Performance comparison of machine learning algorithms and number of independent components used in fMRI decoding of belief vs. disbelief

Source-free collaborative domain adaptation via multi-perspective feature enrichment for functional MRI analysis

Resting-state functional MRI (rs-fMRI) is increasingly employed in multi-site research to analyze neurological disorders, but there exists cross-site/domain data heterogeneity caused by site effects such as differences in scanners/protocols. Existing domain adaptation methods that reduce fMRI heterogeneity generally require accessing source domain data, which is challenging due to privacy concerns and/or data storage burdens. To this end, we propose a source-free collaborative domain adaptation (SCDA) framework using only a pretrained source model and unlabeled target data. Specifically, a multi-perspective feature enrichment method (MFE) is developed to dynamically exploit target fMRIs from multiple views. To facilitate efficient source-to-target knowledge transfer without accessing source data, we initialize MFE using parameters of a pretrained source model. We also introduce an unsupervised pretraining strategy using 3,806 unlabeled fMRIs from three large-scale auxiliary databases. Experimental results on three public and one private datasets show the efficacy of our method in cross-scanner and cross-study prediction.

Intrusion detection in machine learning based E-shaped structure with algorithms, strategies and applications in wireless sensor networks

In the everyday world of computer applications, from the cloud to the Internet of Things, distributed sensor networks are essential (IoT). These computer application devices are often connected to Arduino network connection and microcontrollers such sensors and actuators. Thus, a defensive network with an IDS serves as the need for contemporary networks. The intrusion detection system has unavoidably evolved throughout the years, but despite this, it remains a difficult study topic since the current intrusion detection system uses signature-based approaches rather than anomaly detection. Therefore, improving the current intrusion detection system is challenging since it is difficult to find zero-day attacks in IoT networks when dealing with varied data sources. Filtered Deep Learning Model for Intrusion Detection with a Data Communication Approach is presented in this study. The five steps that make up the suggested model are Initialization of Sensor Networks, Cluster Formation and Head Selection, Connectivity, Attack Detection, and Data Broker. It was discovered that the suggested model for intrusion detection outperformed both the current Deep Learning Neural Net and Artificial Neural Network. In comparison to the most popular algorithms, experimental findings revealed a superior result of 96.12 % accuracy. The E-shaped patch antenna is a brand-new single-patch wide-band microstrip antenna that is presented in this research. A microstrip antenna's patch has two parallel slots built into it to increase its bandwidth. Investigating the behaviour of the currents on the patch allows for the exploration of the wide-band mechanism. A broad bandwidth is achieved by optimising the slot's length, breadth, and location. Finally, a 40.3 % E-shaped patch antenna is developed, made, and tested to resonate at 7.5 and 8.5 GHz for wireless communications. Additionally displayed are the reflection coefficient, VSWR, radiation pattern and directivity.

Bayesian Networks in Radiology

A Bayesian network is a graphical model that uses probability theory to represent relationships among its variables. The model is a directed acyclic graph whose nodes represent variables, such as the presence of a disease or an imaging finding. Connections between nodes express causal influences between variables as probability values. Bayesian networks can learn their structure (nodes and connections) and/or conditional probability values from data. Bayesian networks offer several advantages: (a) they can efficiently perform complex inferences, (b) reason from cause to effect or vice versa, (c) assess counterfactual data, (d) integrate observations with canonical ("textbook") knowledge, and (e) explain their reasoning. Bayesian networks have been employed in a wide variety of applications in radiology, including diagnosis and treatment planning. Unlike deep learning approaches, Bayesian networks have not been applied to computer vision. However, hybrid artificial intelligence systems have combined deep learning models with Bayesian networks, where the deep learning model identifies findings in medical images and the Bayesian network formulates and explains a diagnosis from those findings. One can apply a Bayesian network's probabilistic knowledge to integrate clinical and imaging findings to support diagnosis, treatment planning, and clinical decision-making. This article reviews the fundamental principles of Bayesian networks and summarizes their applications in radiology. Keywords: Bayesian Network, Machine Learning, Abdominal Imaging, Musculoskeletal Imaging, Breast Imaging, Neurologic Imaging, Radiology Education Supplemental material is available for this article. © RSNA, 2023.

Prediction of suicidality in bipolar disorder using variability of intrinsic brain activity and machine learning

Bipolar disorder (BD) is associated with marked suicidal susceptibility, particularly during a major depressive episode. However, the evaluation of suicidal risk remains challenging since it relies mainly on self-reported information from patients. Hence, it is necessary to complement neuroimaging features with advanced machine learning techniques in order to predict suicidal behavior in BD patients. In this study, a total of 288 participants, including 75 BD suicide attempters, 101 BD nonattempters and 112 healthy controls, underwent a resting-state functional magnetic resonance imaging (rs-fMRI). Intrinsic brain activity was measured by amplitude of low-frequency fluctuation (ALFF). We trained and tested a two-level k-nearest neighbors (k-NN) model based on resting-state variability of ALFF with fivefold cross-validation. BD suicide attempters had increased dynamic ALFF values in the right anterior cingulate cortex, left thalamus and right precuneus. Compared to other machine learning methods, our proposed framework had a promising performance with 83.52% accuracy, 78.75% sensitivity and 87.50% specificity. The trained models could also replicate and validate the results in an independent cohort with 72.72% accuracy. These findings based on a relatively large data set, provide a promising way of combining fMRI data with machine learning technique to reliably predict suicide attempt at an individual level in bipolar depression. Overall, this work might enhance our understanding of the neurobiology of suicidal behavior by detecting clinically defined disruptions in the dynamics of instinct brain activity.

Constrained IoT-Based Machine Learning for Accurate Glycemia Forecasting in Type 1 Diabetes Patients

Individuals with diabetes mellitus type 1 (DM1) tend to check their blood sugar levels multiple times daily and utilize this information to predict their future glycemic levels. Based on these predictions, patients decide on the best approach to regulate their glucose levels with considerations such as insulin dosage and other related factors. Nevertheless, modern developments in Internet of Things (IoT) technology and innovative biomedical sensors have enabled the constant gathering of glucose level data using continuous glucose monitoring (CGM) in addition to other biomedical signals. With the use of machine learning (ML) algorithms, glycemic level patterns can be modeled, enabling accurate forecasting of this variable. Constrained devices have limited computational power, making it challenging to run complex machine learning algorithms directly on these devices. However, by leveraging edge computing, using lightweight machine learning algorithms, and performing preprocessing and feature extraction, it is possible to run machine learning algorithms on constrained devices despite these limitations. In this paper we test the burdens of some constrained IoT devices, probing that it is feasible to locally predict glycemia using a smartphone, up to 45 min in advance and with acceptable accuracy using random forest.

Explainable AI: A review of applications to neuroimaging data

Deep neural networks (DNNs) have transformed the field of computer vision and currently constitute some of the best models for representations learned via hierarchical processing in the human brain. In medical imaging, these models have shown human-level performance and even higher in the early diagnosis of a wide range of diseases. However, the goal is often not only to accurately predict group membership or diagnose but also to provide explanations that support the model decision in a context that a human can readily interpret. The limited transparency has hindered the adoption of DNN algorithms across many domains. Numerous explainable artificial intelligence (XAI) techniques have been developed to peer inside the "black box" and make sense of DNN models, taking somewhat divergent approaches. Here, we suggest that these methods may be considered in light of the interpretation goal, including functional or mechanistic interpretations, developing archetypal class instances, or assessing the relevance of certain features or mappings on a trained model in a post-hoc capacity. We then focus on reviewing recent applications of post-hoc relevance techniques as applied to neuroimaging data. Moreover, this article suggests a method for comparing the reliability of XAI methods, especially in deep neural networks, along with their advantages and pitfalls.

Decoding Task-Based fMRI Data with Graph Neural Networks, Considering Individual Differences

Task fMRI provides an opportunity to analyze the working mechanisms of the human brain during specific experimental paradigms. Deep learning models have increasingly been applied for decoding and encoding purposes study to representations in task fMRI data. More recently, graph neural networks, or neural networks models designed to leverage the properties of graph representations, have recently shown promise in task fMRI decoding studies. Here, we propose an end-to-end graph convolutional network (GCN) framework with three convolutional layers to classify task fMRI data from the Human Connectome Project dataset. We compared the predictive performance of our GCN model across four of the most widely used node embedding algorithms-NetMF, RandNE, Node2Vec, and Walklets-to automatically extract the structural properties of the nodes in the functional graph. The empirical results indicated that our GCN framework accurately predicted individual differences (0.978 and 0.976) with the NetMF and RandNE embedding methods, respectively. Furthermore, to assess the effects of individual differences, we tested the classification performance of the model on sub-datasets divided according to gender and fluid intelligence. Experimental results indicated significant differences in the classification predictions of gender, but not high/low fluid intelligence fMRI data. Our experiments yielded promising results and demonstrated the superior ability of our GCN in modeling task fMRI data.

A lightweight learning-based decoding algorithm for intraneural vagus nerve activity classification in pigs

OBJECTIVE

Bioelectronic medicine is an emerging field that aims at developing closed-loop neuromodulation protocols for the autonomic nervous system (ANS) to treat a wide range of disorders. When designing a closed-loop protocol for real time modulation of the ANS, the computational execution time and the memory and power demands of the decoding step are important factors to consider. In the context of cardiovascular and respiratory diseases, these requirements may partially explain why closed-loop clinical neuromodulation protocols that adapt stimulation parameters on patient's clinical characteristics are currently missing.

APPROACH

Here, we developed a lightweight learning-based decoder for the classification of cardiovascular and respiratory functional challenges from neural signals acquired through intraneural electrodes implanted in the cervical vagus nerve (VN) of 5 anaesthetized pigs. Our algorithm is based on signal temporal windowing, 9 handcrafted features, and Random Forest (RF) model for classification. Temporal windowing ranging from 50 ms to 1 sec, compatible in duration with cardio-respiratory dynamics, was applied to the data in order to mimic a pseudo real-time scenario.

MAIN RESULTS

We were able to achieve high balanced accuracy (BA) values over the whole range of temporal windowing duration. We identified 500 ms as the optimal temporal windowing duration for both BA values and computational execution time processing, achieving more than 86% for BA and a computational execution time of only ∼6.8 ms. Our algorithm outperformed in terms of balanced accuracy and computational execution time a state of the art decoding algorithm tested on the same dataset [1]. We found that RF outperformed other machine learning models such as Support Vector Machines, K-Nearest Neighbors, and Multi-Layer Perceptrons.

SIGNIFICANCE

Our approach could represent an important step towards the implementation of a closed-loop neuromodulation protocol relying on a single intraneural interface able to perform real-time decoding tasks and selective modulation of the VN.

Strategy to Predict High and Low Frequency Behaviors Using Triaxial Accelerometers in Grazing of Beef Cattle

Simple Summary

Monitoring animal activity in production systems is an important tool for obtaining information on health, production, and reproduction. In this study, we evaluated the use of accelerometers with different strategies to predict the grazing behavior of Nelore cattle. This research was conducted in an environment both more challenging and representative of the practices adopted in livestock production systems in Brazil. The results of this study showed that the use of the Random Forest algorithm, together with techniques for resampling the training data of the models, classified the studied behaviors with high accuracy, especially for important, and less frequent activities such as water consumption frequency.

Abstract

Knowledge of animal behavior can be indicative of the well-being, health, productivity, and reproduction of animals. The use of accelerometers to classify and predict animal behavior can be a tool for continuous animal monitoring. Therefore, the aim of this study was to provide strategies for predicting more and less frequent beef cattle grazing behaviors. The behavior activities observed were grazing, ruminating, idle, water consumption frequency (WCF), feeding (supplementation) and walking. Three Machine Learning algorithms: Random Forest (RF), Support Vector Machine (SVM) and Naïve Bayes Classifier (NBC) and two resample methods: under and over-sampling, were tested. Overall accuracy was higher for RF models trained with the over-sampled dataset. The greatest sensitivity (0.808) for the less frequent behavior (WCF) was observed in the RF algorithm trained with the under-sampled data. The SVM models only performed efficiently when classifying the most frequent behavior (idle). The greatest predictor in the NBC algorithm was for ruminating behavior, with the over-sampled training dataset. The results showed that the behaviors of the studied animals were classified with high accuracy and specificity when the RF algorithm trained with the resampling methods was used. Resampling training datasets is a strategy to be considered, especially when less frequent behaviors are of interest.

A Novel Intrusion Detection Approach Using Machine Learning Ensemble for IoT Environments

The Internet of Things (IoT) has gained significant importance due to its applicability in diverse environments. Another reason for the influence of the IoT is its use of a flexible and scalable framework. The extensive and diversified use of the IoT in the past few years has attracted cyber-criminals. They exploit the vulnerabilities of the open-source IoT framework due to the absentia of robust and standard security protocols, hence discouraging existing and potential stakeholders. The authors propose a binary classifier approach developed from a machine learning ensemble method to filter and dump malicious traffic to prevent malicious actors from accessing the IoT network and its peripherals. The gradient boosting machine (GBM) ensemble approach is used to train the binary classifier using pre-processed recorded data packets to detect the anomaly and prevent the IoT networks from zero-day attacks. The positive class performance metrics of the model resulted in an accuracy of 98.27%, a precision of 96.40%, and a recall of 95.70%. The simulation results prove the effectiveness of the proposed model against cyber threats, thus making it suitable for critical applications for the IoT.

Identification of Methamphetamine Abstainers by Resting-State Functional Magnetic Resonance Imaging

Methamphetamine (MA) can cause brain structural and functional impairment, but there are few studies on whether this difference will sustain on MA abstainers. The purpose of this study is to investigate the correlation of brain networks in MA abstainers. In this study, 47 people detoxified for at least 14 months and 44 normal people took a resting-state functional magnetic resonance imaging (RS-fMRI) scan. A dynamic (i.e., time-varying) functional connectivity (FC) is obtained by applying sliding windows in the time courses on the independent components (ICs). The windowed correlation data for each IC were then clustered by k-means. The number of subjects in each cluster was used as a new feature for individual identification. The results show that the classifier achieved satisfactory performance (82.3% accuracy, 77.7% specificity, and 85.7% sensitivity). We find that there are significant differences in the brain networks of MA abstainers and normal people in the time domain, but the spatial differences are not obvious. Most of the altered functional connections (time-varying) are identified to be located at dorsal default mode network. These results have shown that changes in the correlation of the time domain may play an important role in identifying MA abstainers. Therefore, our findings provide valuable insights in the identification of MA and elucidate the pathological mechanism of MA from a resting-state functional integration point of view.

Predicting postoperative surgical site infection with administrative data: a random forests algorithm

Background

Since primary data collection can be time-consuming and expensive, surgical site infections (SSIs) could ideally be monitored using routinely collected administrative data. We derived and internally validated efficient algorithms to identify SSIs within 30 days after surgery with health administrative data, using Machine Learning algorithms.

Methods

All patients enrolled in the National Surgical Quality Improvement Program from the Ottawa Hospital were linked to administrative datasets in Ontario, Canada. Machine Learning approaches, including a Random Forests algorithm and the high-performance logistic regression, were used to derive parsimonious models to predict SSI status. Finally, a risk score methodology was used to transform the final models into the risk score system. The SSI risk models were validated in the validation datasets.

Results

Of 14,351 patients, 795 (5.5%) had an SSI. First, separate predictive models were built for three distinct administrative datasets. The final model, including hospitalization diagnostic, physician diagnostic and procedure codes, demonstrated excellent discrimination (C statistics, 0.91, 95% CI, 0.90–0.92) and calibration (Hosmer-Lemeshow χ² statistics, 4.531, p = 0.402).

Conclusion

We demonstrated that health administrative data can be effectively used to identify SSIs. Machine learning algorithms have shown a high degree of accuracy in predicting postoperative SSIs and can integrate and utilize a large amount of administrative data. External validation of this model is required before it can be routinely used to identify SSIs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01369-9.

A Review on Lung Cancer Diagnosis Using Data Mining Algorithms

Due to the serious consequences of lung cancer, medical associations use computer-aided diagnostic procedures to diagnose this disease more accurately. Despite the damaging effects of lung cancer on the body, the lifetime of cancer patients can be extended by early diagnosis. Data mining techniques are practical in diagnosing lung cancer in its first stages. This paper surveys a number of leading data mining-based cancer diagnosis approaches. Moreover, this review draws a comparison between data mining approaches in terms of selection criteria and presents the advantages and disadvantages of each method.

Predicting intraocular pressure using systemic variables or fundus photography with deep learning in a health examination cohort

The purpose of the current study was to predict intraocular pressure (IOP) using color fundus photography with a deep learning (DL) model, or, systemic variables with a multivariate linear regression model (MLM), along with least absolute shrinkage and selection operator regression (LASSO), support vector machine (SVM), and Random Forest: (RF). Training dataset included 3883 examinations from 3883 eyes of 1945 subjects and testing dataset 289 examinations from 289 eyes from 146 subjects. With the training dataset, MLM was constructed to predict IOP using 35 systemic variables and 25 blood measurements. A DL model was developed to predict IOP from color fundus photographs. The prediction accuracy of each model was evaluated through the absolute error and the marginal R-squared (mR2), using the testing dataset. The mean absolute error with MLM was 2.29 mmHg, which was significantly smaller than that with DL (2.70 dB). The mR2 with MLM was 0.15, whereas that with DL was 0.0066. The mean absolute error (between 2.24 and 2.30 mmHg) and mR2 (between 0.11 and 0.15) with LASSO, SVM and RF were similar to or poorer than MLM. A DL model to predict IOP using color fundus photography proved far less accurate than MLM using systemic variables.

Temporal embedding and spatiotemporal feature selection boost multi-voxel pattern analysis decoding accuracy

BACKGROUND

In fMRI decoding, temporal embedding of spatial features of the brain allows the incorporation of brain activity dynamics into the multivariate pattern classification process, and provides enriched information about stimulus-specific response patterns and potentially improved prediction accuracy.

NEW METHOD

This study investigates the possibility of enhancing the classification performance by exploring temporal embedding, to identify the optimum combination of spatiotemporal features based on their classification performance. We investigated the importance of spatiotemporal feature selection using a slow event-related design adapted from the classic Haxby study (Haxby et al., 2001). Data were collected using a multiband fMRI sequence with temporal resolution of 0.568 s.

COMPARISON WITH EXISTING METHODS

A wide range of spatiotemporal observations were created as various combinations of spatiotemporal features. Using both random forest, and support vector machine, classifiers prediction accuracies for these combinations were then compared with the single spatial multivariate pattern approach that uses only a single temporal observation.

RESULTS

Our findings showed that, on average, spatiotemporal feature selection improved prediction accuracy. Moreover, the random forest algorithm outperformed the support vector machine and benefitted from temporal information to a greater extent.

CONCLUSIONS

As expected, the most influential temporal durations were found to be around the peak of the hemodynamic response function, a few seconds after the stimuli onset until -4 s after the peak of the hemodynamic response function. The superiority of spatiotemporal feature selection over single time-point spatial approaches invites future work to design optimal approaches that incorporate spatiotemporal dependencies into feature selection for decoding.

Machine Learning in Neuroimaging: A New Approach to Understand Acupuncture for Neuroplasticity

The effects of acupuncture facilitating neural plasticity for treating diseases have been identified by clinical and experimental studies. In the last two decades, the application of neuroimaging techniques in acupuncture research provided visualized evidence for acupuncture promoting neuroplasticity. Recently, the integration of machine learning (ML) and neuroimaging techniques becomes a focus in neuroscience and brings a new and promising approach to understand the facilitation of acupuncture on neuroplasticity at the individual level. This review is aimed at providing an overview of this rapidly growing field by introducing the commonly used ML algorithms in neuroimaging studies briefly and analyzing the characteristics of the acupuncture studies based on ML and neuroimaging, so as to provide references for future research.

Feasibility of Training a Random Forest Model With Incomplete User-Specific Data for Devising a Control Strategy for Active Biomimetic Ankle

Intelligent control strategies for active biomimetic prostheses could exploit the inter-joint coordination of limbs in human gait in order to mimic the functioning of a biological joint. A machine learning regression model could be employed to learn an input-output relationship between the coordinated limb motion in human gait and predict the motion of a particular limb/joint given the motion of other limbs/joints. Such a model could be potentially used as a controller for an intelligent prosthesis which aims to restore the functioning similar to an intact biological joint. For this, the model needs to be tailored for each user by learning the gait pattern specific to the user. The challenge of training such machine learning regression models in prosthetic control is that, the desired reference output cannot be obtained from an amputee due to the missing limb. In this study, we investigate the feasibility of using two different methods for training a random forest algorithm using incomplete amputee-specific data to predict the ankle kinematics and dynamics from hip, knee, and shank kinematics. First is an inter-subject approach which learns a generalized input-output relationship from a group of able-bodied individuals and then applies this generalized relationship to amputees. Second is a subject-specific approach which maps the amputee's inputs to a desired normative reference output calculated from able-bodied individuals. The subject-specific model outperformed the inter-subject model in predicting the ankle angle and moment in most cases and can be potentially used for devising a control strategy for an intelligent biomimetic ankle.

Tools of the Trade Multivoxel pattern analysis in fMRI: a practical introduction for social and affective neuroscientists

The family of neuroimaging analytical techniques known as multivoxel pattern analysis (MVPA) has dramatically increased in popularity over the past decade, particularly in social and affective neuroscience research using functional magnetic resonance imaging (fMRI). MVPA examines patterns of neural responses, rather than analyzing single voxel- or region-based values, as is customary in conventional univariate analyses. Here, we provide a practical introduction to MVPA and its most popular variants (namely, representational similarity analysis (RSA) and decoding analyses, such as classification using machine learning) for social and affective neuroscientists of all levels, particularly those new to such methods. We discuss how MVPA differs from traditional mass-univariate analyses, the benefits MVPA offers to social neuroscientists, experimental design and analysis considerations, step-by-step instructions for how to implement specific analyses in one's own dataset and issues that are currently facing research using MVPA methods.

Artificial intelligence in drug combination therapy

Currently, the development of medicines for complex diseases requires the development of combination drug therapies. It is necessary because in many cases, one drug cannot target all necessary points of intervention. For example, in cancer therapy, a physician often meets a patient having a genomic profile including more than five molecular aberrations. Drug combination therapy has been an area of interest for a while, for example the classical work of Loewe devoted to the synergism of drugs was published in 1928-and it is still used in calculations for optimal drug combinations. More recently, over the past several years, there has been an explosion in the available information related to the properties of drugs and the biomedical parameters of patients. For the drugs, hundreds of 2D and 3D molecular descriptors for medicines are now available, while for patients, large data sets related to genetic/proteomic and metabolomics profiles of the patients are now available, as well as the more traditional data relating to the histology, history of treatments, pretreatment state of the organism, etc. Moreover, during disease progression, the genetic profile can change. Thus, the ability to optimize drug combinations for each patient is rapidly moving beyond the comprehension and capabilities of an individual physician. This is the reason, that biomedical informatics methods have been developed and one of the more promising directions in this field is the application of artificial intelligence (AI). In this review, we discuss several AI methods that have been successfully implemented in several instances of combination drug therapy from HIV, hypertension, infectious diseases to cancer. The data clearly show that the combination of rule-based expert systems with machine learning algorithms may be promising direction in this field.

Detection of Flares by Decrease in Physical Activity, Collected Using Wearable Activity Trackers in Rheumatoid Arthritis or Axial Spondyloarthritis: An Application of Machine Learning Analyses in Rheumatology

OBJECTIVE

Flares in rheumatoid arthritis (RA) and axial spondyloarthritis (SpA) may influence physical activity. The aim of this study was to assess longitudinally the association between patient-reported flares and activity-tracker-provided steps per minute, using machine learning.

METHODS

This prospective observational study (ActConnect) included patients with definite RA or axial SpA. For a 3-month time period, physical activity was assessed continuously by number of steps/minute, using a consumer grade activity tracker, and flares were self-assessed weekly. Machine-learning techniques were applied to the data set. After intrapatient normalization of the physical activity data, multiclass Bayesian methods were used to calculate sensitivities, specificities, and predictive values of the machine-generated models of physical activity in order to predict patient-reported flares.

RESULTS

Overall, 155 patients (1,339 weekly flare assessments and 224,952 hours of physical activity assessments) were analyzed. The mean ± SD age for patients with RA (n = 82) was 48.9 ± 12.6 years and was 41.2 ± 10.3 years for those with axial SpA (n = 73). The mean ± SD disease duration was 10.5 ± 8.8 years for patients with RA and 10.8 ± 9.1 years for those with axial SpA. Fourteen patients with RA (17.1%) and 41 patients with axial SpA (56.2%) were male. Disease was well-controlled (Disease Activity Score in 28 joints mean ± SD 2.2 ± 1.2; Bath Ankylosing Spondylitis Disease Activity Index score mean ± SD 3.1 ± 2.0), but flares were frequent (22.7% of all weekly assessments). The model generated by machine learning performed well against patient-reported flares (mean sensitivity 96% [95% confidence interval (95% CI) 94-97%], mean specificity 97% [95% CI 96-97%], mean positive predictive value 91% [95% CI 88-96%], and negative predictive value 99% [95% CI 98-100%]). Sensitivity analyses were confirmatory.

CONCLUSION

Although these pilot findings will have to be confirmed, the correct detection of flares by machine-learning processing of activity tracker data provides a framework for future studies of remote-control monitoring of disease activity, with great precision and minimal patient burden.

Fungal Traits Important for Soil Aggregation

Soil structure, the complex arrangement of soil into aggregates and pore spaces, is a key feature of soils and soil biota. Among them, filamentous saprobic fungi have well-documented effects on soil aggregation. However, it is unclear what properties, or traits, determine the overall positive effect of fungi on soil aggregation. To achieve progress, it would be helpful to systematically investigate a broad suite of fungal species for their trait expression and the relation of these traits to soil aggregation. Here, we apply a trait-based approach to a set of 15 traits measured under standardized conditions on 31 fungal strains including Ascomycota, Basidiomycota, and Mucoromycota, all isolated from the same soil. We find large differences among these fungi in their ability to aggregate soil, including neutral to positive effects, and we document large differences in trait expression among strains. We identify biomass density, i.e., the density with which a mycelium grows (positive effects), leucine aminopeptidase activity (negative effects) and phylogeny as important factors explaining differences in soil aggregate formation (SAF) among fungal strains; importantly, growth rate was not among the important traits. Our results point to a typical suite of traits characterizing fungi that are good soil aggregators, and our findings illustrate the power of employing a trait-based approach to unravel biological mechanisms underpinning soil aggregation. Such an approach could now be extended also to other soil biota groups. In an applied context of restoration and agriculture, such trait information can inform management, for example to prioritize practices that favor the expression of more desirable fungal traits.

Evaluation of machine learning algorithms performance for the prediction of early multiple sclerosis from resting-state FMRI connectivity data

Machine Learning application on clinical data in order to support diagnosis and prognostic evaluation arouses growing interest in scientific community. However, choice of right algorithm to use was fundamental to perform reliable and robust classification. Our study aimed to explore if different kinds of Machine Learning technique could be effective to support early diagnosis of Multiple Sclerosis and which of them presented best performance in distinguishing Multiple Sclerosis patients from control subjects. We selected following algorithms: Random Forest, Support Vector Machine, Naïve-Bayes, K-nearest-neighbor and Artificial Neural Network. We applied the Independent Component Analysis to resting-state functional-MRI sequence to identify brain networks. We found 15 networks, from which we extracted the mean signals used into classification. We performed feature selection tasks in all algorithms to obtain the most important variables. We showed that best discriminant network between controls and early Multiple Sclerosis, was the sensori-motor I, according to early manifestation of motor/sensorial deficits in Multiple Sclerosis. Moreover, in classification performance, Random Forest and Support Vector Machine showed same 5-fold cross-validation accuracies (85.7%) using only this network, resulting to be best approaches. We believe that these findings could represent encouraging step toward the translation to clinical diagnosis and prognosis.

Relationship between Vision-Related Quality of Life and Central 10° of the Binocular Integrated Visual Field in Advanced Glaucoma

To investigate the relationships between sensitivity loss in various subfields of the central 10° of the binocular integrated visual field (IVF) and vision-related quality of life (VRQoL) in 172 patients with advanced glaucoma. Using the Random Forest algorithm, which controls for inter-correlations among various subfields of the IVF, we analysed the relationships among the Rasch analysis-derived person ability index (RADPAI), age, best-corrected visual acuity (BCVA), mean total deviations (mTDs) of eight quadrant subfields in the IVF measured with the Humphrey Field Analyzer (HFA) 10-2 program (10-2 IVF), and mTDs of the upper/lower hemifields in the IVF measured with the HFA 24-2 program (24-2 IVF). Significant contributors to RADPAIs were as follows: the inner and outer lower-right quadrants of the 10-2 IVF contributed to the dining and total tasks; the lower-left quadrant of the 10-2 IVF contributed to the walking, going out and total tasks; the lower hemifield of the 24-2 IVF contributed to the walking, going out, dining, miscellaneous and total tasks; and BCVA contributed more to the letter, sentence, dressing and miscellaneous tasks than to others. The impact of damage in different 10-2 IVF subfields differed significantly across daily tasks in patients with advanced glaucoma.

On the Possibility of Predicting Glycaemia 'On the Fly' with Constrained IoT Devices in Type 1 Diabetes Mellitus Patients

Type 1 Diabetes Mellitus (DM1) patients are used to checking their blood glucose levels several times per day through finger sticks and, by subjectively handling this information, to try to predict their future glycaemia in order to choose a proper strategy to keep their glucose levels under control, in terms of insulin dosages and other factors. However, recent Internet of Things (IoT) devices and novel biosensors have allowed the continuous collection of the value of the glucose level by means of Continuous Glucose Monitoring (CGM) so that, with the proper Machine Learning (ML) algorithms, glucose evolution can be modeled, thus permitting a forecast of this variable. On the other hand, glycaemia dynamics require that such a model be user-centric and should be recalculated continuously in order to reflect the exact status of the patient, i.e., an ‘on-the-fly’ approach. In order to avoid, for example, the risk of being disconnected from the Internet, it would be ideal if this task could be performed locally in constrained devices like smartphones, but this would only be feasible if the execution times were fast enough. Therefore, in order to analyze if such a possibility is viable or not, an extensive, passive, CGM study has been carried out with 25 DM1 patients in order to build a solid dataset. Then, some well-known univariate algorithms have been executed in a desktop computer (as a reference) and two constrained devices: a smartphone and a Raspberry Pi, taking into account only past glycaemia data to forecast glucose levels. The results indicate that it is possible to forecast, in a smartphone, a 15-min horizon with a Root Mean Squared Error (RMSE) of 11.65 mg/dL in just 16.15 s, employing a 10-min sampling of the past 6 h of data and the Random Forest algorithm. With the Raspberry Pi, the computational effort increases to 56.49 s assuming the previously mentioned parameters, but this can be improved to 34.89 s if Support Vector Machines are applied, achieving in this case an RMSE of 19.90 mg/dL. Thus, this paper concludes that local on-the-fly forecasting of glycaemia would be affordable with constrained devices.

Application of Graph Theory for Identifying Connectivity Patterns in Human Brain Networks: A Systematic Review

Background: Analysis of the human connectome using functional magnetic resonance imaging (fMRI) started in the mid-1990s and attracted increasing attention in attempts to discover the neural underpinnings of human cognition and neurological disorders. In general, brain connectivity patterns from fMRI data are classified as statistical dependencies (functional connectivity) or causal interactions (effective connectivity) among various neural units. Computational methods, especially graph theory-based methods, have recently played a significant role in understanding brain connectivity architecture. Objectives: Thanks to the emergence of graph theoretical analysis, the main purpose of the current paper is to systematically review how brain properties can emerge through the interactions of distinct neuronal units in various cognitive and neurological applications using fMRI. Moreover, this article provides an overview of the existing functional and effective connectivity methods used to construct the brain network, along with their advantages and pitfalls. Methods: In this systematic review, the databases Science Direct, Scopus, arXiv, Google Scholar, IEEE Xplore, PsycINFO, PubMed, and SpringerLink are employed for exploring the evolution of computational methods in human brain connectivity from 1990 to the present, focusing on graph theory. The Cochrane Collaboration's tool was used to assess the risk of bias in individual studies. Results: Our results show that graph theory and its implications in cognitive neuroscience have attracted the attention of researchers since 2009 (as the Human Connectome Project launched), because of their prominent capability in characterizing the behavior of complex brain systems. Although graph theoretical approach can be generally applied to either functional or effective connectivity patterns during rest or task performance, to date, most articles have focused on the resting-state functional connectivity. Conclusions: This review provides an insight into how to utilize graph theoretical measures to make neurobiological inferences regarding the mechanisms underlying human cognition and behavior as well as different brain disorders.

Application of Graph Theory for Identifying Connectivity Patterns in Human Brain Networks: A Systematic Review

Background: Analysis of the human connectome using functional magnetic resonance imaging (fMRI) started in the mid-1990s and attracted increasing attention in attempts to discover the neural underpinnings of human cognition and neurological disorders. In general, brain connectivity patterns from fMRI data are classified as statistical dependencies (functional connectivity) or causal interactions (effective connectivity) among various neural units. Computational methods, especially graph theory-based methods, have recently played a significant role in understanding brain connectivity architecture. Objectives: Thanks to the emergence of graph theoretical analysis, the main purpose of the current paper is to systematically review how brain properties can emerge through the interactions of distinct neuronal units in various cognitive and neurological applications using fMRI. Moreover, this article provides an overview of the existing functional and effective connectivity methods used to construct the brain network, along with their advantages and pitfalls. Methods: In this systematic review, the databases Science Direct, Scopus, arXiv, Google Scholar, IEEE Xplore, PsycINFO, PubMed, and SpringerLink are employed for exploring the evolution of computational methods in human brain connectivity from 1990 to the present, focusing on graph theory. The Cochrane Collaboration's tool was used to assess the risk of bias in individual studies. Results: Our results show that graph theory and its implications in cognitive neuroscience have attracted the attention of researchers since 2009 (as the Human Connectome Project launched), because of their prominent capability in characterizing the behavior of complex brain systems. Although graph theoretical approach can be generally applied to either functional or effective connectivity patterns during rest or task performance, to date, most articles have focused on the resting-state functional connectivity. Conclusions: This review provides an insight into how to utilize graph theoretical measures to make neurobiological inferences regarding the mechanisms underlying human cognition and behavior as well as different brain disorders.

Individual differences in rate of acquiring stable neural representations of tasks in fMRI

Task-related functional magnetic resonance imaging (fMRI) is a widely-used tool for studying the neural processing correlates of human behavior in both healthy and clinical populations. There is growing interest in mapping individual differences in fMRI task behavior and neural responses. By utilizing neuroadaptive task designs accounting for such individual differences, task durations can be personalized to potentially optimize neuroimaging study outcomes (e.g., classification of task-related brain states). To test this hypothesis, we first retrospectively tracked the volume-by-volume changes of beta weights generated from general linear models (GLM) for 67 adult subjects performing a stop-signal task (SST). We then modeled the convergence of the volume-by-volume changes of beta weights according to their exponential decay (ED) in units of half-life. Our results showed significant differences in beta weight convergence estimates of optimal stopping times (OSTs) between go following successful stop trials and failed stop trials for both cocaine dependent (CD) and control group (Con), and between go following successful stop trials and go following failed stop trials for Con group. Further, we implemented support vector machine (SVM) classification for 67 CD/Con labeled subjects and compared the classification accuracies of fMRI-based features derived from (1) the full fMRI task versus (2) the fMRI task truncated to multiples of the unit of half-life. Among the computed binary classification accuracies, two types of task durations based on 2 half-lives significantly outperformed the accuracies using fully acquired trials, supporting this length as the OST for the SST. In conclusion, we demonstrate the potential of a neuroadaptive task design that can be widely applied to personalizing other task-based fMRI experiments in either dynamic real-time fMRI applications or within fMRI preprocessing pipelines.

PredPsych: A toolbox for predictive machine learning-based approach in experimental psychology research

Recent years have seen an increased interest in machine learning-based predictive methods for analyzing quantitative behavioral data in experimental psychology. While these methods can achieve relatively greater sensitivity compared to conventional univariate techniques, they still lack an established and accessible implementation. The aim of current work was to build an open-source R toolbox - "PredPsych" - that could make these methods readily available to all psychologists. PredPsych is a user-friendly, R toolbox based on machine-learning predictive algorithms. In this paper, we present the framework of PredPsych via the analysis of a recently published multiple-subject motion capture dataset. In addition, we discuss examples of possible research questions that can be addressed with the machine-learning algorithms implemented in PredPsych and cannot be easily addressed with univariate statistical analysis. We anticipate that PredPsych will be of use to researchers with limited programming experience not only in the field of psychology, but also in that of clinical neuroscience, enabling computational assessment of putative bio-behavioral markers for both prognosis and diagnosis.

Root traits are more than analogues of leaf traits: the case for diaspore mass

Root traits are often thought to be analogues of leaf traits along the plant economics spectrum. But evolutionary pressures have most likely shaped above- and belowground patterns differentially. Here, we aimed to identify the most important aboveground traits for explaining root traits without an a priori focus on known concepts. We measured morphological root traits in a glasshouse experiment on 141 common Central European grassland species. Using random forest algorithms, we built predictive models of six root traits from 97 aboveground morphological, ecological and life history traits. Root tissue density was best predicted by leaf dry matter content, whereas traits related to root fineness were best predicted by diaspore mass: the heavier the diaspore, the coarser the root system. Specific leaf area (SLA) was not an important predictor for any of the root traits. This study confirms the hypothesis that root traits are more than analogues of leaf traits within a plant economics spectrum. The results reveal a novel ecological pattern and highlight the power of root data to close important knowledge gaps in trait-based ecology.

Ten Key Observations on the Analysis of Resting-state Functional MR Imaging Data Using Independent Component Analysis

For more than 20 years, the powerful, flexible family of independent component analysis (ICA) techniques has been used to examine spatial, temporal, and subject variation in functional magnetic resonance (fMR) imaging data. This article provides an overview of 10 key principles in the basic and advanced application of ICA to resting-state fMR imaging. ICA's core advantages include robustness to artifact; false-positives and autocorrelation; adaptability to variant study designs; agnosticism to the temporal evolution of fMR imaging signals; and ability to extract, identify, and analyze neural networks. ICA remains in the vanguard of fMRI methods development.

On the use of on-cow accelerometers for the classification of behaviours in dairy barns

Analysing behaviours can provide insight into the health and overall well-being of dairy cows. Automatic monitoring systems using e.g., accelerometers are becoming increasingly important to accurately quantify cows' behaviours as the herd size increases. The aim of this study is to automatically classify cows' behaviours by comparing leg- and neck-mounted accelerometers, and to study the effect of the sampling rate and the number of accelerometer axes logged on the classification performances. Lying, standing, and feeding behaviours of 16 different lactating dairy cows were logged for 6h with 3D-accelerometers. The behaviours were simultaneously recorded using visual observation and video recordings as a reference. Different features were extracted from the raw data and machine learning algorithms were used for the classification. The classification models using combined data of the neck- and the leg-mounted accelerometers have classified the three behaviours with high precision (80-99%) and sensitivity (87-99%). For the leg-mounted accelerometer, lying behaviour was classified with high precision (99%) and sensitivity (98%). Feeding was classified more accurately by the neck-mounted versus the leg-mounted accelerometer (precision 92% versus 80%; sensitivity 97% versus 88%). Standing was the most difficult behaviour to classify when only one accelerometer was used. In addition, the classification performances were not highly influenced when only X, X and Z, or Z and Y axes were used for the classification instead of three axes, especially for the neck-mounted accelerometer. Moreover, the accuracy of the models decreased with about 20% when the sampling rate was decreased from 1Hz to 0.05Hz.

Ten Key Observations on the Analysis of Resting-state Functional MR Imaging Data Using Independent Component Analysis

For more than 20 years, the powerful, flexible family of independent component analysis (ICA) techniques has been used to examine spatial, temporal, and subject variation in functional magnetic resonance (fMR) imaging data. This article provides an overview of 10 key principles in the basic and advanced application of ICA to resting-state fMR imaging. ICA's core advantages include robustness to artifact; false-positives and autocorrelation; adaptability to variant study designs; agnosticism to the temporal evolution of fMR imaging signals; and ability to extract, identify, and analyze neural networks. ICA remains in the vanguard of fMRI methods development.

Methods for cleaning the BOLD fMRI signal

Blood oxygen-level-dependent functional magnetic resonance imaging (BOLD fMRI) has rapidly become a popular technique for the investigation of brain function in healthy individuals, patients as well as in animal studies. However, the BOLD signal arises from a complex mixture of neuronal, metabolic and vascular processes, being therefore an indirect measure of neuronal activity, which is further severely corrupted by multiple non-neuronal fluctuations of instrumental, physiological or subject-specific origin. This review aims to provide a comprehensive summary of existing methods for cleaning the BOLD fMRI signal. The description is given from a methodological point of view, focusing on the operation of the different techniques in addition to pointing out the advantages and limitations in their application. Since motion-related and physiological noise fluctuations are two of the main noise components of the signal, techniques targeting their removal are primarily addressed, including both data-driven approaches and using external recordings. Data-driven approaches, which are less specific in the assumed model and can simultaneously reduce multiple noise fluctuations, are mainly based on data decomposition techniques such as principal and independent component analysis. Importantly, the usefulness of strategies that benefit from the information available in the phase component of the signal, or in multiple signal echoes is also highlighted. The use of global signal regression for denoising is also addressed. Finally, practical recommendations regarding the optimization of the preprocessing pipeline for the purpose of denoising and future venues of research are indicated. Through the review, we summarize the importance of signal denoising as an essential step in the analysis pipeline of task-based and resting state fMRI studies.

Decoding the encoding of functional brain networks: An fMRI classification comparison of non-negative matrix factorization (NMF), independent component analysis (ICA), and sparse coding algorithms

BACKGROUND

Brain networks in fMRI are typically identified using spatial independent component analysis (ICA), yet other mathematical constraints provide alternate biologically-plausible frameworks for generating brain networks. Non-negative matrix factorization (NMF) would suppress negative BOLD signal by enforcing positivity. Spatial sparse coding algorithms (L1 Regularized Learning and K-SVD) would impose local specialization and a discouragement of multitasking, where the total observed activity in a single voxel originates from a restricted number of possible brain networks.

NEW METHOD

The assumptions of independence, positivity, and sparsity to encode task-related brain networks are compared; the resulting brain networks within scan for different constraints are used as basis functions to encode observed functional activity. These encodings are then decoded using machine learning, by using the time series weights to predict within scan whether a subject is viewing a video, listening to an audio cue, or at rest, in 304 fMRI scans from 51 subjects.

RESULTS AND COMPARISON WITH EXISTING METHOD

The sparse coding algorithm of L1 Regularized Learning outperformed 4 variations of ICA (p<0.001) for predicting the task being performed within each scan using artifact-cleaned components. The NMF algorithms, which suppressed negative BOLD signal, had the poorest accuracy compared to the ICA and sparse coding algorithms. Holding constant the effect of the extraction algorithm, encodings using sparser spatial networks (containing more zero-valued voxels) had higher classification accuracy (p<0.001). Lower classification accuracy occurred when the extracted spatial maps contained more CSF regions (p<0.001).

CONCLUSION

The success of sparse coding algorithms suggests that algorithms which enforce sparsity, discourage multitasking, and promote local specialization may capture better the underlying source processes than those which allow inexhaustible local processes such as ICA. Negative BOLD signal may capture task-related activations.

Validating the Usefulness of the "Random Forests" Classifier to Diagnose Early Glaucoma With Optical Coherence Tomography

PURPOSE

To validate the usefulness of the "Random Forests" classifier to diagnose early glaucoma with spectral-domain optical coherence tomography (SDOCT).

METHODS

design: Comparison of diagnostic algorithms.

SETTING

Multiple institutional practices.

STUDY PARTICIPANTS

Training dataset included 94 eyes of 94 open-angle glaucoma (OAG) patients and 84 eyes of 84 normal subjects and testing dataset included 114 eyes of 114 OAG patients and 82 eyes of 82 normal subjects. In both groups, OAG eyes with mean deviation (MD) values better than -5.0 dB were included.

OBSERVATION PROCEDURE

Using the training dataset, classifiers were built to discriminate between glaucoma and normal eyes using 84 OCT measurements using the Random Forests method, multiple logistic regression models based on backward or bidirectional stepwise model selection, a least absolute shrinkage and selection operator regression (LASSO) model, and a Ridge regression model.

MAIN OUTCOME MEASURES

Diagnostic accuracy.

RESULTS

With the testing data, the area under the receiver operating characteristic curve (AROC) with the Random Forests method (93.0%) was significantly (P < .05) larger than those with other models of the stepwise model selections (71.9%), LASSO model (89.6%), and Ridge model (89.2%).

CONCLUSION

It is useful to analyze multiple SDOCT parameters concurrently using the Random Forests method to diagnose glaucoma in early stages.

Diffusion Tensor Imaging of TBI: Potentials and Challenges

Neuroimaging plays a critical role in the setting in traumatic brain injury (TBI). Diffusion tensor imaging (DTI) is an advanced magnetic resonance imaging technique that is capable of providing rich information on the brain's neuroanatomic connectome. The purpose of this article is to systematically review the role of DTI and advanced diffusion techniques in the setting of TBI, including diffusion kurtosis imaging (DKI), neurite orientation dispersion and density imaging, diffusion spectrum imaging, and q-ball imaging. We discuss clinical applications of DTI and review the DTI literature as it pertains to TBI. Despite the continued advancements in DTI and related diffusion techniques over the past 20 years, DTI techniques are sensitive for TBI at the group level only and there is insufficient evidence that DTI plays a role at the individual level. We conclude by discussing future directions in DTI research in TBI including the role of machine learning in the pattern classification of TBI.

Dimensionality of ICA in resting-state fMRI investigated by feature optimized classification of independent components with SVM

Different machine learning algorithms have recently been used for assisting automated classification of independent component analysis (ICA) results from resting-state fMRI data. The success of this approach relies on identification of artifact components and meaningful functional networks. A limiting factor of ICA is the uncertainty of the number of independent components (NIC). We aim to develop a framework based on support vector machines (SVM) and optimized feature-selection for automated classification of independent components (ICs) and use the framework to investigate the effects of input NIC on the ICA results. Seven different resting-state fMRI datasets were studied. 18 features were devised by mimicking the empirical criteria for manual evaluation. The five most significant (p < 0.01) features were identified by general linear modeling and used to generate a classification model for the framework. This feature-optimized classification of ICs with SVM (FOCIS) framework was used to classify both group and single subject ICA results. The classification results obtained using FOCIS and previously published FSL-FIX were compared against manually evaluated results. On average the false negative rate in identifying artifact contaminated ICs for FOCIS and FSL-FIX were 98.27 and 92.34%, respectively. The number of artifact and functional network components increased almost linearly with the input NIC. Through tracking, we demonstrate that incrementing NIC affects most ICs when NIC < 33, whereas only a few limited ICs are affected by direct splitting when NIC is incremented beyond NIC > 40. For a given IC, its changes with increasing NIC are individually specific irrespective whether the component is a potential resting-state functional network or an artifact component. Using FOCIS, we investigated experimentally the ICA dimensionality of resting-state fMRI datasets and found that the input NIC can critically affect the ICA results of resting-state fMRI data.

Multimodal neuroimaging based classification of autism spectrum disorder using anatomical, neurochemical, and white matter correlates

Neuroimaging techniques, such as fMRI, structural MRI, diffusion tensor imaging (DTI), and proton magnetic resonance spectroscopy (1H-MRS) have uncovered evidence for widespread functional and anatomical brain abnormalities in autism spectrum disorder (ASD) suggesting it to be a system-wide neural systems disorder. Nevertheless, most previous studies have focused on examining one index of neuropathology through a single neuroimaging modality, and seldom using multiple modalities to examine the same cohort of individuals. The current study aims to bring together multiple brain imaging modalities (structural MRI, DTI, and 1H-MRS) to investigate the neural architecture in the same set of individuals (19 high-functioning adults with ASD and 18 typically developing (TD) peers). Morphometry analysis revealed increased cortical thickness in ASD participants, relative to typical controls, across the left cingulate, left pars opercularis of the inferior frontal gyrus, left inferior temporal cortex, and right precuneus, and reduced cortical thickness in right cuneus and right precentral gyrus. ASD adults also had reduced fractional anisotropy (FA) and increased radial diffusivity (RD) for two clusters on the forceps minor of the corpus callosum, revealed by DTI analyses. 1H-MRS results showed a reduction in the N-acetylaspartate/Creatine ratio in dorsal anterior cingulate cortex (dACC) in ASD participants. A decision tree classification analysis across the three modalities resulted in classification accuracy of 91.9% with FA, RD, and cortical thickness as key predictors. Examining the same cohort of adults with ASD and their TD peers, this study found alterations in cortical thickness, white matter (WM) connectivity, and neurochemical concentration in ASD. These findings underscore the potential for multimodal imaging to better inform on the neural characteristics most relevant to the disorder.

A review of feature reduction techniques in neuroimaging

Machine learning techniques are increasingly being used in making relevant predictions and inferences on individual subjects neuroimaging scan data. Previous studies have mostly focused on categorical discrimination of patients and matched healthy controls and more recently, on prediction of individual continuous variables such as clinical scores or age. However, these studies are greatly hampered by the large number of predictor variables (voxels) and low observations (subjects) also known as the curse-of-dimensionality or small-n-large-p problem. As a result, feature reduction techniques such as feature subset selection and dimensionality reduction are used to remove redundant predictor variables and experimental noise, a process which mitigates the curse-of-dimensionality and small-n-large-p effects. Feature reduction is an essential step before training a machine learning model to avoid overfitting and therefore improving model prediction accuracy and generalization ability. In this review, we discuss feature reduction techniques used with machine learning in neuroimaging studies.

Bayesian networks in neuroscience: a survey

Bayesian networks are a type of probabilistic graphical models lie at the intersection between statistics and machine learning. They have been shown to be powerful tools to encode dependence relationships among the variables of a domain under uncertainty. Thanks to their generality, Bayesian networks can accommodate continuous and discrete variables, as well as temporal processes. In this paper we review Bayesian networks and how they can be learned automatically from data by means of structure learning algorithms. Also, we examine how a user can take advantage of these networks for reasoning by exact or approximate inference algorithms that propagate the given evidence through the graphical structure. Despite their applicability in many fields, they have been little used in neuroscience, where they have focused on specific problems, like functional connectivity analysis from neuroimaging data. Here we survey key research in neuroscience where Bayesian networks have been used with different aims: discover associations between variables, perform probabilistic reasoning over the model, and classify new observations with and without supervision. The networks are learned from data of any kind-morphological, electrophysiological, -omics and neuroimaging-, thereby broadening the scope-molecular, cellular, structural, functional, cognitive and medical- of the brain aspects to be studied.

Discriminating between glaucoma and normal eyes using optical coherence tomography and the 'Random Forests' classifier

PURPOSE

To diagnose glaucoma based on spectral domain optical coherence tomography (SD-OCT) measurements using the 'Random Forests' method.

METHODS

SD-OCT was conducted in 126 eyes of 126 open angle glaucoma (OAG) patients and 84 eyes of 84 normal subjects. The Random Forests method was then applied to discriminate between glaucoma and normal eyes using 151 OCT parameters including thickness measurements of circumpapillary retinal nerve fiber layer (cpRNFL), the macular RNFL (mRNFL) and the ganglion cell layer-inner plexiform layer combined (GCIPL). The area under the receiver operating characteristic curve (AROC) was calculated using the Random Forests method adopting leave-one-out cross validation. For comparison, AROCs were calculated based on each one of the 151 OCT parameters.

RESULTS

The AROC obtained with the Random Forests method was 98.5% [95% Confidence interval (CI): 97.1-99.9%], which was significantly larger than the AROCs derived from any single OCT parameter (maxima were: 92.8 [CI: 89.4-96.2] %, 94.3 [CI: 91.1-97.6] % and 91.8 [CI: 88.2-95.4] % for cpRNFL-, mRNFL- and GCIPL-related parameters, respectively; P<0.05, DeLong's method with Holm's correction for multiple comparisons). The partial AROC above specificity of 80%, for the Random Forests method was equal to 18.5 [CI: 16.8-19.6] %, which was also significantly larger than the AROCs of any single OCT parameter (P<0.05, Bootstrap method with Holm's correction for multiple comparisons).

CONCLUSIONS

The Random Forests method, analyzing multiple SD-OCT parameters concurrently, significantly improves the diagnosis of glaucoma compared with using any single SD-OCT measurement.

Evaluation of various machine learning methods to predict vision-related quality of life from visual field data and visual acuity in patients with glaucoma

BACKGROUND/AIMS

We investigated whether it was useful to use machine learning algorithms to predict patients' vision related quality of life (VRQoL) from visual field (VF) and visual acuity (VA).

METHODS

VRQoL was surveyed in 164 glaucomatous patients using the Sumi questionnaire. Their VRQoL score was predicted using machine learning algorithms (Random Forest, gradient boosting, support vector machine) based on total deviation (TD) values from integrated VF (IVF), VA, age and gender. For comparison, VRQoL score was predicted using standard linear regression with mean of IVF, TD values, and VA, and also the stepwise model selection by Akaike Information Criterion. Prediction error was calculated as root mean of the squared prediction error (RMSE) associated with the leave one out cross validation.

RESULTS

RMSEs associated with general VRQoL score were smaller for the machine learning algorithms (1.99 to 2.21) compared with the standard linear model and the stepwise model selection (2.35 to 3.20). A similar tendency was found in each individual VRQoL task score.

CONCLUSIONS

We found that it was advantageous to use machine learning methods to predict VRQoL accurately. These statistical methods could be used to help clinicians better understand patients' VRQoL without the need for extra tests other than standard VA and VF.

Cross-sectional study: Does combining optical coherence tomography measurements using the 'Random Forest' decision tree classifier improve the prediction of the presence of perimetric deterioration in glaucoma suspects?

OBJECTIVES

To develop a classifier to predict the presence of visual field (VF) deterioration in glaucoma suspects based on optical coherence tomography (OCT) measurements using the machine learning method known as the 'Random Forest' algorithm.

DESIGN

Case-control study.

PARTICIPANTS

293 eyes of 179 participants with open angle glaucoma (OAG) or suspected OAG.

INTERVENTIONS

Spectral domain OCT (Topcon 3D OCT-2000) and perimetry (Humphrey Field Analyser, 24-2 or 30-2 SITA standard) measurements were conducted in all of the participants. VF damage (Ocular Hypertension Treatment Study criteria (2002)) was used as a 'gold-standard' to classify glaucomatous eyes. The 'Random Forest' method was then used to analyse the relationship between the presence/absence of glaucomatous VF damage and the following variables: age, gender, right or left eye, axial length plus 237 different OCT measurements.

MAIN OUTCOME MEASURES

The area under the receiver operating characteristic curve (AROC) was then derived using the probability of glaucoma as suggested by the proportion of votes in the Random Forest classifier. For comparison, five AROCs were derived based on: (1) macular retinal nerve fibre layer (m-RNFL) alone; (2) circumpapillary (cp-RNFL) alone; (3) ganglion cell layer and inner plexiform layer (GCL+IPL) alone; (4) rim area alone and (5) a decision tree method using the same variables as the Random Forest algorithm.

RESULTS

The AROC from the combined Random Forest classifier (0.90) was significantly larger than the AROCs based on individual measurements of m-RNFL (0.86), cp-RNFL (0.77), GCL+IPL (0.80), rim area (0.78) and the decision tree method (0.75; p<0.05).

CONCLUSIONS

Evaluating OCT measurements using the Random Forest method provides an accurate prediction of the presence of perimetric deterioration in glaucoma suspects.

Single trial decoding of belief decision making from EEG and fMRI data using independent components features

The complex task of assessing the veracity of a statement is thought to activate uniquely distributed brain regions based on whether a subject believes or disbelieves a given assertion. In the current work, we present parallel machine learning methods for predicting a subject's decision response to a given propositional statement based on independent component (IC) features derived from EEG and fMRI data. Our results demonstrate that IC features outperformed features derived from event related spectral perturbations derived from any single spectral band, yet were similar to accuracy across all spectral bands combined. We compared our diagnostic IC spatial maps with our conventional general linear model (GLM) results, and found that informative ICs had significant spatial overlap with our GLM results, yet also revealed unique regions like amygdala that were not statistically significant in GLM analyses. Overall, these results suggest that ICs may yield a parsimonious feature set that can be used along with a decision tree structure for interpretation of features used in classifying complex cognitive processes such as belief and disbelief across both fMRI and EEG neuroimaging modalities.