Random Forest Algorithm for the Classification of Neuroimaging Data in Alzheimer's Disease: A Systematic Review

Overlap between Central and Peripheral Transcriptomes in Parkinson’s Disease but Not Alzheimer’s Disease

Most neurodegenerative disorders take decades to develop, and their early detection is challenged by confounding non-pathological ageing processes. Therefore, the discovery of genes and molecular pathways in both peripheral and brain tissues that are highly predictive of disease evolution is necessary. To find genes that influence Alzheimer’s disease (AD) and Parkinson’s disease (PD) pathogenesis, human RNA-Seq transcriptomic data from Brodmann Area 9 (BA9) of the dorsolateral prefrontal cortex (DLPFC), whole blood (WB), and peripheral blood mononuclear cells (PBMC) were analysed using a combination of differential gene expression and a random forest-based machine learning algorithm. The results suggest that there is little overlap between PD and AD, and the AD brain signature is unique mainly compared to blood-based samples. Moreover, the AD-BA9 was characterised by changes in ‘nervous system development’ with Myocyte-specific enhancer factor 2C (Mef2C), encoding a transcription factor that induces microglia activation, a prominent feature. The peripheral AD transcriptome was associated with alterations in ‘viral process’, and FYN, which has been previously shown to link amyloid-beta and tau, was the prominent feature. However, in the absence of any overlap with the central transcriptome, it is unclear whether peripheral FYN levels reflect AD severity or progression. In PD, central and peripheral signatures are characterised by anomalies in ‘exocytosis’ and specific genes related to the SNARE complex, including Vesicle-associated membrane protein 2 (VAMP2), Syntaxin 1A (STX1A), and p21-activated kinase 1 (PAK1). This is consistent with our current understanding of the physiological role of alpha-synuclein and how alpha-synuclein oligomers compromise vesicle docking and neurotransmission. Overall, the results describe distinct disease-specific pathomechanisms, both within the brain and peripherally, for the two most common neurodegenerative disorders.

Prediction of Early Alzheimer Disease by Hippocampal Volume Changes under Machine Learning Algorithm

This research was aimed at discussing the application value of different machine learning algorithms in the prediction of early Alzheimer's disease (AD), which was based on hippocampal volume changes in magnetic resonance imaging (MRI). In the research, the 84 cases in American Alzheimer's disease neuroimaging initiative (ADNI) database were selected as the research data. Based on the scoring results of cognitive function, all cases were divided into three groups, including cognitive function normal (normal group), early mild cognitive impairment (e-MCI group), and later mild cognitive impairment (l-MCI group) groups. Each group included 28 cases. The features of hippocampal volume changes in MRI images of the patients in different groups were extracted. The samples of training set and test set were established. Besides, the established support vector machine (SVM), decision tree (DT), and random forest (RF) prediction models were used to predict e-MCI. Metalinear regression was utilized to analyze MRI feature data, and the predictive accuracy, sensitivity, and specificity of different models were calculated. The result showed that the volumes of hippocampal left CA1, left CA2-3, left CA4-DG, left presubiculum, left tail, right CA2-3, right CA4-DG, right presubiculum, and right tail in e-MCI group were all smaller than those in normal group (P < 0.01). The corresponding volume of hippocampal subregions in l-MCI group was remarkably reduced compared with that in normal group (P < 0.001). The volumes of regions left CA1, left CA2-3, left CA4-DG, right CA2-3, right CA4-DG, and right presubiculum were all positively correlated with logical memory test-delay recall (LMT-DR) score (R 2 = 0.1702, 0.3779, 0.1607, 0.1620, 0.0426, and 0.1309; P < 0.001). The predictive accuracy of training set sample by DT, SVM, and RF was 86.67%, 93.33%, and 98.33%, respectively. Based on the changes in the volumes of left CA4-DG, right CA2-3, and right CA4-DG, the predictive accuracy of e-MCI and l-MCI by RF model was both higher than those by DT model (P < 0.01). Besides, the predictive accuracy, sensitivity, and specificity of e-MCI by RF model was all notably higher than those by DT model (P < 0.01). The above results demonstrated that the effective early AD prediction models were established by the volume changes in hippocampal subregions, which was based on RF in the research. The establishment of early AD prediction models offered certain reference basis to the diagnosis and treatment of AD patients.

Implementation of a Heart Disease Risk Prediction Model Using Machine Learning

Cardiovascular disease prediction aids practitioners in making more accurate health decisions for their patients. Early detection can aid people in making lifestyle changes and, if necessary, ensuring effective medical care. Machine learning (ML) is a plausible option for reducing and understanding heart symptoms of disease. The chi-square statistical test is performed to select specific attributes from the Cleveland heart disease (HD) dataset. Support vector machine (SVM), Gaussian Naive Bayes, logistic regression, LightGBM, XGBoost, and random forest algorithm have been employed for developing heart disease risk prediction model and obtained the accuracy as 80.32%, 78.68%, 80.32%, 77.04%, 73.77%, and 88.5%, respectively. The data visualization has been generated to illustrate the relationship between the features. According to the findings of the experiments, the random forest algorithm achieves 88.5% accuracy during validation for 303 data instances with 13 selected features of the Cleveland HD dataset.

Artificial intelligence in brain MRI analysis of Alzheimer's disease over the past 12 years: A systematic review

INTRODUCTION

Multiple structural brain changes in Alzheimer's disease (AD) and mild cognitive impairment (MCI) have been revealed on magnetic resonance imaging (MRI). There is a fast-growing effort in applying artificial intelligence (AI) to analyze these data. Here, we review and evaluate the AI studies in brain MRI analysis with synthesis.

METHODS

A systematic review of the literature, spanning the years from 2009 to 2020, was completed using the PubMed database. AI studies using MRI imaging to investigate normal aging, mild cognitive impairment, and AD-dementia were retrieved for review. Bias assessment was completed using the PROBAST criteria.

RESULTS

97 relevant studies were included in the review. The studies were typically focused on the classification of AD, MCI, and normal aging (71% of the reported studies) and the prediction of MCI conversion to AD (25%). The best performance was achieved by using the deep learning-based convolution neural network algorithms (weighted average accuracy 89%), in contrast to 76-86% using Logistic Regression, Support Vector Machines, and other AI methods.

DISCUSSION

The synthesized evidence is paramount to developing sophisticated AI approaches to reliably capture and quantify multiple subtle MRI changes in the whole brain that exemplify the complexity and heterogeneity of AD and brain aging.

Artificial intelligence approaches to the biochemistry of oxidative stress: Current state of the art

Artificial intelligence (AI) and machine learning models are today frequently used for classification and prediction of various biochemical processes and phenomena. In recent years, numerous research efforts have been focused on developing such models for assessment, categorization, and prediction of oxidative stress. Supervised machine learning can successfully automate the process of evaluation and quantification of oxidative damage in biological samples, as well as extract useful data from the abundance of experimental results. In this concise review, we cover the possible applications of neural networks, decision trees and regression analysis as three common strategies in machine learning. We also review recent works on the various weaknesses and limitations of artificial intelligence in biochemistry and related scientific areas. Finally, we discuss future innovative approaches on the ways how AI can contribute to the automation of oxidative stress measurement and diagnosis of diseases associated with oxidative damage.

CT-based radiomics analysis of different machine learning models for differentiating benign and malignant parotid tumors

OBJECTIVES

This study aimed to explore and validate the value of different radiomics models for differentiating benign and malignant parotid tumors preoperatively.

METHODS

This study enrolled 388 patients with pathologically confirmed parotid tumors (training cohort: n = 272; test cohort: n = 116). Radiomics features were extracted from CT images of the non-enhanced, arterial, and venous phases. After dimensionality reduction and selection, radiomics models were constructed by logistic regression (LR), support vector machine (SVM), and random forest (RF). The best radiomic model was selected by using ROC curve analysis. Univariate and multivariable logistic regression was applied to analyze clinical-radiological characteristics and identify variables for developing a clinical model. A combined model was constructed by incorporating radiomics and clinical features. Model performances were assessed by ROC curve analysis, and decision curve analysis (DCA) was used to estimate the models' clinical values.

RESULTS

In total, 2874 radiomic features were extracted from CT images. Ten radiomics features were deemed valuable by dimensionality reduction and selection. Among radiomics models, the SVM model showed greater predictive efficiency and robustness, with AUCs of 0.844 in the training cohort; and 0.840 in the test cohort. Ultimate clinical features constructed a clinical model. The discriminatory capability of the combined model was the best (AUC, training cohort: 0.904; test cohort: 0.854). Combined model DCA revealed optimal clinical efficacy.

CONCLUSIONS

The combined model incorporating radiomics and clinical features exhibited excellent ability to distinguish benign and malignant parotid tumors, which may provide a noninvasive and efficient method for clinical decision making.

KEY POINTS

The current study is the first to compare the value of different radiomics models (LR, SVM, and RF) for preoperative differentiation of benign and malignant parotid tumors. A CT-based combined model, integrating clinical-radiological and radiomics features, is conducive to distinguishing benign and malignant parotid tumors, thereby improving diagnostic performance and aiding treatment.

PlasmidHostFinder: Prediction of Plasmid Hosts Using Random Forest

Plasmids play a major role facilitating the spread of antimicrobial resistance between bacteria. Understanding the host range and dissemination trajectories of plasmids is critical for surveillance and prevention of antimicrobial resistance. Identification of plasmid host ranges could be improved using automated pattern detection methods compared to homology-based methods due to the diversity and genetic plasticity of plasmids. In this study, we developed a method for predicting the host range of plasmids using machine learning-specifically, random forests. We trained the models with 8,519 plasmids from 359 different bacterial species per taxonomic level; the models achieved Matthews correlation coefficients of 0.662 and 0.867 at the species and order levels, respectively. Our results suggest that despite the diverse nature and genetic plasticity of plasmids, our random forest model can accurately distinguish between plasmid hosts. This tool is available online through the Center for Genomic Epidemiology (https://cge.cbs.dtu.dk/services/PlasmidHostFinder/). IMPORTANCE Antimicrobial resistance is a global health threat to humans and animals, causing high mortality and morbidity while effectively ending decades of success in fighting against bacterial infections. Plasmids confer extra genetic capabilities to the host organisms through accessory genes that can encode antimicrobial resistance and virulence. In addition to lateral inheritance, plasmids can be transferred horizontally between bacterial taxa. Therefore, detection of the host range of plasmids is crucial for understanding and predicting the dissemination trajectories of extrachromosomal genes and bacterial evolution as well as taking effective countermeasures against antimicrobial resistance.

Lipid level alteration in human and cellular models of alpha synuclein mutations

Lipid profiles in biological fluids from patients with Parkinson's disease (PD) are increasingly investigated in search of biomarkers. However, the lipid profiles in genetic PD remain to be determined, a gap of knowledge of particular interest in PD associated with mutant α-synuclein (SNCA), given the known relationship between this protein and lipids. The objective of this research is to identify serum lipid composition from SNCA A53T mutation carriers and to compare these alterations to those found in cells and transgenic mice carrying the same genetic mutation. We conducted an unbiased lipidomic analysis of 530 lipid species from 34 lipid classes in serum of 30 participants with SNCA mutation with and without PD and 30 healthy controls. The primary analysis was done between 22 PD patients with SNCA+ (SNCA+/PD+) and 30 controls using machine-learning algorithms and traditional statistics. We also analyzed the lipid composition of human clonal-cell lines and tissue from transgenic mice overexpressing the same SNCA mutation. We identified specific lipid classes that best discriminate between SNCA+/PD+ patients and healthy controls and found certain lipid species, mainly from the glycerophosphatidylcholine and triradylglycerol classes, that are most contributory to this discrimination. Most of these alterations were also present in human derived cells and transgenic mice carrying the same mutation. Our combination of lipidomic and machine learning analyses revealed alterations in glycerophosphatidylcholine and triradylglycerol in sera from PD patients as well as cells and tissues expressing mutant α-Syn. Further investigations are needed to establish the pathogenic significance of these α-Syn-associated lipid changes.

Feature aggregation graph convolutional network based on imaging genetic data for diagnosis and pathogeny identification of Alzheimer's disease

The roles of brain regions activities and gene expressions in the development of Alzheimer's disease (AD) remain unclear. Existing imaging genetic studies usually has the problem of inefficiency and inadequate fusion of data. This study proposes a novel deep learning method to efficiently capture the development pattern of AD. First, we model the interaction between brain regions and genes as node-to-node feature aggregation in a brain region-gene network. Second, we propose a feature aggregation graph convolutional network (FAGCN) to transmit and update the node feature. Compared with the trivial graph convolutional procedure, we replace the input from the adjacency matrix with a weight matrix based on correlation analysis and consider common neighbor similarity to discover broader associations of nodes. Finally, we use a full-gradient saliency graph mechanism to score and extract the pathogenetic brain regions and risk genes. According to the results, FAGCN achieved the best performance among both traditional and cutting-edge methods and extracted AD-related brain regions and genes, providing theoretical and methodological support for the research of related diseases.

Investigation of serum markers of esophageal squamous cell carcinoma based on machine learning methods

Esophageal squamous cell carcinoma (ESCC) is one of the malignant tumors with high mortality in humans, and there is a lack of effective and convenient early diagnosis methods. By analyzing the serum miRNA expression data in ESCC tumor samples and normal samples, on the basis of the max-relevance and min-redundancy (mRMR) feature selection and the incremental feature selection (IFS) method, a random forest classifier constructed by 5 feature miRNAs was acquired in our study. The receiver operator characteristic (ROC) curve showed that the model was able to distinguish samples. Principal component analysis (PCA) and sample hierarchical cluster analysis showed that 5 feature miRNAs could well distinguish ESCC patients from healthy individuals. The expression levels of miR-663a, miR-5100 and miR-221-3p all showed a higher expression level in ESCC patients than those in healthy individuals. On the contrary, miR-6763-5p and miR-7111-5p both showed lower expression levels in ESCC patients than those in healthy individuals. In addition, the collected clinical serum samples were used for qRT-PCR analysis. It was uncovered that the expression trends of the 5 feature miRNAs followed a similar pattern with those in the training set. The above findings indicated that the 5 feature miRNAs may be serum tumor markers of ESCC. This study offers new insights for the early diagnosis of ESCC.

Predicting surgical decision-making in vestibular schwannoma using tree-based machine learning

OBJECTIVE

Vestibular schwannomas (VSs) are the most common neoplasm of the cerebellopontine angle in adults. Though these lesions are generally slow growing, their growth patterns and associated symptoms can be unpredictable, which may complicate the decision to pursue conservative management versus active intervention. Additionally, surgical decision-making can be controversial because of limited high-quality evidence and multiple quality-of-life considerations. Machine learning (ML) is a powerful tool that utilizes data sets to essentialize multidimensional clinical processes. In this study, the authors trained multiple tree-based ML algorithms to predict the decision for active treatment versus MRI surveillance of VS in a single institutional cohort. In doing so, they sought to assess which preoperative variables carried the most weight in driving the decision for intervention and could be used to guide future surgical decision-making through an evidence-based approach.

METHODS

The authors reviewed the records of patients who had undergone evaluation by neurosurgery and otolaryngology with subsequent active treatment (resection or radiation) for unilateral VS in the period from 2009 to 2021, as well as those of patients who had been evaluated for VS and were managed conservatively throughout 2021. Clinical presentation, radiographic data, and management plans were abstracted from each patient record from the time of first evaluation until the last follow-up or surgery. Each encounter with the patient was treated as an instance involving a management decision that depended on demographics, symptoms, and tumor profile. Decision tree and random forest classifiers were trained and tested to predict the decision for treatment versus imaging surveillance on the basis of unseen data using an 80/20 pseudorandom split. Predictor variables were tuned to maximize performance based on lowest Gini impurity indices. Model performance was optimized using fivefold cross-validation.

RESULTS

One hundred twenty-four patients with 198 rendered decisions concerning management were included in the study. In the decision tree analysis, only a maximum tumor dimension threshold of 1.6 cm and progressive symptoms were required to predict the decision for treatment with 85% accuracy. Optimizing maximum dimension thresholds and including age at presentation boosted accuracy to 88%. Random forest analysis (n = 500 trees) predicted the decision for treatment with 80% accuracy. Factors with the highest variable importance based on multiple measures of importance, including mean minimal conditional depth and largest Gini impurity reduction, were maximum tumor dimension, age at presentation, Koos grade, and progressive symptoms at presentation.

CONCLUSIONS

Tree-based ML was used to predict which factors drive the decision for active treatment of VS with 80%–88% accuracy. The most important factors were maximum tumor dimension, age at presentation, Koos grade, and progressive symptoms. These results can assist in surgical decision-making and patient counseling. They also demonstrate the power of ML algorithms in extracting useful insights from limited data sets.

Identification of Radiomic Features as an Imaging Marker to Differentiate Benign and Malignant Breast Masses Based on Magnetic Resonance Imaging

Background - Breast cancer is one of the most common cancers among women globally and early identification is known to increase patient outcomes. Therefore, the main aim of this study is to identify the essential radiomic features as an image marker and compare the diagnostic feasibility of feature parameters derived from radiomics analysis and conventional Magnetic Resonance Imaging (MRI) to differentiate benign and malignant breast masses.

Methods and Material - T1-weighted Dynamic Contrast-Enhanced (DCE) breast MR axial images of 151 (benign (79) and malignant (72)) patients were chosen. Regions of interest were selected using both manual and semi-automatic segmentation from each lesion. 382 radiomic features computed on the selected regions. A random forest model was employed to detect the most important features that differentiate benign and malignant breast masses. The ten most important radiomics features were obtained from manual and semi-automatic segmentation based on the Gini index to train a support vector machine. MATLAB and IBM SPSS Statistics Subscription software used for statistical analysis.

Results - The accuracy (sensitivity) of the models built from the ten most significant features obtained from manual and semi-automatic segmentation were 0.815 (0.84), 0.821 (0.87), respectively. The top 10 features obtained from manual delineation and semi-automatic segmentation showed a significant difference (p<0.05) between benign and malignant breast lesions.

Conclusion - This radiomics analysis based on DCE-BMRI revealed distinct radiomic features to differentiate benign and malignant breast masses. Therefore, the radiomics analysis can be used as a supporting tool in detecting breast MRI lesions.

Persistent Overactive Cytotoxic Immune Response in a Spanish Cohort of Individuals With Long-COVID: Identification of Diagnostic Biomarkers

Long-COVID is a new emerging syndrome worldwide that is characterized by the persistence of unresolved signs and symptoms of COVID-19 more than 4 weeks after the infection and even after more than 12 weeks. The underlying mechanisms for Long-COVID are still undefined, but a sustained inflammatory response caused by the persistence of SARS-CoV-2 in organ and tissue sanctuaries or resemblance with an autoimmune disease are within the most considered hypotheses. In this study, we analyzed the usefulness of several demographic, clinical, and immunological parameters as diagnostic biomarkers of Long-COVID in one cohort of Spanish individuals who presented signs and symptoms of this syndrome after 49 weeks post-infection, in comparison with individuals who recovered completely in the first 12 weeks after the infection. We determined that individuals with Long-COVID showed significantly increased levels of functional memory cells with high antiviral cytotoxic activity such as CD8+ TEMRA cells, CD8±TCRγδ+ cells, and NK cells with CD56+CD57+NKG2C+ phenotype. The persistence of these long-lasting cytotoxic populations was supported by enhanced levels of CD4+ Tregs and the expression of the exhaustion marker PD-1 on the surface of CD3+ T lymphocytes. With the use of these immune parameters and significant clinical features such as lethargy, pleuritic chest pain, and dermatological injuries, as well as demographic factors such as female gender and O+ blood type, a Random Forest algorithm predicted the assignment of the participants in the Long-COVID group with 100% accuracy. The definition of the most accurate diagnostic biomarkers could be helpful to detect the development of Long-COVID and to improve the clinical management of these patients.

Editorial for the Special Issue on “Machine Learning in Healthcare and Biomedical Application”

In the last decade, Machine Learning (ML) has indisputably had a pervasive application in healthcare and biomedical applications [...]

Vaccine hesitancy in the post-vaccination COVID-19 era: a machine learning and statistical analysis driven study

Background The COVID-19 pandemic has badly affected people of all ages globally. Therefore, its vaccine has been developed and made available for public use in unprecedented times. However, because of various levels of hesitancy, it did not have general acceptance. The main objective of this work is to identify the risk associated with the COVID-19 vaccines by developing a prognosis tool that will help in enhancing its acceptability and therefore, reducing the lethality of SARS-CoV-2. Methods: The obtained raw VAERS dataset has three files indicating medical history, vaccination status, and post vaccination symptoms respectively with more than 354 thousand samples. After pre-processing, this raw dataset has been merged into one with 85 different attributes however, the whole analysis has been subdivided into three scenarios ((i) medical history (ii) reaction of vaccination (iii) combination of both). Further, Machine Learning (ML) models which includes Linear Regression (LR), Random Forest (RF), Naive Bayes (NB), Light Gradient Boosting Algorithm (LGBM), and Multilayer feed-forward perceptron (MLP) have been employed to predict the most probable outcome and their performance has been evaluated based on various performance parameters. Also, the chi-square (statistical), LR, RF, and LGBM have been utilized to estimate the most probable attribute in the dataset that resulted in death, hospitalization, and COVID-19. Results: For the above mentioned scenarios, all the models estimates different attributes (such as cardiac arrest, Cancer, Hyperlipidemia, Kidney Disease, Diabetes, Atrial Fibrillation, Dementia, Thyroid, etc.) for death, hospitalization, and COVID-19 even after vaccination. Further, for prediction, LGBM outperforms all the other developed models in most of the scenarios whereas, LR, RF, NB, and MLP perform satisfactorily in patches. Conclusion: The male population in the age group of 50-70 has been found most susceptible to this virus. Also, people with existing serious illnesses have been found most vulnerable. Therefore, they must be vaccinated in close observations. Generally, no serious adverse effect of the vaccine has been observed therefore, people must vaccinate themselves without any hesitation at the earliest. Also, the model developed using LGBM establishes its supremacy over all the other prediction models. Therefore, it can be very helpful for the policymakers in administrating and prioritizing the population for the different vaccination programs.

Prediction of Atrial Fibrillation in Hospitalized Elderly Patients With Coronary Heart Disease and Type 2 Diabetes Mellitus Using Machine Learning: A Multicenter Retrospective Study

Background

The objective of this study was to use machine learning algorithms to construct predictive models for atrial fibrillation (AF) in elderly patients with coronary heart disease (CHD) and type 2 diabetes mellitus (T2DM).

Methods

The diagnosis and treatment data of elderly patients with CHD and T2DM, who were treated in four tertiary hospitals in Chongqing, China from 2015 to 2021, were collected. Five machine learning algorithms: logistic regression, logistic regression+least absolute shrinkage and selection operator, classified regression tree (CART), random forest (RF) and extreme gradient lifting (XGBoost) were used to construct the prediction models. The area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and accuracy were used as the comparison measures between different models.

Results

A total of 3,858 elderly patients with CHD and T2DM were included. In the internal validation cohort, XGBoost had the highest AUC (0.743) and sensitivity (0.833), and RF had the highest specificity (0.753) and accuracy (0.735). In the external verification, RF had the highest AUC (0.726) and sensitivity (0.686), and CART had the highest specificity (0.925) and accuracy (0.841). Total bilirubin, triglycerides and uric acid were the three most important predictors of AF.

Conclusion

The risk prediction models of AF in elderly patients with CHD and T2DM based on machine learning algorithms had high diagnostic value. The prediction models constructed by RF and XGBoost were more effective. The results of this study can provide reference for the clinical prevention and treatment of AF.

A systematic analysis of diagnostic performance for Alzheimer's disease using structural MRI

Background

Alzheimer's disease (AD) is one of the most common neurodegenerative disorders in the elderly. Although numerous structural magnetic resonance imaging (sMRI) studies have reported diagnostic models that could distinguish AD from normal controls (NCs) with 80-95% accuracy, limited efforts have been made regarding the clinically practical computer-aided diagnosis (CAD) system for AD.

Objective

To explore the potential factors that hinder the clinical translation of the AD-related diagnostic models based on sMRI.

Methods

To systematically review the diagnostic models for AD based on sMRI, we identified relevant studies published in the past 15 years on PubMed, Web of Science, Scopus, and Ovid. To evaluate the heterogeneity and publication bias among those studies, we performed subgroup analysis, meta-regression, Begg's test, and Egger's test.

Results

According to our screening criterion, 101 studies were included. Our results demonstrated that high diagnostic accuracy for distinguishing AD from NC was obtained in recently published studies, accompanied by significant heterogeneity. Meta-analysis showed that many factors contributed to the heterogeneity of high diagnostic accuracy of AD using sMRI, which included but was not limited to the following aspects: (i) different datasets; (ii) different machine learning models, e.g. traditional machine learning or deep learning model; (iii) different cross-validation methods, e.g. k-fold cross-validation leads to higher accuracies than leave-one-out cross-validation, but both overestimate the accuracy when compared to validation in independent samples; (iv) different sample sizes; and (v) the publication times. We speculate that these complicated variables might be the adverse factor for developing a clinically applicable system for the early diagnosis of AD.

Conclusions

Our findings proved that previous studies reported promising results for classifying AD from NC with different models using sMRI. However, considering the many factors hindering clinical radiology practice, there would still be a long way to go to improve.

Predicting Motor Responsiveness to Deep Brain Stimulation with Machine Learning

Deep brain stimulation is a complex movement disorder intervention that requires highly invasive brain surgery. Clinicians struggle to predict how patients will respond to this treatment. To address this problem, we are working toward developing a clinical tool to help neurologists predict deep brain stimulation response. We analyzed a cohort of 105 Parkinson's patients who underwent deep brain stimulation at Vanderbilt University Medical Center. We developed binary and multicategory models for predicting likelihood of motor symptom reduction after undergoing deep brain stimulation. We compared the performances of our best models to predictions made by neurologist experts in movement disorders. The strongest binary classification model achieved a 10-fold cross validation AUC of 0.90, outperforming the best neurologist predictions (0.56). These results are promising for future clinical applications, though more work is necessary to validate these findings in a larger cohort and taking into consideration broader quality of life outcome measures.

Development of machine learning models for mortality risk prediction after cardiac surgery

BACKGROUND

We developed machine learning models that combine preoperative and intraoperative risk factors to predict mortality after cardiac surgery.

METHODS

Machine learning involving random forest, neural network, support vector machine, and gradient boosting machine was developed and compared with the risk scores of EuroSCORE I and II, Society of Thoracic Surgeons (STS), as well as a logistic regression model. Clinical data were collected from patients undergoing adult cardiac surgery at the First Medical Centre of Chinese PLA General Hospital between December 2008 and December 2017. The primary outcome was post-operative mortality. Model performance was estimated using several metrics, including sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC). The visualization algorithm was implemented using Shapley's additive explanations.

RESULTS

A total of 5,443 patients were enrolled during the study period. The mean EuroSCORE II score was 3.7%, and the actual in-hospital mortality rate was 2.7%. For predicting operative mortality after cardiac surgery, the AUC scores were 0.87, 0.79, 0.81, and 0.82 for random forest, neural network, support vector machine, and gradient boosting machine, compared with 0.70, 0.73, 0.71, and 0.74 for EuroSCORE I and II, STS, and logistic regression model. Shapley's additive explanations analysis of random forest yielded the top-20 predictors and individual-level explanations for each prediction.

CONCLUSIONS

Machine learning models based on available clinical data may be superior to clinical scoring tools in predicting postoperative mortality in patients following cardiac surgery. Explanatory models show the potential to provide personalized risk profiles for individuals by accounting for the contribution of influencing factors. Additional prospective multicenter studies are warranted to confirm the clinical benefit of these machine learning-driven models.

Machine Learning-Based Assessment of Cognitive Impairment Using Time-Resolved Near-Infrared Spectroscopy and Basic Blood Test

We have demonstrated that machine learning allows us to predict cognitive function in aged people using near-infrared spectroscopy (NIRS) data or basic blood test data. However, the following points are not yet clear: first, whether there are differences in prediction accuracy between NIRS and blood test data; second, whether there are differences in prediction accuracy for cognitive function in linear models and non-linear models; and third, whether there are changes in prediction accuracy when both NIRS and blood test data are added to the input layer. We used a linear regression model (LR) for the linear model and random forest (RF) and deep neural network (DNN) for the non-linear model. We studied 250 participants (mean age = 73.3 ± 12.6 years) and assessed cognitive function using the Mini Mental State Examination (MMSE) (mean MMSE scores = 22.9 ± 6.1). We used time-resolved NIRS (TNIRS) to measure absolute concentrations of hemoglobin and optical pathlength at rest in the bilateral prefrontal cortices. A basic blood test was performed on the same day. We compared predicted MMSE scores and grand truth MMSE scores; prediction accuracies were evaluated using mean absolute error (MAE) and mean absolute percentage error (MAPE). We found that (1) the DNN-based prediction using TNIRS data exhibited lower MAE and MAPE compared with those using blood test data, (2) the difference in MAPE between TNIRS and blood test data was only 0.3%, (3) adding TNIRS data to the blood test data of the input layer only improved MAPE by 1.0% compared to the use of blood test data alone, whereas the use of the blood test data alone exhibited the prediction accuracy with 81.8% sensitivity and 91.3% specificity (N = 202, repeated five-fold cross validation). Given these findings and the benefits of using blood test data (low cost and large-scale screening possible), we concluded that the DNN model using blood test data is still the most suitable for mass screening.

An Exploration: Alzheimer's Disease Classification Based on Convolutional Neural Network

Alzheimer's disease (AD) is the most generally known neurodegenerative disorder, leading to a steady deterioration in cognitive ability. Deep learning models have shown outstanding performance in the diagnosis of AD, and these models do not need any handcrafted feature extraction over conventional machine learning algorithms. Since the 2012 AlexNet accomplishment, the convolutional neural network (CNN) has been progressively utilized by the medical community to assist practitioners to early diagnose AD. This paper explores the current cutting edge applications of CNN on single and multimodality (combination of two or more modalities) neuroimaging data for the classification of AD. An exhaustive systematic search is conducted on four notable databases: Google Scholar, IEEE Xplore, ACM Digital Library, and PubMed in June 2021. The objective of this study is to examine the effectiveness of classification approaches on AD to analyze different kinds of datasets, neuroimaging modalities, preprocessing techniques, and data handling methods. However, CNN has achieved great success in the classification of AD; still, there are a lot of challenges particularly due to scarcity of medical imaging data and its possible scope in this field.

Evaluating Pointwise Reliability of Machine Learning prediction

Interest in Machine Learning applications to tackle clinical and biological problems is increasing. This is driven by promising results reported in many research papers, the increasing number of AI-based software products, and by the general interest in Artificial Intelligence to solve complex problems. It is therefore of importance to improve the quality of machine learning output and add safeguards to support their adoption. In addition to regulatory and logistical strategies, a crucial aspect is to detect when a Machine Learning model is not able to generalize to new unseen instances, which may originate from a population distant to that of the training population or from an under-represented subpopulation. As a result, the prediction of the machine learning model for these instances may be often wrong, given that the model is applied outside its "reliable" space of work, leading to a decreasing trust of the final users, such as clinicians. For this reason, when a model is deployed in practice, it would be important to advise users when the model's predictions may be unreliable, especially in high-stakes applications, including those in healthcare. Yet, reliability assessment of each machine learning prediction is still poorly addressed. Here, we review approaches that can support the identification of unreliable predictions, we harmonize the notation and terminology of relevant concepts, and we highlight and extend possible interrelationships and overlap among concepts. We then demonstrate, on simulated and real data for ICU in-hospital death prediction, a possible integrative framework for the identification of reliable and unreliable predictions. To do so, our proposed approach implements two complementary principles, namely the density principle and the local fit principle. The density principle verifies that the instance we want to evaluate is similar to the training set. The local fit principle verifies that the trained model performs well on training subsets that are more similar to the instance under evaluation. Our work can contribute to consolidating work in machine learning especially in medicine.

Developing a Predictive Model for Depressive Disorders Using Stacking Ensemble and Naive Bayesian Nomogram: Using Samples Representing South Korea

This study provided baseline data for preventing depression in female older adults living alone by understanding the degree of their depressive disorders and factors affecting these depressive disorders by analyzing epidemiological survey data representing South Koreans. To achieve the study objective, this study explored the main risk factors of depressive disorders using the stacking ensemble machine technique. Moreover, this study developed a nomogram that could help primary physicians easily interpret high-risk groups of depressive disorders in primary care settings based on the major predictors derived from machine learning. This study analyzed 582 female older adults (≥60 years old) living alone. The depressive disorder, a target variable, was measured using the Korean version of Patient Health Questionnaire-9. This study developed five single predictive models (GBM, Random Forest, Adaboost, SVM, XGBoost) and six stacking ensemble models (GBM + Bayesian regression, RandomForest + Bayesian regression, Adaboost + Bayesian regression, SVM + Bayesian regression, XGBoost + Bayesian regression, GBM + RandomForest + Adaboost + SVM + XGBoost + Bayesian regression) to predict depressive disorders. The naive Bayesian nomogram confirmed that stress perception, subjective health, n-6 fatty acid, n-3 fatty acid, mean hours of sitting per day, and mean daily sleep hours were six major variables related to the depressive disorders of female older adults living alone. Based on the results of this study, it is required to evaluate the multiple risk factors for depression including various measurable factors such as social support.

Random Forest Model in the Diagnosis of Dementia Patients with Normal Mini-Mental State Examination Scores

Background: Mini-Mental State Examination (MMSE) is the most widely used tool in cognitive screening. Some individuals with normal MMSE scores have extensive cognitive impairment. Systematic neuropsychological assessment should be performed in these patients. This study aimed to optimize the systematic neuropsychological test battery (NTB) by machine learning and develop new classification models for distinguishing mild cognitive impairment (MCI) and dementia among individuals with MMSE ≥ 26. Methods: 375 participants with MMSE ≥ 26 were assigned a diagnosis of cognitively unimpaired (CU) (n = 67), MCI (n = 174), or dementia (n = 134). We compared the performance of five machine learning algorithms, including logistic regression, decision tree, SVM, XGBoost, and random forest (RF), in identifying MCI and dementia. Results: RF performed best in identifying MCI and dementia. Six neuropsychological subtests with high-importance features were selected to form a simplified NTB, and the test time was cut in half. The AUC of the RF model was 0.89 for distinguishing MCI from CU, and 0.84 for distinguishing dementia from nondementia. Conclusions: This simplified cognitive assessment model can be useful for the diagnosis of MCI and dementia in patients with normal MMSE. It not only optimizes the content of cognitive evaluation, but also improves diagnosis and reduces missed diagnosis.

Detecting Cognitive Impairment Status Using Keystroke Patterns and Physical Activity Data among the Older Adults: A Machine Learning Approach

Cognitive impairment has a significantly negative impact on global healthcare and the community. Holding a person's cognition and mental retention among older adults is improbable with aging. Early detection of cognitive impairment will decline the most significant impact of extended disease to permanent mental damage. This paper aims to develop a machine learning model to detect and differentiate cognitive impairment categories like severe, moderate, mild, and normal by analyzing neurophysical and physical data. Keystroke and smartwatch have been used to extract individuals' neurophysical and physical data, respectively. An advanced ensemble learning algorithm named Gradient Boosting Machine (GBM) is proposed to classify the cognitive severity level (absence, mild, moderate, and severe) based on the Standardised Mini-Mental State Examination (SMMSE) questionnaire scores. The statistical method "Pearson's correlation" and the wrapper feature selection technique have been used to analyze and select the best features. Then, we have conducted our proposed algorithm GBM on those features. And the result has shown an accuracy of more than 94%. This paper has added a new dimension to the state-of-the-art to predict cognitive impairment by implementing neurophysical data and physical data together.

A 'Mini Linguistic State Examination' to classify primary progressive aphasia

There are few available methods for qualitatively evaluating patients with primary progressive aphasia. Commonly adopted approaches are time-consuming, of limited accuracy or designed to assess different patient populations. This paper introduces a new clinical test-the Mini Linguistic State Examination-which was designed uniquely to enable a clinician to assess and subclassify both classical and mixed presentations of primary progressive aphasia. The adoption of a novel assessment method (error classification) greatly amplifies the clinical information that can be derived from a set of standard linguistic tasks and allows a five-dimensional profile to be defined. Fifty-four patients and 30 matched controls were recruited. Five domains of language competence (motor speech, phonology, semantics, syntax and working memory) were assessed using a sequence of 11 distinct linguistic assays. A random forest classification was used to assess the diagnostic accuracy for predicting primary progressive aphasia subtypes and create a decision tree as a guide to clinical classification. The random forest prediction model was 96% accurate overall (92% for the logopenic variant, 93% for the semantic variant and 98% for the non-fluent variant). The derived decision tree produced a correct classification of 91% of participants whose data were not included in the training set. The Mini Linguistic State Examination is a new cognitive test incorporating a novel and powerful, yet straightforward, approach to scoring. Rigorous assessment of its diagnostic accuracy confirmed excellent matching of primary progressive aphasia syndromes to clinical gold standard diagnoses. Adoption of the Mini Linguistic State Examination by clinicians will have a decisive impact on the consistency and uniformity with which patients can be described clinically. It will also facilitate screening for cohort-based research, including future therapeutic trials, and is suitable for describing, quantifying and monitoring language deficits in other brain disorders.

Development and Validation of a Random Forest Risk Prediction Pneumothorax Model in Percutaneous Transthoracic Needle Biopsy

BACKGROUND Computed tomography (CT)-guided percutaneous transthoracic needle biopsy (PTNB) is an effective means for diagnosing various thoracic diseases. Pneumothorax is the most common complication, and when it becomes life-threatening, urgent medical intervention is required. The purpose of this study was to develop and validate a model that can be used to predict postoperative pneumothorax following CT-guided PTNB. MATERIAL AND METHODS We enrolled 245 patients who completed CT-guided PTNB to develop the model. A random forest (RF) model was built using 15 risk factors (15-RFs). The 7 most critical risk factors (7-RFs) were extracted by feature selection and used to build a new model. The independent external validation data contained 97 patients. Logistic regression (LR), support vector machine (SVM), and decision tree (DT) models were also developed using both 15-RFs and 7-RFs, and their performance was compared with the RF models. RESULTS The length of the aerated lung traversed was identified as the most important risk factor for developing pneumothorax, followed by angle of pleural puncture, lesion depth, lesion size, age, procedure time, and sex. The RF model demonstrated better performance in the development and validation datasets when compared with the LR, SVM, and DT based on 15-RFs and 7-RFs. According to DeLong's test for difference in ROC curves, the RF models based on the 15-RFs and 7-RFs achieved similar classification performance (P>0.05). CONCLUSIONS This study demonstrated the feasibility of using the 7-RFs RF model for predicting postoperative pneumothorax before patients undergo CT-guided PTNB.

Machine learning trained with quantitative susceptibility mapping to detect mild cognitive impairment in Parkinson's disease

BACKGROUND

Cognitive decline is commonly observed in Parkinson's disease (PD). Identifying PD with mild cognitive impairment (PD-MCI) is crucial for early initiation of therapeutic interventions and preventing cognitive decline.

OBJECTIVE

We aimed to develop a machine learning model trained with magnetic susceptibility values based on the multi-atlas label-fusion method to classify PD without dementia into PD-MCI and normal cognition (PD-CN).

METHODS

This multicenter observational cohort study retrospectively reviewed 61 PD-MCI and 59 PD-CN cases for the internal validation cohort and 22 PD-MCI and 21 PD-CN cases for the external validation cohort. The multi-atlas method parcellated the quantitative susceptibility mapping (QSM) images into 20 regions of interest and extracted QSM-based magnetic susceptibility values. Random forest, extreme gradient boosting, and light gradient boosting were selected as machine learning algorithms.

RESULTS

All classifiers demonstrated substantial performances in the classification task, particularly the random forest model. The accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve for this model were 79.1%, 77.3%, 81.0%, and 0.78, respectively. The QSM values in the caudate nucleus, which were important features, were inversely correlated with the Montreal Cognitive Assessment scores (right caudate nucleus: r = -0.573, 95% CI: -0.801 to -0.298, p = 0.003; left caudate nucleus: r = -0.659, 95% CI: -0.894 to -0.392, p < 0.001).

CONCLUSIONS

Machine learning models trained with QSM values successfully classified PD without dementia into PD-MCI and PD-CN groups, suggesting the potential of QSM values as an auxiliary biomarker for early evaluation of cognitive decline in patients with PD.

Evaluation of the Effectiveness of Herbal Components Based on Their Regulatory Signature on Carcinogenic Cancer Cells

Predicting cancer cells’ response to a plant-derived agent is critical for the drug discovery process. Recently transcriptomes advancements have provided an opportunity to identify regulatory signatures to predict drug activity. Here in this study, a combination of meta-analysis and machine learning models have been used to determine regulatory signatures focusing on differentially expressed transcription factors (TFs) of herbal components on cancer cells. In order to increase the size of the dataset, six datasets were combined in a meta-analysis from studies that had evaluated the gene expression in cancer cell lines before and after herbal extract treatments. Then, categorical feature analysis based on the machine learning methods was applied to examine transcription factors in order to find the best signature/pattern capable of discriminating between control and treated groups. It was found that this integrative approach could recognize the combination of TFs as predictive biomarkers. It was observed that the random forest (RF) model produced the best combination rules, including AIP/TFE3/VGLL4/ID1 and AIP/ZNF7/DXO with the highest modulating capacity. As the RF algorithm combines the output of many trees to set up an ultimate model, its predictive rules are more accurate and reproducible than other trees. The discovered regulatory signature suggests an effective procedure to figure out the efficacy of investigational herbal compounds on particular cells in the drug discovery process.

Generating diagnostic profiles of cognitive decline and dementia using magnetoencephalography

Accurate identification of the underlying cause(s) of cognitive decline and dementia is challenging due to significant symptomatic overlap between subtypes. This study presents a multi-class classification framework for subjects with subjective cognitive decline, mild cognitive impairment, Alzheimer's disease, dementia with Lewy bodies, fronto-temporal dementia and cognitive decline due to psychiatric illness, trained on source-localized resting-state magnetoencephalography data. Diagnostic profiles, describing probability estimates for each of the 6 diagnoses, were assigned to individual subjects. A balanced accuracy rate of 41% and multi-class area under the curve value of 0.75 were obtained for 6-class classification. Classification primarily depended on posterior relative delta, theta and beta power and amplitude-based functional connectivity in the beta and gamma frequency band. Dementia with Lewy bodies (sensitivity: 100%, precision: 20%) and Alzheimer's disease subjects (sensitivity: 51%, precision: 90%) could be classified most accurately. Fronto-temporal dementia subjects (sensitivity: 11%, precision: 3%) were most frequently misclassified. Magnetoencephalography biomarkers hold promise to increase diagnostic accuracy in a noninvasive manner. Diagnostic profiles could provide an intuitive tool to clinicians and may facilitate implementation of the classifier in the memory clinic.

Machine learning-based statistical analysis for early stage detection of cervical cancer

Cervical cancer (CC) is the most common type of cancer in women and remains a significant cause of mortality, particularly in less developed countries, although it can be effectively treated if detected at an early stage. This study aimed to find efficient machine-learning-based classifying models to detect early stage CC using clinical data. We obtained a Kaggle data repository CC dataset which contained four classes of attributes including biopsy, cytology, Hinselmann, and Schiller. This dataset was split into four categories based on these class attributes. Three feature transformation methods, including log, sine function, and Z-score were applied to these datasets. Several supervised machine learning algorithms were assessed for their performance in classification. A Random Tree (RT) algorithm provided the best classification accuracy for the biopsy (98.33%) and cytology (98.65%) data, whereas Random Forest (RF) and Instance-Based K-nearest neighbor (IBk) provided the best performance for Hinselmann (99.16%), and Schiller (98.58%) respectively. Among the feature transformation methods, logarithmic gave the best performance for biopsy datasets whereas sine function was superior for cytology. Both logarithmic and sine functions performed the best for the Hinselmann dataset, while Z-score was best for the Schiller dataset. Various Feature Selection Techniques (FST) methods were applied to the transformed datasets to identify and prioritize important risk factors. The outcomes of this study indicate that appropriate system design and tuning, machine learning methods and classification are able to detect CC accurately and efficiently in its early stages using clinical data.

Parsimonious Models of Precipitation Phase Derived from Random Forest Knowledge: Intercomparing Logistic Models, Neural Networks, and Random Forest Models

The precipitation phase (PP) affects the hydrologic cycle which in turn affects the climate system. A lower ratio of snow to rain due to climate change affects timing and duration of the stream flow. Thus, more knowledge about the PP occurrence and drivers is necessary and especially important in cities dependent on water coming from glaciers, such as Quito, the capital of Ecuador (2.5 million inhabitants), depending in part on the Antisana glacier. The logistic models (LM) of PP rely only on air temperature and relative humidity to predict PP. However, the processes related to PP are far more complex. The aims of this study were threefold: (i) to compare the performance of random forest (RF) and artificial neural networks (ANN) to derive PP in relation to LM; (ii) to identify the main drivers of PP occurrence using RF; and (iii) to develop LM using meteorological drivers derived from RF. The results show that RF and ANN outperformed LM in predicting PP in 8 out of 10 metrics. RF indicated that temperature, dew point temperature, and specific humidity are more important than wind or radiation for PP occurrence. With these predictors, parsimonious and efficient models were developed showing that data mining may help in understanding complex processes and complements expert knowledge.

Improved Random Forest Algorithm Based on Decision Paths for Fault Diagnosis of Chemical Process with Incomplete Data

Fault detection and diagnosis (FDD) has received considerable attention with the advent of big data. Many data-driven FDD procedures have been proposed, but most of them may not be accurate when data missing occurs. Therefore, this paper proposes an improved random forest (RF) based on decision paths, named DPRF, utilizing correction coefficients to compensate for the influence of incomplete data. In this DPRF model, intact training samples are firstly used to grow all the decision trees in the RF. Then, for each test sample that possibly contains missing values, the decision paths and the corresponding nodes importance scores are obtained, so that for each tree in the RF, the reliability score for the sample can be inferred. Thus, the prediction results of each decision tree for the sample will be assigned to certain reliability scores. The final prediction result is obtained according to the majority voting law, combining both the predicting results and the corresponding reliability scores. To prove the feasibility and effectiveness of the proposed method, the Tennessee Eastman (TE) process is tested. Compared with other FDD methods, the proposed DPRF model shows better performance on incomplete data.

Machine Learning in Modeling of Mouse Behavior

Mouse behavior is a primary outcome in evaluations of therapeutic efficacy. Exhaustive, continuous, multiparametric behavioral phenotyping is a valuable tool for understanding the pathophysiological status of mouse brain diseases. Automated home cage behavior analysis produces highly granulated data both in terms of number of features and sampling frequency. Previously, we demonstrated several ways to reduce feature dimensionality. In this study, we propose novel approaches for analyzing 33-Hz data generated by CleverSys software. We hypothesized that behavioral patterns within short time windows are reflective of physiological state, and that computer modeling of mouse behavioral routines can serve as a predictive tool in classification tasks. To remove bias due to researcher decisions, our data flow is indifferent to the quality, value, and importance of any given feature in isolation. To classify day and night behavior, as an example application, we developed a data preprocessing flow and utilized logistic regression (LG), support vector machines (SVM), random forest (RF), and one-dimensional convolutional neural networks paired with long short-term memory deep neural networks (1DConvBiLSTM). We determined that a 5-min video clip is sufficient to classify mouse behavior with high accuracy. LG, SVM, and RF performed similarly, predicting mouse behavior with 85% accuracy, and combining the three algorithms in an ensemble procedure increased accuracy to 90%. The best performance was achieved by combining the 1DConv and BiLSTM algorithms yielding 96% accuracy. Our findings demonstrate that computer modeling of the home-cage ethome can clearly define mouse physiological state. Furthermore, we showed that continuous behavioral data can be analyzed using approaches similar to natural language processing. These data provide proof of concept for future research in diagnostics of complex pathophysiological changes that are accompanied by changes in behavioral profile.

Machine learning methods for predicting progression from mild cognitive impairment to Alzheimer's disease dementia: a systematic review

BACKGROUND

An increase in lifespan in our society is a double-edged sword that entails a growing number of patients with neurocognitive disorders, Alzheimer's disease being the most prevalent. Advances in medical imaging and computational power enable new methods for the early detection of neurocognitive disorders with the goal of preventing or reducing cognitive decline. Computer-aided image analysis and early detection of changes in cognition is a promising approach for patients with mild cognitive impairment, sometimes a prodromal stage of Alzheimer's disease dementia.

METHODS

We conducted a systematic review following PRISMA guidelines of studies where machine learning was applied to neuroimaging data in order to predict whether patients with mild cognitive impairment might develop Alzheimer's disease dementia or remain stable. After removing duplicates, we screened 452 studies and selected 116 for qualitative analysis.

RESULTS

Most studies used magnetic resonance image (MRI) and positron emission tomography (PET) data but also magnetoencephalography. The datasets were mainly extracted from the Alzheimer's disease neuroimaging initiative (ADNI) database with some exceptions. Regarding the algorithms used, the most common was support vector machine with a mean accuracy of 75.4%, but convolutional neural networks achieved a higher mean accuracy of 78.5%. Studies combining MRI and PET achieved overall better classification accuracy than studies that only used one neuroimaging technique. In general, the more complex models such as those based on deep learning, combined with multimodal and multidimensional data (neuroimaging, clinical, cognitive, genetic, and behavioral) achieved the best performance.

CONCLUSIONS

Although the performance of the different methods still has room for improvement, the results are promising and this methodology has a great potential as a support tool for clinicians and healthcare professionals.

Gene clusters based on OLIG2 and CD276 could distinguish molecular profiling in glioblastoma

Background

The molecular profiling of glioblastoma (GBM) based on transcriptomic analysis could provide precise treatment and prognosis. However, current subtyping (classic, mesenchymal, neural, proneural) is time-consuming and cost-intensive hindering its clinical application. A simple and efficient method for classification was imperative.

Methods

In this study, to simplify GBM subtyping more efficiently, we applied a random forest algorithm to conduct 26 genes as a cluster featured with hub genes, OLIG2 and CD276. Functional enrichment analysis and Protein–protein interaction were performed using the genes in this gene cluster. The classification efficiency of the gene cluster was validated by WGCNA and LASSO algorithms, and tested in GSE84010 and Gravandeel’s GBM datasets.

Results

The gene cluster (n = 26) could distinguish mesenchymal and proneural excellently (AUC = 0.92), which could be validated by multiple algorithms (WGCNA, LASSO) and datasets (GSE84010 and Gravandeel’s GBM dataset). The gene cluster could be functionally enriched in DNA elements and T cell associated pathways. Additionally, five genes in the signature could predict the prognosis well (p = 0.0051 for training cohort, p = 0.065 for test cohort).

Conclusions

Our study proved the accuracy and efficiency of random forest classifier for GBM subtyping, which could provide a convenient and efficient method for subtyping Proneural and Mesenchymal GBM.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-021-03083-y.

Aggregate Trends of Apolipoprotein E on Cognition in Transgenic Alzheimer's Disease Mice

Background:

Apolipoprotein E (APOE) genotypes typically increase risk of amyloid-β deposition and onset of clinical Alzheimer’s disease (AD). However, cognitive assessments in APOE transgenic AD mice have resulted in discord.

Objective:

Analysis of 31 peer-reviewed AD APOE mouse publications (n = 3,045 mice) uncovered aggregate trends between age, APOE genotype, gender, modulatory treatments, and cognition.

Methods:

T-tests with Bonferroni correction (significance = p < 0.002) compared age-normalized Morris water maze (MWM) escape latencies in wild type (WT), APOE2 knock-in (KI2), APOE3 knock-in (KI3), APOE4 knock-in (KI4), and APOE knock-out (KO) mice. Positive treatments (t+) to favorably modulate APOE to improve cognition, negative treatments (t–) to perturb etiology and diminish cognition, and untreated (t0) mice were compared. Machine learning with random forest modeling predicted MWM escape latency performance based on 12 features: mouse genotype (WT, KI2, KI3, KI4, KO), modulatory treatment (t+, t–, t0), mouse age, and mouse gender (male = g_m; female = g_f, mixed gender = g_mi).

Results:

KI3 mice performed significantly better in MWM, but KI4 and KO performed significantly worse than WT. KI2 performed similarly to WT. KI4 performed significantly worse compared to every other genotype. Positive treatments significantly improved cognition in WT, KI4, and KO compared to untreated. Interestingly, negative treatments in KI4 also significantly improved mean MWM escape latency. Random forest modeling resulted in the following feature importance for predicting superior MWM performance: [KI3, age, g_m, KI4, t0, t+, KO, WT, g_mi, t–, g_f, KI2] = [0.270, 0.094, 0.092, 0.088, 0.077, 0.074, 0.069, 0.061, 0.058, 0.054, 0.038, 0.023].

Conclusion:

APOE3, age, and male gender was most important for predicting superior mouse cognitive performance.

Real-time prediction of intradialytic relative blood volume: a proof-of-concept for integrated cloud computing infrastructure

BACKGROUND

Inadequate refilling from extravascular compartments during hemodialysis can lead to intradialytic symptoms, such as hypotension, nausea, vomiting, and cramping/myalgia. Relative blood volume (RBV) plays an important role in adapting the ultrafiltration rate which in turn has a positive effect on intradialytic symptoms. It has been clinically challenging to identify changes RBV in real time to proactively intervene and reduce potential negative consequences of volume depletion. Leveraging advanced technologies to process large volumes of dialysis and machine data in real time and developing prediction models using machine learning (ML) is critical in identifying these signals.

METHOD

We conducted a proof-of-concept analysis to retrospectively assess near real-time dialysis treatment data from in-center patients in six clinics using Optical Sensing Device (OSD), during December 2018 to August 2019. The goal of this analysis was to use real-time OSD data to predict if a patient's relative blood volume (RBV) decreases at a rate of at least - 6.5 % per hour within the next 15 min during a dialysis treatment, based on 10-second windows of data in the previous 15 min. A dashboard application was constructed to demonstrate how reporting structures may be developed to alert clinicians in real time of at-risk cases. Data was derived from three sources: (1) OSDs, (2) hemodialysis machines, and (3) patient electronic health records.

RESULTS

Treatment data from 616 in-center dialysis patients in the six clinics was curated into a big data store and fed into a Machine Learning (ML) model developed and deployed within the cloud. The threshold for classifying observations as positive or negative was set at 0.08. Precision for the model at this threshold was 0.33 and recall was 0.94. The area under the receiver operating curve (AUROC) for the ML model was 0.89 using test data.

CONCLUSIONS

The findings from our proof-of concept analysis demonstrate the design of a cloud-based framework that can be used for making real-time predictions of events during dialysis treatments. Making real-time predictions has the potential to assist clinicians at the point of care during hemodialysis.

Lipidomics Prediction of Parkinson's Disease Severity: A Machine-Learning Analysis

BACKGROUND

The role of the lipidome as a biomarker for Parkinson's disease (PD) is a relatively new field that currently only focuses on PD diagnosis.

OBJECTIVE

To identify a relevant lipidome signature for PD severity markers.

METHODS

Disease severity of 149 PD patients was assessed by the Unified Parkinson's Disease Rating Scale (UPDRS) and the Montreal Cognitive Assessment (MoCA). The lipid composition of whole blood samples was analyzed, consisting of 517 lipid species from 37 classes; these included all major classes of glycerophospholipids, sphingolipids, glycerolipids, and sterols. To handle the high number of lipids, the selection of lipid species and classes was consolidated via analysis of interrelations between lipidomics and disease severity prediction using the random forest machine-learning algorithm aided by conventional statistical methods.

RESULTS

Specific lipid classes dihydrosphingomyelin (dhSM), plasmalogen phosphatidylethanolamine (PEp), glucosylceramide (GlcCer), dihydro globotriaosylceramide (dhGB3), and to a lesser degree dihydro GM3 ganglioside (dhGM3), as well as species dhSM(20:0), PEp(38:6), PEp(42:7), GlcCer(16:0), GlcCer(24:1), dhGM3(22:0), dhGM3(16:0), and dhGB3(16:0) contribute to PD severity prediction of UPDRS III score. These, together with age, age at onset, and disease duration, also contribute to prediction of UPDRS total score. We demonstrate that certain lipid classes and species interrelate differently with the degree of severity of motor symptoms between men and women, and that predicting intermediate disease stages is more accurate than predicting less or more severe stages.

CONCLUSION

Using machine-learning algorithms and methodologies, we identified lipid signatures that enable prediction of motor severity in PD. Future studies should focus on identifying the biological mechanisms linking GlcCer, dhGB3, dhSM, and PEp with PD severity.

Prediction of Motor Function in Stroke Patients Using Machine Learning Algorithm: Development of Practical Models

BACKGROUND

Machine learning (ML) techniques are being increasingly adopted in the medical field.

OBJECTIVE

We developed a deep neural network (DNN) model and applied 2 well-known ML algorithms, logistic regression and random forest, in predicting motor outcome at 6 months after stroke.

METHODS

In the present study, by using 14 input variables which are easily measured by clinicians, we developed ML models and investigated their applicability to predicting motor outcome in hemiplegic stroke patients. We retrospectively analyzed data of 1,056 consecutive stroke patients. Favorable outcomes of the upper and lower limbs were defined as a modified Brunnstrom classification (MBC) score of ≥5 (able to perform activities of daily living with the affected upper limb) and a functional ambulation category (FAC) score of ≥4 (able to walk without guardian's assistance), respectively. Poor outcomes of the upper and lower limbs were defined as MBC and FAC scores of <5 and <4, respectively. We developed 3 ML algorithms, namely the DNN, logistic regression, and random forest.

RESULTS

Regarding the prediction of upper limb function, for the DNN model, the area under the curve (AUC) was 0.906. For the logistic regression and random forest models, the AUC were 0.874 and 0.882, respectively. For the prediction of lower limb function, for the DNN, logistic regression, and random forest models, the AUCs were 0.822, 0.768, and 0.802, respectively.

CONCLUSIONS

We demonstrated that the ML algorithms, particularly the DNN, can be useful for predicting motor outcomes in the upper and lower limbs at 6 months after stroke.

Nonadditivity in public and inhouse data: implications for drug design

Numerous ligand-based drug discovery projects are based on structure-activity relationship (SAR) analysis, such as Free-Wilson (FW) or matched molecular pair (MMP) analysis. Intrinsically they assume linearity and additivity of substituent contributions. These techniques are challenged by nonadditivity (NA) in protein-ligand binding where the change of two functional groups in one molecule results in much higher or lower activity than expected from the respective single changes. Identifying nonlinear cases and possible underlying explanations is crucial for a drug design project since it might influence which lead to follow. By systematically analyzing all AstraZeneca (AZ) inhouse compound data and publicly available ChEMBL25 bioactivity data, we show significant NA events in almost every second assay among the inhouse and once in every third assay in public data sets. Furthermore, 9.4% of all compounds of the AZ database and 5.1% from public sources display significant additivity shifts indicating important SAR features or fundamental measurement errors. Using NA data in combination with machine learning showed that nonadditive data is challenging to predict and even the addition of nonadditive data into training did not result in an increase in predictivity. Overall, NA analysis should be applied on a regular basis in many areas of computational chemistry and can further improve rational drug design.

Can machine learning improve mortality prediction following cardiac surgery?

OBJECTIVES

Interest in the clinical usefulness of machine learning for risk prediction has bloomed recently. Cardiac surgery patients are at high risk of complications and therefore presurgical risk assessment is of crucial relevance. We aimed to compare the performance of machine learning algorithms over traditional logistic regression (LR) model to predict in-hospital mortality following cardiac surgery.

METHODS

A single-centre data set of prospectively collected information from patients undergoing adult cardiac surgery from 1996 to 2017 was split into 70% training set and 30% testing set. Prediction models were developed using neural network, random forest, naive Bayes and retrained LR based on features included in the EuroSCORE. Discrimination was assessed using area under the receiver operating characteristic curve, and calibration analysis was undertaken using the calibration belt method. Model calibration drift was assessed by comparing Goodness of fit χ2 statistics observed in 2 equal bins from the testing sample ordered by procedure date.

RESULTS

A total of 28 761 cardiac procedures were performed during the study period. The in-hospital mortality rate was 2.7%. Retrained LR [area under the receiver operating characteristic curve 0.80; 95% confidence interval (CI) 0.77-0.83] and random forest model (0.80; 95% CI 0.76-0.83) showed the best discrimination. All models showed significant miscalibration. Retrained LR proved to have the weakest calibration drift.

CONCLUSIONS

Our findings do not support the hypothesis that machine learning methods provide advantage over LR model in predicting operative mortality after cardiac surgery.