A hybrid machine learning approach to cerebral stroke prediction based on imbalanced medical dataset

BACKGROUND AND OBJECTIVE

Cerebral stroke has become a significant global public health issue in recent years. The ideal solution to this concern is to prevent in advance by controlling related metabolic factors. However, it is difficult for medical staff to decide whether special precautions are needed for a potential patient only based on the monitoring of physiological indicators unless they are obviously abnormal. This paper will develop a hybrid machine learning approach to predict cerebral stroke for clinical diagnosis based on the physiological data with incompleteness and class imbalance.

METHODS

Two steps are involved in the whole process. Firstly, random forest regression is adopted to impute missing values before classification. Secondly, an automated hyperparameter optimization(AutoHPO) based on deep neural network(DNN) is applied to stroke prediction on an imbalanced dataset.

RESULTS

The medical dataset contains 43,400 records of potential patients which includes 783 occurrences of stroke. The false negative rate from our prediction approach is only 19.1%, which has reduced by an average of 51.5% in comparison to other traditional approaches. The false positive rate, accuracy and sensitivity predicted by the proposed approach are respectively 33.1, 71.6, and 67.4%.

CONCLUSION

The approach proposed in this paper has effectively reduced the false negative rate with a relatively high overall accuracy, which means a successful decrease in the misdiagnosis rate for stroke prediction. The results are more reliable and valid as the reference in stroke prognosis, and also can be acquired conveniently at a low cost.

The influence of skeletal muscle on appetite regulation

INTRODUCTION

Fat-free mass, of which skeletal muscle is amajor component, correlates positively with energy intake at energy balance. This is due to the effects of metabolically active tissue on energy expenditure which in itself appears to signal to the brain adrive to eat to ensure cellular energy homeostasis. The mechanisms responsible for this drive to eat are unknown but are likely to be related to energy utilization. Here muscle imparts an indirect influence on hunger. The drive to eat is also enhanced after muscle loss secondary to intentional weight loss. The evidence suggests loss of both fat mass and skeletal muscle mass directly influences the trajectory and magnitude of weight regain highlighting their potential role in long-termappetite control. The mechanisms responsible for the potential direct drive to eat stemming from muscle loss are unknown.

AREAS COVERED

The literature pertaining to muscle and appetite at energy balance and after weight loss was examined. Aliterature search was conducted to identify studies related to appetite, muscle, exercise, and weight loss.

EXPERT OPINION

Understanding the mechanisms which link energy expenditure and muscle loss to hunger has the potential to positively impact both the prevention and the treatment of obesity.

Use of a multilayer perceptron to create a prediction model for dressing independence in a small sample at a single facility

[Purpose] This study aimed to assess the accuracy of a prediction model for dressing independence created with a multilayer perceptron in a small sample at a single facility. [Participants and Methods] This retrospective observational study included 82 first-stroke patients. The prediction models for dressing independence at hospital discharge were created using a multilayer perceptron, logistic regression, and a decision tree, and compared for predictive accuracy. Age, dressing performance, trunk function, visuospatial perception, balance, and cognitive function at admission were used as variables. [Results] The area under the receiver operating characteristic curve, classification accuracy, sensitivity, specificity, positive-predictive value, and negative-predictive value for training data were highest with the multilayer perceptron model. Cochran’s Q and multiple comparison tests revealed a significant difference between logistic regression and multilayer perceptron models. Testing of data in 10-fold cross-validation yielded the same results, except for sensitivity. [Conclusion] The present study suggested that higher accuracy could be expected with a multilayer perceptron than with logistic regression and a decision tree when creating a prediction model for independence of activities of daily living in a small sample of stroke patients.

Roles of myokines in exercise-induced improvement of neuropsychiatric function

Exercise is a well-known non-pharmacological intervention to improve brain functions, including cognition, memory, and motor coordination. Contraction of skeletal muscles during exercise releases humoral factors that regulate the whole-body metabolism via interaction with other non-muscle organs. Myokines are muscle-derived effectors that regulate body metabolism by autocrine, paracrine, or endocrine action and were reportedly suggested as "exercise factors" that can improve the brain function. However, several aspects remain to be elucidated, namely the specific activities of myokines related to the whole-body metabolism or brain function, the mechanisms of regulation of other organs or cells, the sources of "exercise factors" that regulate brain function, and their mechanisms of interaction with non-muscle organs. In this paper, we present the physiological functions of myokines secreted by exercise, including regulation of the whole-body metabolism by interaction with other organs and adaptation of skeletal muscles to exercise. In addition, we discuss the functions of myokines that possibly contribute to exercise-induced improvement of brain function. Among several myokines, brain-derived neurotrophic factor (BDNF) is the most studied myokine that regulates adult neurogenesis and synaptic plasticity. However, the source of circulating BDNF and its upstream effector, insulin-like growth factor (IGF-1), and irisin and the effect size of peripheral BDNF, irisin, and IGF-1 released after exercise should be further investigated. Recently, cathepsin B has been reported to be secreted from skeletal muscles and upregulate BDNF following exercise, which was associated with improved cognitive function. We reviewed the level of evidence for the effect of myokine on the brain function. Level of evidence for the association of the change in circulating myokine following exercise and improvement of neuropsychiatric function is lower than the level of evidence for the benefit of exercise on the brain. Therefore, more clinical evidences for the association of myokine release after exercise and their effect on the brain function are required. Finally, we discuss the effect size of the action of myokines on cognitive benefits of exercise, in addition to other contributors, such as improvement of the cardiovascular system or the effect of "exercise factors" released from non-muscle organs, particularly in patients with sarcopenia.

A systematic literature review and classification of knowledge discovery in traditional medicine

INTRODUCTION AND OBJECTIVE

Despite the importance of machine learning methods application in traditional medicine there is a no systematic literature review and a classification for this field. This is the first comprehensive literature review of the application of data mining methods in traditional medicine.

METHOD

We reviewed 5 database between 2000 to 2017 based on the Kitchenham systematic review methodology. 502 articles were identified and reviewed for their relevance to application of machine learning methods in traditional medicine, 42 selected papers were classified and categorized on four dimension; 1) application domain of data mining techniques in traditional medicine; 2) the data mining methods most frequently used in traditional medicine; 3) main strength and limitation of data mining techniques in traditional medicine; 4) the performance evaluation methods in data mining methods in traditional medicine.

RESULT

The result obtained showed that main application domain of data mining techniques in traditional medicine was related to syndrome differentiation. Bayesian Networks (BNs), Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) were recognized as being the methods most frequently applied in traditional medicine. Furthermore, each data mining techniques has its own strength and limitations when applied in traditional medicine. Single scaler methods were frequently used for performance evaluation of data mining methods.

CONCLUSION

Machine learning methods have become an important research field in traditional medicine. Our research provides information about this methods by examining the related articles.

Central irisin administration affords antidepressant-like effect and modulates neuroplasticity-related genes in the hippocampus and prefrontal cortex of mice

Evidence has indicated that the practice of physical exercise has antidepressant effects that might be associated with irisin release and BDNF signaling. In this study we investigated the effects of the central administration of irisin or BDNF in predictive tests of antidepressant properties paralleled with the gene expression of peroxisome proliferator-activated receptor gamma co-activator 1α (PGC-1α), fibronectin type III domain-containing protein 5 (FNDC5) and brain-derived neurotrophic factor (BDNF) in the hippocampus and prefrontal cortex of mice. Irisin (0.5-1 ng/mouse, i.c.v.) reduced the immobility time in the tail suspension test (TST) and forced swim test (FST), without altering locomotion in the open field test (OFT). Irisin reduced the immobility time in the TST up to 6 h after its administration. Irisin administration (6 h) increased PGC-1α mRNA in the hippocampus and prefrontal cortex and reduced (1 h) PGC-1α mRNA in the prefrontal cortex. FNDC5 and BDNF mRNA expression was decreased (1 h) in both structures and remained reduced up to 6 h in the prefrontal cortex. Moreover, BDNF administered at 0.25 μg/mouse, i.c.v. (1 and 6 h before the test) reduced the immobility time in the TST. BDNF administration reduced PGC-1α mRNA in the hippocampus (6 h) and prefrontal cortex (1 and 6 h). It also increased FNDC5 mRNA expression in the hippocampus (1 and 6 h), but reduced the expression of this gene and also BDNF mRNA in the prefrontal cortex (1 and 6 h). None of the treatments altered BDNF protein levels in both structures. In conclusion, irisin presents a behavioral antidepressant profile similar to BDNF, an effect associated with the modulation of gene expression of PGC-1α, FNDC5 and BDNF, reinforcing the pivotal role of these genes in mood regulation.

An update on the role of irisin in the regulation of endocrine and metabolic functions

Irisin is a novel myokine and adipokine that has gained much attention recently due to its mechanisms of action. Irisin is secreted following proteolytic cleavage of its precursor fibronectin type III domain containing 5 (FNDC5). Following its release, irisin exerts its major action by increasing the expression of mitochondrial uncoupling protein 1 (UCP 1), which facilitates the conversion of white adipose tissue (WAT) into beige adipose tissue. Irisin is distributed in various body tissues and several actions have been attributed to its presence in those tissues. It has been suggested that it plays a role in metabolic diseases, ageing, inflammation and neurogenesis. However, the circulating levels of irisin are modulated by several factors such as diet, obesity, exercise, pharmacological agents and different pathological conditions. In this review, we have discussed the mechanisms by which irisin influences the functions of different body systems and how external factors in turn affect the circulating level of irisin. In conclusion, modification of circulating irisin level may help in the management of a variety of endocrine and metabolic disorders.

Towards phenotyping stroke: Leveraging data from a large-scale epidemiological study to detect stroke diagnosis

Objective

1) To develop a machine learning approach for detecting stroke cases and subtypes from hospitalization data, 2) to assess algorithm performance and predictors on real-world data collected by a large-scale epidemiology study in the US; and 3) to identify directions for future development of high-precision stroke phenotypic signatures.

Materials and methods

We utilized 8,131 hospitalization events (ICD-9 codes 430–438) collected from the Greater Cincinnati/Northern Kentucky Stroke Study in 2005 and 2010. Detailed information from patients’ medical records was abstracted for each event by trained research nurses. By analyzing the broad list of demographic and clinical variables, the machine learning algorithms predicted whether an event was a stroke case and, if so, the stroke subtype. The performance was validated on gold-standard labels adjudicated by stroke physicians, and results were compared with stroke classifications based on ICD-9 discharge codes, as well as labels determined by study nurses.

Results

The best performing machine learning algorithm achieved a performance of 88.57%/93.81%/92.80%/93.30%/89.84%/98.01% (accuracy/precision/recall/F-measure/area under ROC curve/area under precision-recall curve) on stroke case detection. For detecting stroke subtypes, the algorithm yielded an overall accuracy of 87.39% and greater than 85% precision on individual subtypes. The machine learning algorithms significantly outperformed the ICD-9 method on all measures (P value<0.001). Their performance was comparable to that of study nurses, with better tradeoff between precision and recall. The feature selection uncovered a subset of predictive variables that could facilitate future development of effective stroke phenotyping algorithms.

Discussion and conclusions

By analyzing a broad array of patient data, the machine learning technologies held promise for improving detection of stroke diagnosis, thus unlocking high statistical power for subsequent genetic and genomic studies.

Synergistic effect of carnosine on browning of adipose tissue in exercised obese rats; a focus on circulating irisin levels

The recent appreciation of the energy burning capacity of brown adipose tissue turns it to an attractive target for anti-obesity therapy. We sought to evaluate the effect of L-carnosine on browning of white adipose tissue in exercised obese rats. Sixty adult male Wistar albino rats, 7-8 week-old weighing 130-150 g, were allocated into six groups; with 10 rats in each, for an experimentation period of 12 weeks: (i) normal control rats fed a standard fat diet (SFD/control), (ii) normal control rats fed a standard diet and injected with L-carnosine (250 mg/kg, i.p,) for 6 weeks (SFD/CAR), (iii) high-fat diet (HFD)-induced obese rats for 12 weeks, (iv) HFD rats subjected to exercise training (HFD/EXE) for 6 weeks, (v) HFD rats injected with L-carnosine (250 mg/kg,i.p.) for 6 weeks (HFD/CAR) and, (vi) HFD rats subjected to exercise training and L-carnosine (HFD/EXE/CAR). At the end of the 12-week-experiment, the body weights and the serum levels of lipid profile, oxidative stress, and inflammatory markers as well as circulating myokines were investigated. Gastrocnemius muscles and inguinal adipose tissues were excised for the measurement of gene expression of muscle irisin, adipose tissue uncoupling protein1 (UCP1), CD137 and the protein level of p38MAPK. In addition, histopathological examination for the studied groups was performed. Both exercise training for 6 weeks and carnosine treatment significantly decreased body weight gain, ameliorated obesity-induced dyslipidemia, reduced the thiobarbituric acid reactive species (TBARS) and TNF-α, while increased total antioxidant capacity and IL-10. Furthermore, increases in serum irisin levels and the expression of adipose uncoupling protein-1 (UCP-1), adipose CD137, p38 MAPK, and muscular fibronectin type III domain-containing protein 5(FNDC5), the precursor of irisin gene expression, were correlated with these carnosine- and exercise-induced physiological improvements. The highest improvement was evident in the combined exercise and carnosine group which indicates that their beneficial effects in obese animals were synergistic. These findings suggest that L-carnosine may induce browning of adipose tissue through irisin stimulation, a phenomenon that could be related to its antioxidant, anti-inflammatory, and anti-obesity effects.

Small interfering RNA mediated knockdown of irisin suppresses food intake and modulates appetite regulatory peptides in zebrafish

Irisin is a myokine encoded in fibronectin type III domain containing 5 (FNDC5). FNDC5 forms an integral part of the muscle post-exercise, and causes an increase in energy expenditure in mammals. Irisin is abundantly expressed in cardiac and skeletal muscles and is secreted upon activation of peroxisome proliferator-activated receptor gamma coactivator-1 (PGC-1 alpha). Irisin regulates feeding behaviour and cardiovascular function in mammals. More recently, irisin has gained importance as a potential biomarker for myocardial infarction due to its abundance in cardiac muscle. The goal of this research was to determine whether irisin influences feeding, and regulates appetite regulatory peptides in zebrafish. Intraperitoneal injection of irisin [0.1, 1, 10 and 100ng/g body weight (BW)] did not affect feeding, but its knockdown using siRNA (10ng/g BW) caused a significant reduction in food intake. Knockdown of irisin reduced ghrelin and orexin-A mRNA expression, and increased cocaine and amphetamine regulated transcript mRNA expression in zebrafish brain and gut. siRNA mediated knockdown of irisin also downregulated brain derived neurotrophic factor mRNA in zebrafish. The role of endogenous irisin on food intake is likely mediated by its actions on other metabolic peptides. Collectively, these results indicate that unaltered endogenous irisin is required to maintain food intake in zebrafish.

Classification techniques on computerized systems to predict and/or to detect Apnea: A systematic review

BACKGROUND AND OBJECTIVE

Sleep apnea syndrome (SAS), which can significantly decrease the quality of life is associated with a major risk factor of health implications such as increased cardiovascular disease, sudden death, depression, irritability, hypertension, and learning difficulties. Thus, it is relevant and timely to present a systematic review describing significant applications in the framework of computational intelligence-based SAS, including its performance, beneficial and challenging effects, and modeling for the decision-making on multiple scenarios.

METHODS

This study aims to systematically review the literature on systems for the detection and/or prediction of apnea events using a classification model.

RESULTS

Forty-five included studies revealed a combination of classification techniques for the diagnosis of apnea, such as threshold-based (14.75%) and machine learning (ML) models (85.25%). In addition, the ML models, were clustered in a mind map, include neural networks (44.26%), regression (4.91%), instance-based (11.47%), Bayesian algorithms (1.63%), reinforcement learning (4.91%), dimensionality reduction (8.19%), ensemble learning (6.55%), and decision trees (3.27%).

CONCLUSIONS

A classification model should provide an auto-adaptive and no external-human action dependency. In addition, the accuracy of the classification models is related with the effective features selection. New high-quality studies based on randomized controlled trials and validation of models using a large and multiple sample of data are recommended.

Association between completed suicide and environmental temperature in a Mexican population, using the Knowledge Discovery in Database approach

BACKGROUND AND OBJECTIVE

Suicide is a worldwide health problem and climatological characteristics have been associated with suicide behavior. However, approaches such as the Knowledge Discovery in Database are not frequently used to search for an association between climatological characteristics and suicide. The aim of the present study was to assess the association between weather data and suicide in a Mexican population.

METHODS

We used the information of 1357 patients who completed suicide from 2005 to 2012. Alternatively, weather data were provided by the National Water Commission. We used the Knowledge Discovery in Database approach with an Apriori algorithm and the data analyses were performed with the Waikato Environment for Knowledge Analysis software. One hundred rules of association were generated with a confidence of 0.86 and support of 1.

RESULTS

We found an association between environmental temperature and suicide: days with no rain and temperatures between 30 °C and 40 °C (86-104 °F) were related to males completing suicide by hanging.

CONCLUSIONS

In the prevention of suicidal behavior, the Knowledge Discovery in Database could be used to establish climatological characteristics and their association with suicide. This approach must be considered in future prevention strategies.

Different medical data mining approaches based prediction of ischemic stroke

AIM

Medical data mining (also called knowledge discovery process in medicine) processes for extracting patterns from large datasets. In the current study, we intend to assess different medical data mining approaches to predict ischemic stroke.

MATERIALS AND METHODS

The collected dataset from Turgut Ozal Medical Centre, Inonu University, Malatya, Turkey, comprised the medical records of 80 patients and 112 healthy individuals with 17 predictors and a target variable. As data mining approaches, support vector machine (SVM), stochastic gradient boosting (SGB) and penalized logistic regression (PLR) were employed. 10-fold cross validation resampling method was utilized, and model performance evaluation metrics were accuracy, area under ROC curve (AUC), sensitivity, specificity, positive predictive value and negative predictive value. The grid search method was used for optimizing tuning parameters of the models.

RESULTS

The accuracy values with 95% CI were 0.9789 (0.9470-0.9942) for SVM, 0.9737 (0.9397-0.9914) for SGB and 0.8947 (0.8421-0.9345) for PLR. The AUC values with 95% CI were 0.9783 (0.9569-0.9997) for SVM, 0.9757 (0.9543-0.9970) for SGB and 0.8953 (0.8510-0.9396) for PLR.

CONCLUSIONS

The results of the current study demonstrated that the SVM produced the best predictive performance compared to the other models according to the majority of evaluation metrics. SVM and SGB models explained in the current study could yield remarkable predictive performance in the classification of ischemic stroke.

A novel method of artery stenosis diagnosis using transfer function and support vector machine based on transmission line model: A numerical simulation and validation study

BACKGROUND AND OBJECTIVE

Transfer function (TF) is an important parameter for the analysis and understanding of hemodynamics when arterial stenosis exists in human arterial tree. Aimed to validate the feasibility of using TF to diagnose arterial stenosis, the forward problem and inverse problem were simulated and discussed.

METHODS

A calculation method of TF between ascending aorta and any other artery was proposed based on a 55 segment transmission line model (TLM) of human artery tree. The effects of artery stenosis on TF were studied in two aspects: stenosis degree and position. The degree of arterial stenosis was specified to be 10-90% in three representative arteries: carotid, aorta and iliac artery, respectively. In order to validate the feasibility of diagnosis of artery stenosis using TF and support vector machine (SVM), a database of TF was established to simulate the real conditions of artery stenosis based on the TLM model. And a diagnosis model of artery stenosis was built by using SVM and the database.

RESULTS

The simulating results showed the modulus and phase of TF were decreasing sharply from frequency 2 to 10Hz with the stenosis degree increasing and displayed their unique and nonlinear characteristics when frequency is higher than 10Hz. The diagnosis results showed the average accuracy was above 76% for the stenosis from 10% to 90% degree, and the diagnosis accuracies of moderate (50%) and serious (90%) stenosis were 87% and 99%, respectively. When the stenosis degree increased to 90%, the accuracy of stenosis localization reached up to 94% for most of arteries.

CONCLUSIONS

The proposed method of combining TF and SVM is a theoretically feasible method for diagnosis of artery stenosis.