ROC Area Research Articles

Are merchant views useful for patellofemoral osteoarthritis?

Purpose: To evaluate the effectiveness of merchant view radiographs (MVR) to identify and grade patellofemoral osteoarthritis (PFOA) relative to MRI. Methods: A cohort of patients who underwent partial meniscectomy at a single large institution aged at least 45 years with a knee MRI within the preceding 6 months and with MVR within the preceding year was selected. PFOA was evaluated on MVR using KL grading (0-4) and minimum joint space width (mJSW) by a musculoskeletal (MSK) radiologist or a trained medical student. Size and depth of cartilage loss in the patellofemoral compartment was graded on MRI by an MSK radiologist using clinically used grading systems. MRI findings were classified into none, mild, moderate, or severe categories apriori. Positive and negative predictive values (PPV and NPV) and ROC curves for KL grading were calculated for detecting PFOA on MRI as none or mild (negative) versus moderate or severe (positive). Prediction models were built for comparing ROC curves for KL and mJSW variables. Results: 442 patients were identified with MVR and MRI. MVR KL grades of 0 or 1 had relatively good NPV for OA on MRI (81.3%, 95% CI: [75.1%, 78.5%]). MVR KL grades of 4 had similarly good PPV (86.7%, 95% CI: [69.5%,99%] ). MVR KL grades of 2 had relatively poor NPV (68.0%) and KL grades of 3 had similarly poor PPV (69.0%). MVR KL grades of 2 and 3 accounted for 62.6% of the radiographs. ROC areas were poor to fair for KL grading (ROC area 0.696) and for a mJSW model (ROC area 0.686). A model with both KL and JSW (ROC area 0.709) was not significantly better than just KL grading (p=0.623). Conclusions: MVR may be diagnostically useful for PFOA when they are essentially normal (KL 0 or 1) or very abnormal (KL 4). MRI should be more strongly considered if there is mild to moderate radiographic PFOA (KL 2 or 3), as radiographs are less accurate at these grades.Tabled 1MRI vs X-ray PFOAKL grade PF = 0KL grade PF = 1KL grade PF = 2KL grade PF = 3KL grade PF = 4MEDIAN JSW_PFMRI=non213059717.6MRI=mild2942901117.1MRI=moderate713461326.4MRI=severe262427115.4 Open table in a new tab

Ganglion cell layer analysis with deep learning in glaucoma diagnosis

Abstract Objective To determine and compare the diagnostic precision in glaucoma of two deep learning models using infrared images of the optic nerve, eye fundus, and the ganglion cell layer (GCL). Methods We have selected a sample of normal and glaucoma patients. Three infrared images were registered with a spectral-domain optical coherence tomography (SD-OCT). The first corresponds to the confocal scan image of the fundus, the second is a cut-out of the first centered on the optic nerve, and the third was the SD-OCT image of the GCL. Our deep learning models are developed on the MatLab platform with the ResNet50 and VGG19 pre-trained neural networks. Results 498 eyes of 298 patients were collected. Of the 498 eyes, 312 are glaucoma and 186 are normal. In the test, the precision of the models was 96% (ResNet50) and 96% (VGG19) for the GCL images, 90% (ResNet50) and 90% (VGG19) for the optic nerve images and 82% (ResNet50) and 84% (VGG19) for the fundus images. The ROC area in the test was 0.96 (ResNet50) and 0.97 (VGG19) for the GCL images, 0.87 (ResNet50) and 0.88 (VGG19) for the optic nerve images, and 0.79 (ResNet50) and 0.81 (VGG19) for the fundus images. Conclusions Both deep learning models, applied to the GCL images, achieve high diagnostic precision, sensitivity and specificity in the diagnosis of glaucoma.

Prediction efficacy of feature classification of solitary pulmonary nodules based on CT radiomics

ObjectiveTo investigate the relationship between CT radiomic features, pathological classification of pulmonary nodules, and evaluate the prediction effect of different stratified progressive radiomic models on the pathological classification of pulmonary nodules. MethodsAltogether, 189 patients pathologically confirmed with pulmonary nodules from July 2017 to August 2019 who had complete data were enrolled, including 71 patients with benign nodules, 51 with malignant non-invasive nodules, and 67 with invasive nodules. Three CT radiomic models were established respectively. Model 1 classified benign and malignant nodules (including malignant non-invasive and invasive nodules). Model 2 classified malignant non-invasive and invasive nodules. Model 3 classified benign, malignant non-invasive, and invasive nodules. High-throughput feature collection was carried out for all delineated regions of interest (ROIs), and the best models were established by screening features and classifiers using intelligent methods. ROC curves and areas under the curve (AUCs) were used to evaluate the prediction efficacy of the models by calculating the sensitivity, specificity, accuracies, positive predictive values, and negative predictive values. ResultsThrough Models 1, 2, and 3, we screened out 20, 2, and 20 radiomic features, respectively, and plotted the ROC curves. In the test group, the AUC values were 0.85, 0.89, and 0.84, respectively; the sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were 79.66 %, 70.42 %, 84.59 %, and 81.74 % and 67.57％ for Model 1, 88.06 %, 74.51 %, 82.2 %, 81.94 %, and 82.61 % for Model 2, and 71.34 %, 85.05 %, 70.37 %, 83.2 %, and 76.3 % for Model 3. ConclusionThe radiomic feature models based on CT images could well reflect the differences between benign nodules, malignant non-invasive nodules, and invasive nodules, and assist in their classification.

Detection of Streptococcus equi subsp. equi in guttural pouch lavage samples using a loop-mediated isothermal nucleic acid amplification microfluidic device.

BackgroundRapid point‐of‐care (POC) detection of Streptococcus equi subsp. equi (S. equi) would theoretically reduce the spread of strangles by identifying index and carrier horses.HypothesisThat the eqbE isothermal amplification (LAMP) assay, and the same eqbE LAMP assay tested in a microfluidic device format, are comparable to a triplex real‐time quantitative polymerase chain reaction (qPCR) assay that is commonly used in diagnostic labs.SamplesSixty‐eight guttural pouch lavage (GPL) specimens from horses recovering from strangles.MethodsGuttural pouch lavage specimens were tested for S. equi retrospectively using the benchtop eqbE LAMP, the eqbE LAMP microfluidic device, and compared to the triplex qPCR, that detects 2 S. equi‐specific genes, eqbE and SEQ2190, as the reference standard using the receiver operating characteristic area under the curve (ROC).ResultsThe 27/68 specimens were positive by benchtop eqbE LAMP, 31/64 by eqbE LAMP microfluidic device, and 12/67 by triplex qPCR. Using the triplex PCR as the reference, the benchtop eqbE LAMP showed excellent discrimination (ROC Area = 0.813, 95% confidence interval [CI] = 0.711‐0.915) as did the LAMP microfluidic device (ROC Area = 0.811, 95% CI = 0.529‐0.782). There was no significant difference between the benchtop LAMP and LAMP microfluidic device (ROC Area 0.813 ± 0.055 vs 0.811 ± 0.034, P = .97).ConclusionsThe eqbE LAMP microfluidic device detected S. equi in GPL specimens from convalescent horses. This assay shows potential for development as a POC device for rapid, sensitive, accurate, and cost‐efficient detection of S. equi.

Clinical variables that help in predicting the presence of autoantibodies in patients with acute encephalitis

ObjectiveTo identify clinical variables that could predict the presence of autoantibodies in patients with acute encephalitis. MethodsAn observational, retrospective study from May 2011 to May 2017. Clinical, EEG, brain MRI data, and antibodies against human neuronal antigens (NMDAR, GABAR, AMPAR, LGI1, CASPR2, and GAD) from 158 patients with criteria for possible autoimmune encephalitis were analyzed to create a predictive model for this disease. ResultsWe analyzed 158 samples, of which 18 cases were positive for anti-NMDAR, 2 for anti-LGI1, and 2 for anti-GAD. Seven of the 18 positive NMDAR patients were children, and 12 were female. Behavioral disorder, epileptic seizures, movement disorder, and altered level of consciousness were the frequent symptoms with >75 % sensitivity in positive anti-NMDAR patients. Other symptoms, such as language disorder, psychosis, hypoventilation, altered wake and sleep cycle, and cognitive impairment, had a sensitivity >55 %. Abnormal EEG findings had a high sensitivity (99.4 %). Brain MRI suggestive of encephalitis was observed in 7 of the positive cases for NMDAR. Abnormal CSF findings were reported in 12 patients positive for this receptor (sensitivity 70.6 %). With 7 of these symptoms, we obtained a sensitivity of 70 % and specificity of 81 % for the presence of anti-NMDAR antibodies (ROC Area 82 %). However, to predict that a patient with subacute encephalitis may have an autoimmune cause, the patient should include clinical manifestations such as movement disorder, behavioral disorder, hypoventilation, dysautonomia, and alteration of the wake and sleep cycle. Children were significantly more likely than adults with autoimmune encephalitis to experience chorea and status epilepticus (p < 0.05). ConclusionsAnti-NMDAR encephalitis was more frequent in females and children. The repertoire of autoimmune encephalitis in children is different from adults. The presence of subacute behavioral changes, epileptic seizures, movement disorders, altered consciousness, hypoventilation, dysautonomia, and altered wake and sleep cycle predicted autoimmune encephalitis in our series.

651 Clinical efficacy of a machine learning model for prediction of shoulder dystocia

A newly introduced shoulder dystocia (SD) machine learning (ML) model (Tsur et al, 2019) uses maternal demographics, obstetric history, and sonographic evaluation within 5 weeks from delivery. Our primary aim was to externally validate the calculator for prediction of SD or birth injury in patients who attempted a vaginal delivery. Secondary objective was to assess the clinical utility in optimizing the number of cesarean delivery (CD) for suspected large for gestational age. We conducted a multicenter retrospective study of all non-anomalous singleton pregnancies with a sonographic estimated fetal weight (EFW) ≥ 35 weeks between January 2013 and June 2019. All people who attempted labor or underwent a cesarean delivery due to suspected fetal macrosomia with EFW ≥4000g within 4 weeks from delivery were included. A matched number of people with EFW between 3500g and 3999g were also included in the analysis. The ROC area under the curve (AUC) was used as the primary metric of the model performance and was compared to EFW (ACOG recommendations). A total of 490 people met inclusion criteria, of whom 381 (77.8%) attempted labor and 109 (22.2%) had a CD for suspected fetal macrosomia. We had 19 cases of SD or birth injury in our cohort. The ML model achieved an AUC of 0.77, significantly better than EFW alone (0.61, p-value = 0.028). Furthermore, we evaluated the clinical performance of various risk scores above 0.5 (Table). A risk score of 0.92 could identify and prevent 53% of SD or neonatal injury cases while reducing the rate of CD for suspected macrosomia by 6%. In this validation cohort, using a ML model was found to be more accurate than EFW alone in prediction of SD or neonatal birth injury. Based on these findings, we are planning a prospective study investigating efficacy of the ML model compared to the current recommendations in reducing the rate of SD, associated neonatal morbidity and cesarean delivery for the indication of macrosomia.

An Analysis of Multiclass Imbalanced Data Problem in Machine Learning for Network Attack Detections

In the current trend, machine learning has been used widely for network attack detection. The performance of a machine learning model depends on its training dataset and the dataset distribution. Network attack detection is one of the problems that usually suffer from the imbalanced data distribution. However, the effect of this imbalanced data is generally neglected by researchers. Therefore, in this research, we studied the impact of an imbalanced multiclass dataset to the machine learning performance. Five state-of-the-art machine learning algorithms were used in this study, and the classifiers that can classify the minority class or the majority class instances accurately were also identified. In this research, the performances of these classifiers were evaluated by using the performance metrics: true positive rate, false positive rate, precision, F-measure, ROC area, and classification accuracy. The results show that the J48 classifier outperforms the other four classifiers in every aspect. Besides, Naïve Bayes, J48, has also worked as the best classifier that able to classifies the instances of the minority class accurately.

929 Predictive performance of race/ethnicity based growth velocities in pregnancies complicated by diabetes

Recently, the NICHD published a race/ethnicity-based fetal growth velocity percentile calculator derived from a cohort of low-risk pregnancies. Fetal growth velocity is accelerated in pregnancies affected by diabetes. However, use of growth velocity percentiles are untested in this population. We sought to determine the test performance of growth velocity percentiles in pregnancies with diabetes. Retrospective cohort study of all women with pregestational and gestational diabetes with ultrasound data at two third trimester time points (28w0d-31w6d and 32w0d-35w6d) between 2009-2019 at a single academic institution. Self-reported race/ethnicity and fetal biometry were used to calculate abdominal circumference (AC) and estimated fetal weight (EFW) velocities between time points. Traditional AC and EFW percentiles were recorded for the second ultrasound for comparison. ROC curves and areas under the curve (AUC) were compared between parameters for prediction of composite neonatal outcome and LGA. Test statistics were calculated and compared at optimal cut points identified by the Youden J statistic. There were 1,093 pregnancies included in the analysis with 599 (54.8%) and 297 (27%) complicated by adverse composite outcome or LGA birthweight. No differences in AUCs for neonatal composite outcome were found (p=.234; Figure 1). Traditional EFW and AC biometric percentiles were significantly better predictors of LGA birthweight than either velocity parameter (Figure 2). The optimal cut point for EFW velocity (51%) had poor overall performance with a sensitivity 65.7% (CI 60.0%-71.0%) and positive predictive value (PPV) of 35.6% (CI 31.6-39.8%) for LGA. Results were similar for AC velocity (cut point 57%) with a sensitivity of 57.6% (CI 51.7%-63.3%) and PPV of 36.9% (CI 32.5%-41.4%). Neither traditional ultrasound biometry nor race/ethnicity-based growth velocity accurately predicted adverse neonatal outcomes in pregnancies complicated by diabetes. However, traditional ultrasound biometry was superior to both EFW and AC growth velocities for prediction of LGA birth weight.View Large Image Figure ViewerDownload Hi-res image Download (PPT)

Empirical Assessment of Machine Learning Techniques for Software Requirements Risk Prediction

Software risk prediction is the most sensitive and crucial activity of Software Development Life Cycle (SDLC). It may lead to the success or failure of a project. The risk should be predicted earlier to make a software project successful. A model is proposed for the prediction of software requirement risks using requirement risk dataset and machine learning techniques. In addition, a comparison is made between multiple classifiers that are K-Nearest Neighbour (KNN), Average One Dependency Estimator (A1DE), Naïve Bayes (NB), Composite Hypercube on Iterated Random Projection (CHIRP), Decision Table (DT), Decision Table/Naïve Bayes Hybrid Classifier (DTNB), Credal Decision Trees (CDT), Cost-Sensitive Decision Forest (CS-Forest), J48 Decision Tree (J48), and Random Forest (RF) achieve the best suited technique for the model according to the nature of dataset. These techniques are evaluated using various evaluation metrics including CCI (correctly Classified Instances), Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Relative Absolute Error (RAE), Root Relative Squared Error (RRSE), precision, recall, F-measure, Matthew’s Correlation Coefficient (MCC), Receiver Operating Characteristic Area (ROC area), Precision-Recall Curves area (PRC area), and accuracy. The inclusive outcome of this study shows that in terms of reducing error rates, CDT outperforms other techniques achieving 0.013 for MAE, 0.089 for RMSE, 4.498% for RAE, and 23.741% for RRSE. However, in terms of increasing accuracy, DT, DTNB, and CDT achieve better results.

PVT1 signals an androgen-dependent transcriptional repression program in prostate cancer cells and a set of the repressed genes predicts high-risk tumors

BackgroundAndrogen receptor (AR) and polycomb repressive complex 2 (PRC2) are known to co-occupy the loci of genes that are downregulated by androgen-stimulus. Long intergenic non-coding RNA (lincRNA) PVT1 is an overexpressed oncogene that is associated with AR in LNCaP prostate cancer cells, and with PRC2 in HeLa and many other types of cancer cells. The possible involvement of PVT1 in mediating androgen-induced gene expression downregulation in prostate cancer has not been explored.MethodsLNCaP cell line was used. Native RNA-binding-protein immunoprecipitation with anti-AR or anti-EZH2 was followed by RT-qPCR with primers for PVT1. Knockdown of PVT1 with specific GapmeRs (or a control with scrambled GapmeR) was followed by differentially expressed genes (DEGs) determination with Agilent microarrays and with Significance Analysis of Microarrays statistical test. DEGs were tested as a tumor risk classifier with a machine learning Random Forest algorithm run with gene expression data from all TCGA-PRAD (prostate adenocarcinoma) tumors as input. ChIP-qPCR was performed for histone marks at the promoter of one DEG.ResultsWe show that PVT1 knockdown in androgen-stimulated LNCaP cells caused statistically significant expression upregulation/downregulation of hundreds of genes. Interestingly, PVT1 knockdown caused upregulation of 160 genes that were repressed by androgen, including a significantly enriched set of tumor suppressor genes, and among them FAS, NOV/CCN3, BMF, HRK, IFIT2, AJUBA, DRAIC and TNFRSF21. A 121-gene-set (out of the 160) was able to correctly predict the classification of all 293 intermediate- and high-risk TCGA-PRAD tumors, with a mean ROC area under the curve AUC = 0.89 ± 0.04, pointing to the relevance of these genes in cancer aggressiveness. Native RIP-qPCR in LNCaP showed that PVT1 was associated with EZH2, a component of PRC2. PVT1 knockdown followed by ChIP-qPCR showed significant epigenetic remodeling at the enhancer and promoter regions of tumor suppressor gene NOV, one of the androgen-repressed genes that were upregulated upon PVT1 silencing.ConclusionsOverall, we provide first evidence that PVT1 was involved in signaling a genome-wide androgen-dependent transcriptional repressive program of tumor suppressor protein-coding genes in prostate cancer cells. Identification of transcriptional inhibition of tumor suppressor genes by PVT1 highlights the pathway to the investigation of mechanisms that lie behind the oncogenic role of PVT1 in cancer.DiTiwbuLh6zGF35UTdvYfAVideo

Spam detection on Twitter using a support vector machine and users’ features by identifying their interactions

Spam tweets might cause numerous problems for users. An automatic method is introduced as a proposed method to detect spam tweets. This method is based on pre-processing and feature extraction steps. The pre-processing step is significant for our problem due to the specific structure of tweets. The pre-processing step is performed in such a way that after which only the words remain in each tweet that can play a key role in determining whether the tweet is spam or non-spam. In the proposed method, the features are classified into five classes of user profile features, account information features, user activity based features, user interaction based features, and tweet content-based features including 28 different features. In the feature selection step, an optimal subset of these features is selected for the learning process. However, a support vector classifier is used for the learning process by two Gaussian and polynomial kernels. Finally, the proposed method is compared with multi-layer perceptron (MLP), Naive Bayes (NB), random forest (RF), and k-nearest neighbors (KNN) methods in terms of standard criteria. The obtained results show the superiority of the proposed method using support vector machine (SVM) algorithm and polynomial kernel with 0.988 precision, 0.953 efficiency, 0.96 accuracy, F-0.969, and 0.985 ROC area under the curve compared to the other methods, indicating that the proposed method has better performance overall.

Angiographic Scoring System for Predicting Successful Percutaneous Coronary Intervention of In-Stent Chronic Total Occlusion.

The purpose of this study was to develop a scoring model to predict the technical success of recanalizing via antegrade approach in-stent chronic total occlusion (IS-CTO) by percutaneous coronary intervention (PCI). We retrospectively collected data from 474 patients who underwent an uneasy IS-CTO PCI via antegrade approach from January 2015 to December 2018, consecutively. We selected clinical and angiographic factors and utilized a derivation and validation cohort (4:1 sampling ratio) analysis. Factors with strong correlations with technical failure, according to multivariable analysis, were assigned 1 point, and a scoring system with a 4-point maximum was established. The model was then validated with a validation cohort. The overall procedural success rate was 77.4%. On multivariable analysis, the factors that correlated with technical failure were proximal bending (beta coefficient [β] = 2.142), tortuosity (β= 2.622), stent under expansion (β = 3.052), and poor distal landing zone (β = 2.004). The IS-CTO score demonstrated good calibration and excellent predicting capacity in the derivation (receiver-operator characteristic [ROC] area = 0.973 and Hosmer-Lemeshow chi-squared = 5.252; p = 0.072) and validation (ROC area = 0.976 and Hosmer-Lemeshow chi-squared = 0.916; p = 0.632) cohorts. In the validation subset, the IS-CTO score demonstrated superior performance to the Japanese chronic total occlusion score (J-CTO) and PROGRESS CTO scores for predicting technical success (area under the a curve [AUC] 0.976 vs. 0.642 vs. 0.579, respectively; difference in AUC between the IS-CTO score and J-CTO score = 0.334, p < 0.01; difference in AUC between the IS-CTO score and PROGRESS score = 0.397, p < 0.01). Our results suggest that the IS-CTO score system is a helpful tool to predict the technical success of IS-CTO PCI via antegrade approach in china. Graphical Abstract.

A radiomics model of predicting tumor volume change of patients with stage III non-small cell lung cancer after radiotherapy.

To predict the volume change of stage III NSCLC after radiotherapy with 60 Gy.This retrospective study included two independent cohorts, a train cohort of 192 patients, and a test cohort of 31 patients. We developed a radiomics model based on radiomics features and clinical variables. LIFEx package was used to extract radiomics texture features from CT images. The classification method was logistic regression analysis and feature selection was performed by correlation coefficients. Performance metrics of logistic regression include accuracy, precision, the receiver operating characteristic curves, and recall.The combination features of clinical variables and radiomics can predict the tumor volume change after radiotherapy with 88.7% accuracy (88.6% precision, 88.7% recall, and 88.7% ROC area).Radiomics features combined with medical knowledge have a great potential to predict accurately tumor volume change of stage III NSCLC after radiotherapy with 60 Gy.

Validity of a 2-item screening tool to identify families at risk for food insecurity in Brazil.

This manuscript aimed to develop a brief 2-item screening tool to identify Brazilian households that include families with children at risk for food insecurity. Psychometric analyses including sensitivity, specificity, positive and negative predictive value, accuracy, and ROC curves were used to test combinations of questions to determine the most effective screener to assess households at risk for food insecurity when compared to a gold standard scale. Participants included Brazilian National Survey of Demography and Health on Women and Children (PNDS) surveyed households with a valid Brazilian Food Insecurity Scale (EBIA) response. The sample included 3,920 households representing 11,779,686 households when expanded using PNDS sample weights. With overall prevalence of food insecurity at 21%, a Brazilian 2-item food-insecurity screen showed sensitivity of 79.31%, specificity of 92.95%, positive predictive value of 74.62%, negative predictive value of 94.5% and ROC area 86.13%. This screen also presented high convergent validity for children's nutrition and health variables when compared with the gold standard, the EBIA full scale. Based on its ability to detect households at risk for food insecurity, a 2-item screening tool is recommended for widespread adoption as a screening measure throughout Brazil, especially when rapid decision-making has been made fundamental, as under the COVID-19 pandemic. This screener can enable providers to accurately identify families at risk for food insecurity and promptly intervene to prevent or ameliorate adverse health and developmental consequences associated with food insecurity and swiftly respond to crises.

Comparative Analysis of Machine Learning based Filtering Techniques using MovieLens dataset

The primary objective of the present work is two-fold. First, to compare the different filtering techniques namely BayesNet, Decision Table, Logistic, k-NN, JRip, LibSVM, Randomized Filtered Classifier, Random Forest, Random Tree and OneR on the basis of six evaluation metrics including Kappa Statistic, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), F- Measure, ROC Area and Accuracy. Second, to improve the accuracy with the resampling mechanism. To determine the most appropriate algorithm, we have compared them using WEKA tool on a large sample of MovieLens dataset. The performance of ten different training algorithms has been analyzed and presented in this paper. It is found that random forest has received better results in comparison with other techniques with 99.39% accuracy. In addition to this, we have also incorporated the discussion of on-going researches and related works carried in the domain of recommendation systems. The present research shall help the researchers in exploring the best classification techniques for improving accuracy.

Exploring the Impact of Data Poisoning Attacks on Machine Learning Model Reliability

Recent years have seen the widespread adoption of Artificial Intelligence techniques in several domains, including healthcare, justice, assisted driving and Natural Language Processing (NLP) based applications (e.g., the Fake News detection). Those mentioned are just a few examples of some domains that are particularly critical and sensitive to the reliability of the adopted machine learning systems. Therefore, several Artificial Intelligence approaches were adopted as support to realize easy and reliable solutions aimed at improving the early diagnosis, personalized treatment, remote patient monitoring and better decision-making with a consequent reduction of healthcare costs. Recent studies have shown that these techniques are venerable to attacks by adversaries at phases of artificial intelligence. Poisoned data set are the most common attack to the reliability of Artificial Intelligence approaches. Noise, for example, can have a significant impact on the overall performance of a machine learning model. This study discusses the strength of impact of noise on classification algorithms. In detail, the reliability of several machine learning techniques to distinguish correctly pathological and healthy voices by analysing poisoning data was evaluated. Voice samples selected by available database, widely used in research sector, the Saarbruecken Voice Database, were processed and analysed to evaluate the resilience and classification accuracy of these techniques. All analyses are evaluated in terms of accuracy, specificity, sensitivity, F1-score and ROC area.

New Parameter of the Second Half of the P-Wave, P-Wave Duration, and Atrial Conduction Times Predict Atrial Fibrillation during Electrophysiological Studies

Objective: Several P-wave parameters reflect atrial conduction characteristics and have been used to predict atrial fibrillation (AF). The aim of this study was to determine the relationship between maximum P-wave duration (PMax) and new P-wave parameters, with atrial conduction times (CT), and to assess their predictive value of AF during electrophysiological studies (AF-EPS). Subjects and Methods: This was a cross-sectional study in 153 randomly selected patients aged 18–70 years, undergoing EPS. The patients were divided into 2 groups designated as no AF-EPS and AF-EPS, depending on whether AF occurred during EPS or not. Different P-wave parameters and atrial CT were compared for both study groups. Subsequently, the predictive value of the P-wave parameters and the atrial CT for AF-EPS was evaluated. Results: The values of CT, PMax, and maximum Ppeak-Pend interval (Pp-eMax) were significantly higher in patients with AF-EPS. Almost all P-wave parameters were correlated with the left CT. PMax, Pp-eMax, and CT were univariate and multivariate predictors of AF-EPS. The largest ROC area was presented by interatrial CT (0.852; p < 0.001; cutoff value: ≥82.5 ms; sensitivity: 91.1%; specificity: 81.1%). Pp-eMax showed greater sensitivity (79.5%) to discriminate AF-EPS than PMax (72.7%), but the latter had better specificity (60.4% vs. 41.5%). Conclusions: Left atrial CT were directly and significantly correlated with PMax and almost all the parameters of the second half of the P-wave. CT, PMax, and Pp-eMax (new parameter) were good predictors of AF-EPS, although CT did more robustly.

Analyzing Predictive Algorithms in Data Mining for Cardiovascular Disease using WEKA Tool

Cardiovascular Disease (CVD) is the foremost cause of death worldwide that generates a high percentage of Electronic Health Records (EHRs). Analyzing these complex patterns from EHRs is a tedious process. To address this problem, Medical Institutions requires effective Predictive Algorithms for the Prognosis and Diagnosis of the Patients. Under this work, the current state-of-the-art studied to identify leading Predictive Algorithms. Further, these algorithms namely Support Vector Machine (SVM), Naïve Bayes (NB), Decision Tree (DT), Random Forest (RF), Artificial Neural Network (ANN), Logistic Regression (LR), AdaBoost and k-Nearest Neighbors (k-NN) analyzed against the two datasets on open-source WEKA software. This work used two similar structured datasets i.e., Statlog Dataset and Cleveland Dataset. For Pre-Processing of Datasets, The missing values were replaced with the Mean value and later 10 Fold Cross-Validation was utilized for the evaluation. The result of the performance analysis showed that SVM outperforms other algorithms against both datasets. SVM showed an accuracy of 84.156% against the Cleveland dataset and 84.074% against the Statlog dataset. LR showed a ROC Area of 0.9 against both datasets. The findings of the work will help Health Institutions to understand the importance and usage of Predictive Algorithms for the automatic prediction of CVD based on the symptoms.

Identification of a Prognostic Risk Signature of Kidney Renal Clear Cell Carcinoma Based on Regulating the Immune Response Pathway Exploration

Purpose To construct a survival model for predicting the prognosis of patients with kidney renal clear cell carcinoma (KIRC) based on gene expression related to immune response regulation. Materials and Methods KIRC mRNA sequencing data and patient clinical data were downloaded from the TCGA database. The pathways and genes involved in the regulation of the immune response were identified from the GSEA database. A single factor Cox analysis was used to determine the association of mRNA in relation to patient prognosis (P < 0.05). The prognostic risk model was further established using the LASSO regression curve. The survival prognosis model was constructed, and the sensitivity and specificity of the model were evaluated using the ROC curve. Results Compared with normal kidney tissues, there were 28 dysregulated mRNA expressions in KIRC tissues (P < 0.05). Univariate Cox regression analysis revealed that 12 mRNAs were related to the prognosis of patients with renal cell carcinoma. The LASSO regression curve drew a risk signature consisting of six genes: TRAF6, FYN, IKBKG, LAT2, C2, IL4, EREG, TRAF2, and IL12A. The five-year ROC area analysis (AUC) showed that the model has good sensitivity and specificity (AUC >0.712). Conclusion We constructed a risk prediction model based on the regulated immune response-related genes, which can effectively predict the survival of patients with KIRC.

Proteomic Data Analysis for Differential Profiling of the Autoimmune Diseases SLE, RA, SS, and ANCA-Associated Vasculitis.

Early and correct diagnosis of inflammatory rheumatic diseases (IRD) poses a clinical challenge due to the multifaceted nature of symptoms, which also may change over time. The aim of this study was to perform protein expression profiling of four systemic IRDs, systemic lupus erythematosus (SLE), ANCA-associated systemic vasculitis (SV), rheumatoid arthritis (RA), and Sjögren’s syndrome (SS), and healthy controls to identify candidate biomarker signatures for differential classification. A total of 316 serum samples collected from patients with SLE, RA, SS, or SV and from healthy controls were analyzed using 394-plex recombinant antibody microarrays. Differential protein expression profiling was examined using Wilcoxon signed rank test, and condensed biomarker panels were identified using advanced bioinformatics and state-of-the art classification algorithms to pinpoint signatures reflecting each disease (raw data set available at https://figshare.com/s/3bd3848a28ef6e7ae9a9.). In this study, we were able to classify the included individual IRDs with high accuracy, as demonstrated by the ROC area under the curve (ROC AUC) values ranging between 0.96 and 0.80. In addition, the groups of IRDs could be separated from healthy controls at an ROC AUC value of 0.94. Disease-specific candidate biomarker signatures and general autoimmune signature were identified, including several deregulated analytes. This study supports the rationale of using multiplexed affinity-based technologies to reflect the biological complexity of autoimmune diseases. A multiplexed approach for decoding multifactorial complex diseases, such as autoimmune diseases, will play a significant role for future diagnostic purposes, essential to prevent severe organ- and tissue-related damage.