers the interpretability of these methods. Approach. In this study, we propose a cross-dataset feature selection (FS) module using a grouped SVM-based recursive feature elimination approach (G-SVM-RFE) to enhance biomarker consistency across multiple datasets. Additionally, we employ connectome gradient data for ADHD classification. In details, we introduce the G-SVM-RFE method to effectively concentrate gradient components within a few brain regions, thereby increasing the likelihood of identifying these regions as ADHD biomarkers. The cross-dataset FS module is integrated into an existing binary hypothesis testing (BHT) framework. This module utilizes external datasets to identify global regions that yield stable biomarkers. Meanwhile, given a dataset which waits for implementing the classification task as local dataset, we learn its own specific regions to further improve the performance of accuracy on this dataset. Main results. By employing this module, our experiments achieve an average accuracy of 96.7% on diverse datasets. Importantly, the discriminative gradient components primarily originate from the global regions, providing evidence for the significance of these regions. We further identify regions with the high appearance frequencies as biomarkers, where all the used global regions and one local region are recognized. Significance. These biomarkers align with existing research on impaired brain regions in children with ADHD. Thus, our method demonstrates its validity by providing enhanced biological explanations derived from ADHD mechanisms.

An Improved Fake News Detection Model by Applying a Recursive Feature Elimination Approach for Credibility Assessment and Uncertainty

World Health Organization (W.H.O) has coined the word “Infodemic” to refer to the dissemination of fake news during this pandemic, which is considered to be as harmful as the virus itself. Verifying the information available on the internet is a prerequisite to ensuring the ecosystem is maintained which is the driving force behind this work. The primary goal of this study is to address the problem of time-consuming automatically voluminous fake news detection of certain data and consider the uncertainty of data from causal relations using a rich feature set. This research produces significant feature reduction for reduced time and improved accuracy by filtering out significant features using recursive feature selection (RFE). The retained features by the RFE algorithm are also compared to a standard statistical measure of Pearson’s correlation to ensure no information loss while reducing features. The suggested methodology has also defined appropriate class output assurance levels and accurate prediction ambiguity for the fake identification jobs. Comparative analysis for existing methods for feature selection is performed. The result of experimentation testifies the improvement of a 6% increase in precision and a 97% reduction in execution time.

Automated detection of cybersecurity attacks in healthcare systems with recursive feature elimination and multilayer perceptron optimization

Widespread proliferation of interconnected healthcare equipment, accompanying software, operating systems, and networks in the Internet of Medical Things (IoMT) raises the risk of security compromise as the bulk of IoMT devices are not built to withstand internet attacks. In this work, we have developed a cyber-attack and anomaly detection model based on recursive feature elimination (RFE) and multilayer perceptron (MLP). The RFE approach selected optimal features using logistic regression (LR) and extreme gradient boosting regression (XGBRegressor) kernel functions. MLP parameters were adjusted by using a hyperparameter optimization and 10-fold cross-validation approach was performed for performance evaluations. The developed model was performed on various IoMT cybersecurity datasets, and attained the best accuracy rates of 99.99%, 99.94%, 98.12%, and 96.2%, using Edith Cowan University- Internet of Health Things (ECU-IoHT), Intensive Care Unit (ICU Dataset), Telemetry data, Operating systems’ data, and Network data from the testbed IoT/IIoT network (TON-IoT), and Washington University in St. Louis enhanced healthcare monitoring system (WUSTL-EHMS) datasets, respectively. The proposed method has the ability to counter cyber attacks in healthcare applications.

Assessment of Feature Selection for Liquefaction Prediction Based on Recursive Feature Elimination

This paper presents a machine learning model using a random forest (RF) algorithm with the recursive feature elimination (RFE) for the soil liquefaction prediction. The prediction model is tested on 253 CPT-based field data from different earthquakes. RFE, which is one of the feature selection methods, was adopted for eliminating irrelevant features in the mentioned dataset, and then the performance of the RFE-RF (i.e., the model determined by the RFE method) and the RF models with all variables were compared in terms of their performance matrices. The primary focus of this study is to investigate the effectiveness of the feature selection algorithm approach, therefore the raw data that is a benchmark dataset was used to compare the performance of the RFE-RF. The result showed that the RFE approach improved the overall accuracy of the liquefaction prediction.

An Adaptive Genetic Algorithm with Recursive Feature Elimination Approach for Predicting Malaria Vector Gene Expression Data Classification using Support Vector Machine Kernels

As mosquito parasites breed across many parts of the sub-Saharan Africa part of the world, infected cells embrace an unpredictable and erratic life period. Millions of individual parasites have gene expressions. Ribonucleic acid sequencing (RNA-seq) is a popular transcriptional technique that has improved the detection of major genetic probes. The RNA-seq analysis generally requires computational improvements of machine learning techniques since it computes interpretations of gene expressions. For this study, an adaptive genetic algorithm (A-GA) with recursive feature elimination (RFE) (A-GA-RFE) feature selection algorithms was utilized to detect important information from a high-dimensional gene expression malaria vector RNA-seq dataset. Support Vector Machine (SVM) kernels were used as the classification algorithms to evaluate its predictive performances. The feasibility of this study was confirmed by using an RNA-seq dataset from the mosquito Anopheles gambiae. The technique results in related performance had 98.3 and 96.7 % accuracy rates, respectively. HIGHLIGHTS Dimensionality reduction method based of feature selection Classification using Support vector machine Classification of malaria vector dataset using an adaptive GA-RFE-SVM GRAPHICAL ABSTRACT

CT-Based Hand-crafted Radiomic Signatures Can Predict PD-L1 Expression Levels in Non-small Cell Lung Cancer: a Two-Center Study.

Here, we used pre-treatment CT images to develop and evaluate a radiomic signature that can predict the expression of programmed death ligand 1 (PD-L1) in non-small cell lung cancer (NSCLC). We then verified its predictive performance by cross-referencing its results with clinical characteristics. This two-center retrospective analysis included 125 patients with histologically confirmed NSCLC. A total of 1287 hand-crafted radiomic features were observed from manually determined tumor regions. Valuable features were then selected with a ridge regression-based recursive feature elimination approach. Machine learning-based prediction models were then built from this and compared each other. The final radiomic signature was built using logistic regression in the primary cohort, and then tested in a validation cohort. Finally, we compared the efficacy of the radiomic signature to the clinical model and the radiomic-clinical nomogram. Among the 125 patients, 89 were classified as having PD-L1 positive expression. However, there was no significant difference in PD-L1 expression levels determined by clinical characteristics (P = 0.109-0.955). Upon selecting 9 radiomic features, we found that the logistic regression-based prediction model performed the best (AUC = 0.96, P < 0.001). In the external cohort, our radiomic signature showed an AUC of 0.85, which outperformed both the clinical model (AUC = 0.38, P < 0.001) and the radiomics-nomogram model (AUC = 0.61, P < 0.001). Our CT-based hand-crafted radiomic signature model can effectively predict PD-L1 expression levels, providing a noninvasive means of better understanding PD-L1 expression in patients with NSCLC.

Prognostic Modeling of Patients Undergoing Surgery Alone for Esophageal Squamous Cell Carcinoma: A Histopathological and Computed Tomography Based Quantitative Analysis.

ObjectiveTo evaluate the effectiveness of a novel computerized quantitative analysis based on histopathological and computed tomography (CT) images for predicting the postoperative prognosis of esophageal squamous cell carcinoma (ESCC) patients.MethodsWe retrospectively reviewed the medical records of 153 ESCC patients who underwent esophagectomy alone and quantitatively analyzed digital histological specimens and diagnostic CT images. We cut pathological images (6000 × 6000) into 50 × 50 patches; each patient had 14,400 patches. Cluster analysis was used to process these patches. We used the pathological clusters to all patches ratio (PCPR) of each case for pathological features and we obtained 20 PCPR quantitative features. Totally, 125 computerized quantitative (20 PCPR and 105 CT) features were extracted. We used a recursive feature elimination approach to select features. A Cox hazard model with L1 penalization was used for prognostic indexing. We compared the following prognostic models: Model A: clinical features; Model B: quantitative CT and clinical features; Model C: quantitative histopathological and clinical features; and Model D: combined information of clinical, CT, and histopathology. Indices of concordance (C-index) and leave-one-out cross-validation (LOOCV) were used to assess prognostic model accuracy.ResultsFive PCPR and eight CT features were treated as significant indicators in ESCC prognosis. C-indices adjusted for LOOCV were comparable among four models, 0.596 (Model A) vs. 0.658 (Model B) vs. 0.651 (Model C), and improved to 0.711with Model D combining information of clinical, CT, and histopathology (all p<0.05). Using Model D, we stratified patients into low- and high-risk groups. The 3-year overall survival rates of low- and high-risk patients were 38.0% and 25.0%, respectively (p<0.001).ConclusionQuantitative prognostic modeling using a combination of clinical data, histopathological, and CT images can stratify ESCC patients with surgery alone into high-risk and low-risk groups.

PredNTS: Improved and Robust Prediction of Nitrotyrosine Sites by Integrating Multiple Sequence Features.

Nitrotyrosine, which is generated by numerous reactive nitrogen species, is a type of protein post-translational modification. Identification of site-specific nitration modification on tyrosine is a prerequisite to understanding the molecular function of nitrated proteins. Thanks to the progress of machine learning, computational prediction can play a vital role before the biological experimentation. Herein, we developed a computational predictor PredNTS by integrating multiple sequence features including K-mer, composition of k-spaced amino acid pairs (CKSAAP), AAindex, and binary encoding schemes. The important features were selected by the recursive feature elimination approach using a random forest classifier. Finally, we linearly combined the successive random forest (RF) probability scores generated by the different, single encoding-employing RF models. The resultant PredNTS predictor achieved an area under a curve (AUC) of 0.910 using five-fold cross validation. It outperformed the existing predictors on a comprehensive and independent dataset. Furthermore, we investigated several machine learning algorithms to demonstrate the superiority of the employed RF algorithm. The PredNTS is a useful computational resource for the prediction of nitrotyrosine sites. The web-application with the curated datasets of the PredNTS is publicly available.

RETRACTED ARTICLE: High-performance in classification of heart disease using advanced supercomputing technique with cluster-based enhanced deep genetic algorithm

In the world, cardiovascular disease is a common, debilitating disease that affects human lives in many different ways. It is important to detect heart disease effectively and reliably in order to avoid potential heart failures. Since heart failure is a major risk, it should be successfully treated in the early stages of heart disease. Many programs exist that can diagnose heart disease at an early stage using machine learning techniques. But these predictive systems are difficult to predict the heart conditions accurately with a minimum of time. Today, the stochastic gradient boosting with recursive feature elimination approach has been developed for feature selection. The results of the clustering were based on the adaptive Harris Hawk optimization algorithm. The selected features allowed us to better identify people with heart disease because they all have the same features. Classification is achieved using an improved deep genetic algorithm (EDGA). The system enhances the DNN's initial weights by using an augmented genetic algorithm and proposing the best initial weights for the DNN using neural network. The technique is illustrated using the publicly accessible dataset from the UCI machine learning repository. The study found that EDGA is well suited for predicting heart disease.

A Machine Learning Approach to Monitor the Emergence of Late Intrauterine Growth Restriction.

Late intrauterine growth restriction (IUGR) is a fetal pathological condition characterized by chronic hypoxia secondary to placental insufficiency, resulting in an abnormal rate of fetal growth. This pathology has been associated with increased fetal and neonatal morbidity and mortality. In standard clinical practice, late IUGR diagnosis can only be suspected in the third trimester and ultimately confirmed at birth. This study presents a radial basis function support vector machine (RBF-SVM) classification based on quantitative features extracted from fetal heart rate (FHR) signals acquired using routine cardiotocography (CTG) in a population of 160 healthy and 102 late IUGR fetuses. First, the individual performance of each time, frequency, and nonlinear feature was tested. To improve the unsatisfactory results of univariate analysis we firstly adopted a Recursive Feature Elimination approach to select the best subset of FHR-based parameters contributing to the discrimination of healthy vs. late IUGR fetuses. A fine tuning of the RBF-SVM model parameters resulted in a satisfactory classification performance in the training set (accuracy 0.93, sensitivity 0.93, specificity 0.84). Comparable results were obtained when applying the model on a totally independent testing set. This investigation supports the use of a multivariate approach for the in utero identification of late IUGR condition based on quantitative FHR features encompassing different domains. The proposed model allows describing the relationships among features beyond the traditional linear approaches, thus improving the classification performance. This framework has the potential to be proposed as a screening tool for the identification of late IUGR fetuses.

A Machine Learning Approach to EEG-based Prediction of Human Affective States Using Recursive Feature Elimination Method

Emotion recognition, as a branch of affective computing, has attracted great attention in the last decades as it can enable more natural brain-computer interface systems. Electroencephalography (EEG) has proven to be an effective modality for emotion recognition, with which user affective states can be tracked and recorded, especially for primitive emotional events such as arousal and valence. Although brain signals have been shown to correlate with emotional states, the effectiveness of proposed models is somewhat limited. The challenge is improving accuracy, while appropriate extraction of valuable features might be a key to success. This study proposes a framework based on incorporating fractal dimension features and recursive feature elimination approach to enhance the accuracy of EEG-based emotion recognition. The fractal dimension and spectrum-based features to be extracted and used for more accurate emotional state recognition. Recursive Feature Elimination will be used as a feature selection method, whereas the classification of emotions will be performed by the Support Vector Machine (SVM) algorithm. The proposed framework will be tested with a widely used public database, and results are expected to demonstrate higher accuracy and robustness compared to other studies. The contributions of this study are primarily about the improvement of the EEG-based emotion classification accuracy. There is a potential restriction of how generic the results can be as different EEG dataset might yield different results for the same framework. Therefore, experimenting with different EEG dataset and testing alternative feature selection schemes can be very interesting for future work.

MUSP: a high-accuracy map of thein situcrosstalk of ubiquitylation and SUMOylation proteome predicted via the feature enhancement approach

Reversible post-translational modification (PTM) orchestrates various biological processes by changing the properties of proteins. Since many proteins are multiply modified by PTMs, identification of PTM crosstalk site has emerged to be an intriguing topic and attracted much attention. In this study, we systematically deciphered the in situ crosstalk of ubiquitylation and SUMOylation that co-occurs on the same lysine residue. We first collected 3363 ubiquitylation-SUMOylation (UBS) crosstalk site on 1302 proteins and then investigated the prime sequence motifs, the local evolutionary degree and the distribution of structural annotations at the residue and sequence levels between the UBS crosstalk and the single modification sites. Given the properties of UBS crosstalk sites, we thus developed the mUSP classifier to predict UBS crosstalk site by integrating different types of features with two-step feature optimization by recursive feature elimination approach. By using various cross-validations, the mUSP model achieved an average area under the curve (AUC) value of 0.8416, indicating its promising accuracy and robustness. By comparison, the mUSP has significantly better performance with the improvement of 38.41 and 51.48% AUC values compared to the cross-results by the previous single predictor. The mUSP was implemented as a web server available at http://bioinfo.ncu.edu.cn/mUSP/index.html to facilitate the query of our high-accuracy UBS crosstalk results for experimental design and validation.

Elaboration of a multisequence MRI-based radiomics signature for the preoperative prediction of the muscle-invasive status of bladder cancer: a double-center study.

To develop a multisequence MRI-based radiomics signature for the preoperative prediction of the muscle-invasive status of bladder cancer (BCa). This retrospective study involved 106 eligible patients from two independent clinical centers. All patients underwent a preoperative 3.0T MRI scan with T2-weighted image (T2WI) and multi-b-value diffusion-weighted image (DWI) sequences. In total, 1404 radiomics features were extracted from the largest region of the reported tumor locations on the T2WI, DWI, and corresponding apparent diffusion coefficient map (ADC) of each patient. A radiomics signature, namely the Radscore, was then generated using the recursive feature elimination approach and a logistic regression algorithm in a training cohort (n = 64). Its performance was then validated in an independent validation cohort (n = 42). The primary imaging and clinical factors in conjunction with the Radscore were used to determine whether the performance could be further improved. The Radscore, generated by 36 selected radiomics features, demonstrated a favorable ability to predict muscle-invasive BCa status in both the training (AUC 0.880) and validation (AUC 0.813) cohorts. Subsequently, integrating the two independent predictors (including the Radscore and MRI-determined tumor stalk) into a nomogram exhibited more favorable discriminatory performance, with the AUC improved to 0.924 and 0.877 in both cohorts, respectively. The proposed multisequence MRI-based radiomics signature alone could be an effective tool for quantitative prediction of muscle-invasive status of BCa. Integrating the Radscore with MRI-determined tumor stalk could further improve the discriminatory power, realizing more accurate prediction of nonmuscle-invasive and muscle-invasive BCa. • DWI is superior to T2WI sequence in reflecting the heterogeneous differences between NMIBC and MIBC, and multisequence MRI helps in the preoperative prediction of muscle-invasive status of BCa. • Co-occurrence (CM), run-length matrix (RLM), and gray-level size zone matrix (GLSZM) features were the favorable feature categories for the prediction of muscle-invasive status of BCa. • The Radscore (proposed multisequence MRI-based radiomics signature) helps predict preoperatively muscle invasion. Combination with the MRI-determined tumor stalk further improves prediction.

Abstract 2413: Genomewide expression profiling identifies a novel gene-expression signature for recurrence prediction in patients with early-onset colorectal cancer (EOCRC)

Abstract Purpose: Although the incidence of CRC has declined over the last decade, recent evidence indicates that the incidence of early onset CRC (EOCRC; patients younger than age 50) is actually on the rise. The disease presentation in patients with EOCRC is quite distinct, with often the disease diagnosis at more advanced stages, poorer histological grade, and higher local recurrence rates, compared to late onset CRC (LOCRC). Hence, availability of biomarkers that can predict recurrence in high-risk EOCRC patients is much needed as it will lead to improved prognosis, and tailoring personalized treatment regimens in this population. Herein, we have conducted a genomewide biomarker discovery approach, followed by robust bioinformatic analysis to establish a novel gene-based recurrence prediction classifier in patients with EOCRC. Methods: We undertook a comprehensive and systematic biomarker discovery effort by first analyzing six independent, genomewide expression datasets (GSE14333, GSE17538, GSE3311, GSE37892, GSE39084 and GSE39582) of EOCRC patients to identify (discovery dataset; n=81) and validate (validation dataset; n=75) a gene signature for predicting recurrence in EOCRC. Subsequently, the robustness of this signature was analytically evaluated in an in-house clinical training cohort (n=54). Using recursive feature elimination approach based on random forest classification, the signature was further refined and the gene signature was validated in another independent clinical cohort (n=54). Results: We initially identified a 25-gene signature from a total of 3591 mRNAs, which showed significant potential for predicting recurrence free survival (RFS) in both the in-silico discovery (HR, 32.83, 95% confidential interval (CI), 10.73 - 100.51, P&amp;lt;0.001) and validation datasets in the discovery phase (HR, 4.11, 95% CI, 1.77 - 9.56, P &amp;lt;0.001). Subsequently, based on our clinical training cohort, we developed a reduced and optimized 5-gene signature, which demonstrated robust prognostic power in both the clinical training cohort (HR, 3.67; 95% CI, 1.48-9.52; P=0.003) and an independent validation cohort (HR, 7.53; 95% CI, 1.73-51.3; P=0.007). In univariate and multivariate cox regression analysis, our 5-gene signature emerged as an independent predictor of worse RFS. Furthermore, a combination signature comprising of lymph node metastasis and our 5-gene signature demonstrated an even superior predictive performance compared to individual factors in both cohorts (training cohort: HR 4.56; 95% CI, 0.65-0.88; validation cohort: HR 8.25, 0.86; 95% CI, 0.74-0.94). Conclusion: Our novel 5-gene signature provides a robust prognostic tool for patients with EOCRC; which can be clinically transformative in guiding post-surgical decision-making to improve survival and management of young patients suffering from this malignancy. Citation Format: Kensuke Yamamura, Lina Zhu, Raju Kandimalla, Takatoshi Matsuyama, Yusuke Kinugasa, Francesc Balaguer, Hideo Baba, Xin Wang, Ajay Goel. Genomewide expression profiling identifies a novel gene-expression signature for recurrence prediction in patients with early-onset colorectal cancer (EOCRC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2413.

A predictive nomogram for individualized recurrence stratification of bladder cancer using multiparametric MRI and clinical risk factors.

Preoperative prediction of bladder cancer (BCa) recurrence risk is critical for individualized clinical management of BCa patients. To develop and validate a nomogram based on radiomics and clinical predictors for personalized prediction of the first 2 years (TFTY) recurrence risk. Retrospective. Preoperative MRI datasets of 71 BCa patients (34 recurrent) were collected, and divided into training (n = 50) and validation cohorts (n = 21). 3.0T MRI/T2 -weighted (T2 W), multi-b-value diffusion-weighted (DW), and dynamic contrast-enhanced (DCE) sequences. Radiomics features were extracted from the T2 W, DW, apparent diffusion coefficient, and DCE images. A Rad_Score model was constructed using the support vector machine-based recursive feature elimination approach and a logistic regression model. Combined with the important clinical factors, including age, gender, grade, and muscle-invasive status (MIS) of the archived lesion, tumor size and number, surgery, and image signs like stalk and submucosal linear enhancement, a radiomics-clinical nomogram was developed, and its performance was evaluated in the training and the validation cohorts. The potential clinical usefulness was analyzed by the decision curve. Univariate and multivariate analyses were performed to explore the independent predictors for BCa recurrence prediction. Of the 1872 features, the 32 with the highest area under the curve (AUC) of receiver operating characteristic were selected for the Rad_Score calculation. The nomogram developed by two independent predictors, MIS and Rad_Score, showed good performance in the training (accuracy 88%, AUC 0.915, P << 0.01) and validation cohorts (accuracy 80.95%, AUC 0.838, P = 0.009). The decision curve exhibited when the risk threshold was larger than 0.3, more benefit was observed by using the radiomics-clinical nomogram than using the radiomics or clinical model alone. The proposed radiomics-clinical nomogram has potential in the preoperative prediction of TFTY BCa recurrence. 3 Technical Efficacy: Stage 3 J. Magn. Reson. Imaging 2019;50:1893-1904.

SVM recursive feature elimination analyses of structural brain MRI predicts near-term relapses in patients with clinically isolated syndromes suggestive of multiple sclerosis.

Machine learning classification is an attractive approach to automatically differentiate patients from healthy subjects, and to predict future disease outcomes. A clinically isolated syndrome (CIS) is often the first presentation of multiple sclerosis (MS), but it is difficult at onset to predict who will have a second relapse and hence convert to clinically definite MS. In this study, we thus aimed to distinguish CIS converters from non-converters at onset of a CIS, using recursive feature elimination and weight averaging with support vector machines. We also sought to assess the influence of cohort size and cross-validation methods on the accuracy estimate of the classification. We retrospectively collected 400 patients with CIS from six European MAGNIMS MS centres. Patients underwent brain MRI at onset of a CIS according to local standard-of-care protocols. The diagnosis of clinically definite MS at one-year follow-up was the standard against which the accuracy of the model was tested. For each patient, we derived MRI-based features, such as grey matter probability, white matter lesion load, cortical thickness, and volume of specific cortical and white matter regions. Features with little contribution to the classification model were removed iteratively through an interleaved sample bootstrapping and feature averaging approach. Classification of CIS outcome at one-year follow-up was performed with 2-fold, 5-fold, 10-fold and leave-one-out cross-validation for each centre cohort independently and in all patients together. The estimated classification accuracy across centres ranged from 64.9% to 88.1% using 2-fold cross-validation and from 73% to 92.9% using leave-one-out cross-validation. The classification accuracy estimate was higher in single-centre, smaller data sets than in combinations of data sets, being the lowest when all patients were merged together. Regional MRI features such as WM lesions, grey matter probability in the thalamus and the precuneus or cortical thickness in the cuneus and inferior temporal gyrus predicted the occurrence of a second relapse in patients at onset of a CIS using support vector machines. The increased accuracy estimate of the classification achieved with smaller and single-centre samples may indicate a model bias (overfitting) when data points were limited, but also more homogeneous. We provide an overview of classifier performance from a range of cross-validation schemes to give insight into the variability across schemes. The proposed recursive feature elimination approach with weight averaging can be used both in single- and multi-centre data sets in order to bridge the gap between group-level comparisons and making predictions for individual patients.

Machine‐Learning Variables at Different Scales vs. Knowledge‐based Variables for Mapping Multiple Soil Properties

Core IdeasExplored the impact of variables selected by different means on mapping at a watershed scale area.Identified influential topographic variables with appropriate scales for mapping soil properties.Finding the appropriate scales of local attributes is helpful for mapping soil properties accurately.Local attribute scale helped map accuracy vs. use of local and regional attributes at single scale.The influential environmental variables and their appropriate scales for mapping different soil properties are usually different. Comparisons between variables selected using machine learning and knowledge‐based approaches and their impacts for soil mapping are necessary to provide guidelines on selecting environmental variables. This study compared topographic variables with single and multiple scales selected for five soil properties (topsoil clay content, sand content, silt content, topsoil organic matter content (SOM), and soil depth) using a recursive feature elimination (RFE) algorithm and variables selected based on expert knowledge on mapping. One hundred and seventy‐three variables were generated including five single‐scale variables derived with a 3 × 3 neighborhood size and seven multi‐scale variables with various neighborhood sizes. Subsets 1 and 2 were selected from single‐scale variables (Pool 1) and all variables (Pool 2), respectively. A reference Subset 3 was selected based on soil pedogenesis knowledge. Results showed that variables in Subset 1 included both local and regional terrain attributes, but were considerably different to the reference variables. When considering scales, local variables at specific scales became the most important variables. Mapping accuracies using Subset 2 were the highest for all soil properties. This indicates that finding the appropriate scales of local attributes and using common local attributes with appropriate scales is more helpful than using both local and regional attributes at single scale for mapping soil properties accurately. The RFE approach is efficient for selection of variables. Also, the input environmental variables for selection determine whether the selected variables perform better than all input variables without selection for predicting soil properties.