Successful classification of cocaine dependence using brain imaging: a generalizable machine learning approach

Updated Perspectives on the Neurobiology of Substance Use Disorders Using Neuroimaging

Substance use problems impair social functioning, academic achievement, and employability. Psychological, biological, social, and environmental factors can contribute to substance use disorders. In recent years, neuroimaging breakthroughs have helped elucidate the mechanisms of substance misuse and its effects on the brain. Functional magnetic resonance imaging (MRI), positron emission tomography (PET), single-photon emission computed tomography (SPECT), and magnetic resonance spectroscopy (MRS) are all examples. Neuroimaging studies suggest substance misuse affects executive function, reward, memory, and stress systems. Recent neuroimaging research attempts have provided clinicians with improved tools to diagnose patients who misuse substances, comprehend the complicated neuroanatomy and neurobiology involved, and devise individually tailored and monitorable treatment regimens for individuals with substance use disorders. This review describes the most recent developments in drug misuse neuroimaging, including the neurobiology of substance use disorders, neuroimaging, and substance use disorders, established neuroimaging techniques, recent developments with established neuroimaging techniques and substance use disorders, and emerging clinical neuroimaging technology.

Sampling inequalities affect generalization of neuroimaging-based diagnostic classifiers in psychiatry

BACKGROUND

The development of machine learning models for aiding in the diagnosis of mental disorder is recognized as a significant breakthrough in the field of psychiatry. However, clinical practice of such models remains a challenge, with poor generalizability being a major limitation.

METHODS

Here, we conducted a pre-registered meta-research assessment on neuroimaging-based models in the psychiatric literature, quantitatively examining global and regional sampling issues over recent decades, from a view that has been relatively underexplored. A total of 476 studies (n = 118,137) were included in the current assessment. Based on these findings, we built a comprehensive 5-star rating system to quantitatively evaluate the quality of existing machine learning models for psychiatric diagnoses.

RESULTS

A global sampling inequality in these models was revealed quantitatively (sampling Gini coefficient (G) = 0.81, p < .01), varying across different countries (regions) (e.g., China, G = 0.47; the USA, G = 0.58; Germany, G = 0.78; the UK, G = 0.87). Furthermore, the severity of this sampling inequality was significantly predicted by national economic levels (β =  - 2.75, p < .001, R²_adj = 0.40; r =  - .84, 95% CI: - .41 to - .97), and was plausibly predictable for model performance, with higher sampling inequality for reporting higher classification accuracy. Further analyses showed that lack of independent testing (84.24% of models, 95% CI: 81.0-87.5%), improper cross-validation (51.68% of models, 95% CI: 47.2-56.2%), and poor technical transparency (87.8% of models, 95% CI: 84.9-90.8%)/availability (80.88% of models, 95% CI: 77.3-84.4%) are prevailing in current diagnostic classifiers despite improvements over time. Relating to these observations, model performances were found decreased in studies with independent cross-country sampling validations (all p < .001, BF₁₀ > 15). In light of this, we proposed a purpose-built quantitative assessment checklist, which demonstrated that the overall ratings of these models increased by publication year but were negatively associated with model performance.

CONCLUSIONS

Together, improving sampling economic equality and hence the quality of machine learning models may be a crucial facet to plausibly translating neuroimaging-based diagnostic classifiers into clinical practice.

Classifying migraine using PET compressive big data analytics of brain's μ-opioid and D2/D3 dopamine neurotransmission

Introduction: Migraine is a common and debilitating pain disorder associated with dysfunction of the central nervous system. Advanced magnetic resonance imaging (MRI) studies have reported relevant pathophysiologic states in migraine. However, its molecular mechanistic processes are still poorly understood in vivo. This study examined migraine patients with a novel machine learning (ML) method based on their central μ-opioid and dopamine D2/D3 profiles, the most critical neurotransmitters in the brain for pain perception and its cognitive-motivational interface. Methods: We employed compressive Big Data Analytics (CBDA) to identify migraineurs and healthy controls (HC) in a large positron emission tomography (PET) dataset. 198 PET volumes were obtained from 38 migraineurs and 23 HC during rest and thermal pain challenge. 61 subjects were scanned with the selective μ-opioid receptor (μOR) radiotracer [11C]Carfentanil, and 22 with the selective dopamine D2/D3 receptor (DOR) radiotracer [11C]Raclopride. PET scans were recast into a 1D array of 510,340 voxels with spatial and intensity filtering of non-displaceable binding potential (BPND), representing the receptor availability level. We then performed data reduction and CBDA to power rank the predictive brain voxels. Results: CBDA classified migraineurs from HC with accuracy, sensitivity, and specificity above 90% for whole-brain and region-of-interest (ROI) analyses. The most predictive ROIs for μOR were the insula (anterior), thalamus (pulvinar, medial-dorsal, and ventral lateral/posterior nuclei), and the putamen. The latter, putamen (anterior), was also the most predictive for migraine regarding DOR D2/D3 BPND levels. Discussion: CBDA of endogenous μ-opioid and D2/D3 dopamine dysfunctions in the brain can accurately identify a migraine patient based on their receptor availability across key sensory, motor, and motivational processing regions. Our ML-based findings in the migraineur's brain neurotransmission partly explain the severe impact of migraine suffering and associated neuropsychiatric comorbidities.

Classification of FLT3 inhibitors and SAR analysis by machine learning methods

FMS-like tyrosine kinase 3 (FLT3) is a type III receptor tyrosine kinase, which is an important target for anti-cancer therapy. In this work, we conducted a structure-activity relationship (SAR) study on 3867 FLT3 inhibitors we collected. MACCS fingerprints, ECFP4 fingerprints, and TT fingerprints were used to represent the inhibitors in the dataset. A total of 36 classification models were built based on support vector machine (SVM), random forest (RF), eXtreme Gradient Boosting (XGBoost), and deep neural networks (DNN) algorithms. Model 3D_3 built by deep neural networks (DNN) and TT fingerprints performed best on the test set with the highest prediction accuracy of 85.83% and Matthews correlation coefficient (MCC) of 0.72 and also performed well on the external test set. In addition, we clustered 3867 inhibitors into 11 subsets by the K-Means algorithm to figure out the structural characteristics of the reported FLT3 inhibitors. Finally, we analyzed the SAR of FLT3 inhibitors by RF algorithm based on ECFP4 fingerprints. The results showed that 2-aminopyrimidine, 1-ethylpiperidine,2,4-bis(methylamino)pyrimidine, amino-aromatic heterocycle, [(2E)-but-2-enyl]dimethylamine, but-2-enyl, and alkynyl were typical fragments among highly active inhibitors. Besides, three scaffolds in Subset_A (Subset 4), Subset_B, and Subset_C showed a significant relationship to inhibition activity targeting FLT3.

Aberrant topology of white matter networks in patients with methamphetamine dependence and its application in support vector machine-based classification

Brain white matter (WM) networks have been widely studied in neuropsychiatric disorders. However, few studies have evaluated alterations in WM network topological organization in patients with methamphetamine (MA) dependence. Therefore, using machine learning classification methods to analyze WM network topological attributes may give new insights into patients with MA dependence. In the study, diffusion tensor imaging-based probabilistic tractography was used to map the weighted WM networks in 46 MA-dependent patients and 46 control subjects. Using graph-theoretical analyses, the global and regional topological attributes of WM networks for both groups were calculated and compared to determine inter-group differences using a permutation-based general linear model. In addition, the study used a support vector machine (SVM) learning approach to construct a classifier for discriminating subjects with MA dependence from control subjects. Relative to the control group, the MA-dependent group exhibited abnormal topological organization, as evidenced by decreased small-worldness and modularity, and increased nodal efficiency in the right medial superior temporal gyrus, right pallidum, and right ventromedial putamen; the MA-dependent group had the higher hubness scores in 25 regions, which were mainly located in the default mode network. An SVM trained with topological attributes achieved classification accuracy, sensitivity, specificity, and kappa values of 98.09% ± 2.59%, 98.24% ± 4.00%, 97.94% ± 4.26%, and 96.18% ± 5.19% for patients with MA dependence. Our results may suggest altered global WM structural networks in MA-dependent patients. Furthermore, the abnormal WM network topological attributes may provide promising features for the construction of high-efficacy classification models.

Invasive Computational Psychiatry

Computational psychiatry, a relatively new yet prolific field that aims to understand psychiatric disorders with formal theories about the brain, has seen tremendous growth in the past decade. Despite initial excitement, actual progress made by computational psychiatry seems stagnant. Meanwhile, understanding of the human brain has benefited tremendously from recent progress in intracranial neuroscience. Specifically, invasive techniques such as stereotactic electroencephalography, electrocorticography, and deep brain stimulation have provided a unique opportunity to precisely measure and causally modulate neurophysiological activity in the living human brain. In this review, we summarize progress and drawbacks in both computational psychiatry and invasive electrophysiology and propose that their combination presents a highly promising new direction-invasive computational psychiatry. The value of this approach is at least twofold. First, it advances our mechanistic understanding of the neural computations of mental states by providing a spatiotemporally precise depiction of neural activity that is traditionally unattainable using noninvasive techniques with human subjects. Second, it offers a direct and immediate way to modulate brain states through stimulation of algorithmically defined neural regions and circuits (i.e., algorithmic targeting), thus providing both causal and therapeutic insights. We then present depression as a use case where the combination of computational and invasive approaches has already shown initial success. We conclude by outlining future directions as a road map for this exciting new field as well as presenting cautions about issues such as ethical concerns and generalizability of findings.

An Interpretable and Predictive Connectivity-Based Neural Signature for Chronic Cannabis Use

BACKGROUND

Cannabis is one of the most widely used substances in the world, with usage trending upward in recent years. However, although the psychiatric burden associated with maladaptive cannabis use has been well established, reliable and interpretable biomarkers associated with chronic use remain elusive. In this study, we combine large-scale functional magnetic resonance imaging with machine learning and network analysis and develop an interpretable decoding model that offers both accurate prediction and novel insights into chronic cannabis use.

METHODS

Chronic cannabis users (n = 166) and nonusing healthy control subjects (n = 124) completed a cue-elicited craving task during functional magnetic resonance imaging. Linear machine learning methods were used to classify individuals into chronic users and nonusers based on whole-brain functional connectivity. Network analysis was used to identify the most predictive regions and communities.

RESULTS

We obtained high (∼80% out-of-sample) accuracy across 4 different classification models, demonstrating that task-evoked connectivity can successfully differentiate chronic cannabis users from nonusers. We also identified key predictive regions implicating motor, sensory, attention, and craving-related areas, as well as a core set of brain networks that contributed to successful classification. The most predictive networks also strongly correlated with cannabis craving within the chronic user group.

CONCLUSIONS

This novel approach produced a neural signature of chronic cannabis use that is both accurate in terms of out-of-sample prediction and interpretable in terms of predictive networks and their relation to cannabis craving.

Evaluation of Risk of Bias in Neuroimaging-Based Artificial Intelligence Models for Psychiatric Diagnosis: A Systematic Review

IMPORTANCE

Neuroimaging-based artificial intelligence (AI) diagnostic models have proliferated in psychiatry. However, their clinical applicability and reporting quality (ie, feasibility) for clinical practice have not been systematically evaluated.

OBJECTIVE

To systematically assess the risk of bias (ROB) and reporting quality of neuroimaging-based AI models for psychiatric diagnosis.

EVIDENCE REVIEW

PubMed was searched for peer-reviewed, full-length articles published between January 1, 1990, and March 16, 2022. Studies aimed at developing or validating neuroimaging-based AI models for clinical diagnosis of psychiatric disorders were included. Reference lists were further searched for suitable original studies. Data extraction followed the CHARMS (Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies) and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines. A closed-loop cross-sequential design was used for quality control. The PROBAST (Prediction Model Risk of Bias Assessment Tool) and modified CLEAR (Checklist for Evaluation of Image-Based Artificial Intelligence Reports) benchmarks were used to systematically evaluate ROB and reporting quality.

FINDINGS

A total of 517 studies presenting 555 AI models were included and evaluated. Of these models, 461 (83.1%; 95% CI, 80.0%-86.2%) were rated as having a high overall ROB based on the PROBAST. The ROB was particular high in the analysis domain, including inadequate sample size (398 of 555 models [71.7%; 95% CI, 68.0%-75.6%]), poor model performance examination (with 100% of models lacking calibration examination), and lack of handling data complexity (550 of 555 models [99.1%; 95% CI, 98.3%-99.9%]). None of the AI models was perceived to be applicable to clinical practices. Overall reporting completeness (ie, number of reported items/number of total items) for the AI models was 61.2% (95% CI, 60.6%-61.8%), and the completeness was poorest for the technical assessment domain with 39.9% (95% CI, 38.8%-41.1%).

CONCLUSIONS AND RELEVANCE

This systematic review found that the clinical applicability and feasibility of neuroimaging-based AI models for psychiatric diagnosis were challenged by a high ROB and poor reporting quality. Particularly in the analysis domain, ROB in AI diagnostic models should be addressed before clinical application.

Abnormal resting-state functional connectome in methamphetamine-dependent patients and its application in machine-learning-based classification

Brain resting-state functional connectivity (rsFC) has been widely analyzed in substance use disorders (SUDs), including methamphetamine (MA) dependence. Most of these studies utilized Pearson correlation analysis to assess rsFC, which cannot determine whether two brain regions are connected by direct or indirect pathways. Moreover, few studies have reported the application of rsFC-based graph theory in MA dependence. We evaluated alterations in Tikhonov regularization-based rsFC and rsFC-based topological attributes in 46 MA-dependent patients, as well as the correlations between topological attributes and clinical variables. Moreover, the topological attributes selected by least absolute shrinkage and selection operator (LASSO) were used to construct a support vector machine (SVM)-based classifier for MA dependence. The MA group presented a subnetwork with increased rsFC, indicating overactivation of the reward circuit that makes patients very sensitive to drug-related visual cues, and a subnetwork with decreased rsFC suggesting aberrant synchronized spontaneous activity in subregions within the orbitofrontal cortex (OFC) system. The MA group demonstrated a significantly decreased area under the curve (AUC) for the clustering coefficient (Cp) (Pperm &lt; 0.001), shortest path length (Lp) (Pperm = 0.007), modularity (Pperm = 0.006), and small-worldness (σ, Pperm = 0.004), as well as an increased AUC for global efficiency (E.glob) (Pperm = 0.009), network strength (Sp) (Pperm = 0.009), and small-worldness (ω, Pperm &lt; 0.001), implying a shift toward random networks. MA-related increased nodal efficiency (E.nodal) and altered betweenness centrality were also discovered in several brain regions. The AUC for ω was significantly positively associated with psychiatric symptoms. An SVM classifier trained by 36 features selected by LASSO from all topological attributes achieved excellent performance, cross-validated prediction area under the receiver operating characteristics curve, accuracy, sensitivity, specificity, and kappa of 99.03 ± 1.79, 94.00 ± 5.78, 93.46 ± 8.82, 94.52 ± 8.11, and 87.99 ± 11.57%, respectively (Pperm &lt; 0.001), indicating that rsFC-based topological attributes can provide promising features for constructing a high-efficacy classifier for MA dependence.

Practical foundations of machine learning for addiction research. Part II. Workflow and use cases

In a continuum with applied statistics, machine learning offers a wide variety of tools to explore, analyze, and understand addiction data. These tools include algorithms that can leverage useful information from data to build models; these models can solve particular tasks to answer addiction scientific questions. In this second part of a two-part review on machine learning, we explain how to apply machine learning methods to addiction research. Like other analytical tools, machine learning methods require a careful implementation to carry out a reproducible and transparent research process with reliable results. This review describes a workflow to guide the application of machine learning in addiction research, detailing study design, data collection, data pre-processing, modeling, and results communication. How to train, validate, and test a model, detect and characterize overfitting, and determine an adequate sample size are some of the key issues when applying machine learning. We also illustrate the process and particular nuances with examples of how researchers in addiction have applied machine learning techniques with different goals, study designs, or data sources as well as explain the main limitations of machine learning approaches and how to best address them. A good use of machine learning enriches the addiction research toolkit.

Practical foundations of machine learning for addiction research. Part I. Methods and techniques

Machine learning assembles a broad set of methods and techniques to solve a wide range of problems, such as identifying individuals with substance use disorders (SUD), finding patterns in neuroimages, understanding SUD prognostic factors and their association, or determining addiction genetic underpinnings. However, the addiction research field underuses machine learning. This two-part narrative review focuses on machine learning tools and concepts, providing an introductory insight into their capabilities to facilitate their understanding and acquisition by addiction researchers. This first part presents supervised and unsupervised methods such as linear models, naive Bayes, support vector machines, artificial neural networks, and k-means. We illustrate each technique with examples of its use in current addiction research. We also present some open-source programming tools and methodological good practices that facilitate using these techniques. Throughout this work, we emphasize a continuum between applied statistics and machine learning, we show their commonalities, and provide sources for further reading to deepen the understanding of these methods. This two-part review is a primer for the next generation of addiction researchers incorporating machine learning in their projects. Researchers will find a bridge between applied statistics and machine learning, ways to expand their analytical toolkit, recommendations to incorporate well-established good practices in addiction data analysis (e.g., stating the rationale for using newer analytical tools, calculating sample size, improving reproducibility), and the vocabulary to enhance collaboration between researchers who do not conduct data analyses and those who do.

Latent Tuberculosis Infection Diagnosis among Household Contacts in a High Tuberculosis-Burden Area: a Comparison between Transcript Signature and Interferon Gamma Release Assay

Diagnosis of latent tuberculosis infection (LTBI) using biomarkers in order to identify the risk of progressing to active TB and therefore predicting a preventive therapy has been the main bottleneck in eradication of tuberculosis. We compared two assays for the diagnosis of LTBI: transcript signatures and interferon gamma release assay (IGRA), among household contacts (HHCs) in a high tuberculosis-burden population. HHCs of active TB cases were recruited for our study; these were confirmed to be clinically negative for active TB disease. Eighty HHCs were screened by IGRA using QuantiFERON-TB Gold Plus (QFT-Plus) to identify LTBI and uninfected cohorts; further, quantitative levels of transcript for selected six genes (TNFRSF10C, ASUN, NEMF, FCGR1B, GBP1, and GBP5) were determined. Machine learning (ML) was used to construct models of different gene combinations, with a view to identify hidden but significant underlying patterns of their transcript levels. Forty-three HHCs were found to be IGRA positive (LTBI) and thirty-seven were IGRA negative (uninfected). FCGR1B, GBP1, and GBP5 transcripts differentiated LTBI from uninfected among HHCs using Livak method. ML and ROC (Receiver Operator Characteristic) analysis validated this transcript signature to have a specificity of 72.7%. In this study, we compared a quantitative transcript signature with IGRA to assess the diagnostic ability of the two, for detection of LTBI cases among HHCs of a high-TB burden population; we concluded that a three gene (FCGR1B, GBP1, and GBP5) transcript signature can be used as a biomarker for rapid screening.

IMPORTANCE The study compares potential of transcript signature and IGRA to diagnose LTBI. It is first of its kind study to screen household contacts (HHCs) in high TB burden area of India. A transcript signature (FCGR1B, GBP1, & GBP5) is identified as potential biomarker for LTBI. These results can lead to development of point-of-care (POC) like device for LTBI screening in a high TB burdened area.

White matter microstructure differences in individuals with dependence on cocaine, methamphetamine, and nicotine: Findings from the ENIGMA-Addiction working group

BACKGROUND

Nicotine and illicit stimulants are very addictive substances. Although associations between grey matter and dependence on stimulants have been frequently reported, white matter correlates have received less attention.

METHODS

Eleven international sites ascribed to the ENIGMA-Addiction consortium contributed data from individuals with dependence on cocaine (n = 147), methamphetamine (n = 132) and nicotine (n = 189), as well as non-dependent controls (n = 333). We compared the fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD) and mean diffusivity (MD) of 20 bilateral tracts. Also, we compared the performance of various machine learning algorithms in deriving brain-based classifications on stimulant dependence.

RESULTS

The cocaine and methamphetamine groups had lower regional FA and higher RD in several association, commissural, and projection white matter tracts. The methamphetamine dependent group additionally showed lower regional AD. The nicotine group had lower FA and higher RD limited to the anterior limb of the internal capsule. The best performing machine learning algorithm was the support vector machine (SVM). The SVM successfully classified individuals with dependence on cocaine (AUC = 0.70, p < 0.001) and methamphetamine (AUC = 0.71, p < 0.001) relative to non-dependent controls. Classifications related to nicotine dependence proved modest (AUC = 0.62, p = 0.014).

CONCLUSIONS

Stimulant dependence was related to FA disturbances within tracts consistent with a role in addiction. The multivariate pattern of white matter differences proved sufficient to identify individuals with stimulant dependence, particularly for cocaine and methamphetamine.

Machine learning approaches for parsing comorbidity/heterogeneity in antisociality and substance use disorders: A primer

By some accounts, as many as 93% of individuals diagnosed with antisocial personality disorder (ASPD) or psychopathy also meet criteria for some form of substance use disorder (SUD). This high level of comorbidity, combined with an overlapping biopsychosocial profile, and potentially interacting features, has made it difficult to delineate the shared/unique characteristics of each disorder. Moreover, while rarely acknowledged, both SUD and antisociality exist as highly heterogeneous disorders in need of more targeted parcellation. While emerging data-driven nosology for psychiatric disorders (e.g., Research Domain Criteria (RDoC), Hierarchical Taxonomy of Psychopathology (HiTOP)) offers the opportunity for a more systematic delineation of the externalizing spectrum, the interrogation of large, complex neuroimaging-based datasets may require data-driven approaches that are not yet widely employed in psychiatric neuroscience. With this in mind, the proposed article sets out to provide an introduction into machine learning methods for neuroimaging that can help parse comorbid, heterogeneous externalizing samples. The modest machine learning work conducted to date within the externalizing domain demonstrates the potential utility of the approach but remains highly nascent. Within the paper, we make suggestions for how future work can make use of machine learning methods, in combination with emerging psychiatric nosology systems, to further diagnostic and etiological understandings of the externalizing spectrum. Finally, we briefly consider some challenges that will need to be overcome to encourage further progress in the field.

Classification of Gamers Using Multiple Physiological Signals: Distinguishing Features of Internet Gaming Disorder

The proliferating and excessive use of internet games has caused various comorbid diseases, such as game addiction, which is now a major social problem. Recently, the American Psychiatry Association classified “Internet gaming disorder (IGD)” as an addiction/mental disorder. Although many studies have been conducted on the diagnosis, treatment, and prevention of IGD, screening studies for IGD are still scarce. In this study, we classified gamers using multiple physiological signals to contribute to the treatment and prevention of IGD. Participating gamers were divided into three groups based on Young’s Internet Addiction Test score and average game time as follows: Group A, those who rarely play games; Group B, those who enjoy and play games regularly; and Group C, those classified as having IGD. In our game-related cue-based experiment, we obtained self-reported craving scores and multiple physiological data such as electrooculogram (EOG), photoplethysmogram (PPG), and electroencephalogram (EEG) from the users while they watched neutral (natural scenery) or stimulating (gameplay) videos. By analysis of covariance (ANCOVA), 13 physiological features (vertical saccadic movement from EOG, standard deviation of N-N intervals, and PNN50 from PPG, and many EEG spectral power indicators) were determined to be significant to classify the three groups. The classification was performed using a 2-layers feedforward neural network. The fusion of three physiological signals showed the best result compared to other cases (combination of EOG and PPG or EEG only). The accuracy was 0.90 and F-1 scores were 0.93 (Group A), 0.89 (Group B), and 0.88 (Group C). However, the subjective self-reported scores did not show a significant difference among the three groups by ANCOVA analysis. The results indicate that the fusion of physiological signals can be an effective method to objectively classify gamers.

A CNN-Based Autoencoder and Machine Learning Model for Identifying Betel-Quid Chewers Using Functional MRI Features

Betel quid (BQ) is one of the most commonly used psychoactive substances in some parts of Asia and the Pacific. Although some studies have shown brain function alterations in BQ chewers, it is virtually impossible for radiologists’ to visually distinguish MRI maps of BQ chewers from others. In this study, we aimed to construct autoencoder and machine-learning models to discover brain alterations in BQ chewers based on the features of resting-state functional magnetic resonance imaging. Resting-state functional magnetic resonance imaging (rs-fMRI) was obtained from 16 BQ chewers, 15 tobacco- and alcohol-user controls (TA), and 17 healthy controls (HC). We used an autoencoder and machine learning model to identify BQ chewers among the three groups. A convolutional neural network (CNN)-based autoencoder model and supervised machine learning algorithm logistic regression (LR) were used to discriminate BQ chewers from TA and HC. Classifying the brain MRIs of HC, TA controls, and BQ chewers by conducting leave-one-out-cross-validation (LOOCV) resulted in the highest accuracy of 83%, which was attained by LR with two rs-fMRI feature sets. In our research, we constructed an autoencoder and machine-learning model that was able to identify BQ chewers from among TA controls and HC, which were based on data from rs-fMRI, and this might provide a helpful approach for tracking BQ chewers in the future.

Machine Learning Assisted Prediction of Prognostic Biomarkers Associated With COVID-19, Using Clinical and Proteomics Data

With the availability of COVID-19-related clinical data, healthcare researchers can now explore the potential of computational technologies such as artificial intelligence (AI) and machine learning (ML) to discover biomarkers for accurate detection, early diagnosis, and prognosis for the management of COVID-19. However, the identification of biomarkers associated with survival and deaths remains a major challenge for early prognosis. In the present study, we have evaluated and developed AI-based prediction algorithms for predicting a COVID-19 patient's survival or death based on a publicly available dataset consisting of clinical parameters and protein profile data of hospital-admitted COVID-19 patients. The best classification model based on clinical parameters achieved a maximum accuracy of 89.47% for predicting survival or death of COVID-19 patients, with a sensitivity and specificity of 85.71 and 92.45%, respectively. The classification model based on normalized protein expression values of 45 proteins achieved a maximum accuracy of 89.01% for predicting the survival or death, with a sensitivity and specificity of 92.68 and 86%, respectively. Interestingly, we identified 9 clinical and 45 protein-based putative biomarkers associated with the survival/death of COVID-19 patients. Based on our findings, few clinical features and proteins correlate significantly with the literature and reaffirm their role in the COVID-19 disease progression at the molecular level. The machine learning-based models developed in the present study have the potential to predict the survival chances of COVID-19 positive patients in the early stages of the disease or at the time of hospitalization. However, this has to be verified on a larger cohort of patients before it can be put to actual clinical practice. We have also developed a webserver CovidPrognosis, where clinical information can be uploaded to predict the survival chances of a COVID-19 patient. The webserver is available at http://14.139.62.220/covidprognosis/.

SAR and QSAR research on tyrosinase inhibitors using machine learning methods

Tyrosinase is a key rate-limiting enzyme in the process of melanin synthesis, which is closely related to human pigmentation disorders. Tyrosinase inhibitors can down-regulate tyrosinase to effectively reduce melanin synthesis. In this work, we conducted structure-activity relationship (SAR) study on 1097 diverse mushroom tyrosinase inhibitors. We applied five kinds of machine learning methods to develop 15 classification models. Model 5B built by fully connected neural networks and ECFP4 fingerprints achieved the highest prediction accuracy of 91.36% and Matthews correlation coefficient (MCC) of 0.81 on the test set. The applicability domains (AD) of classification models were defined by d S T D - P R O method. Moreover, we clustered the 1097 inhibitors into eight subsets by K-Means to figure out inhibitors' structural features. In addition, 10 quantitative structure-activity relationship (QSAR) models were constructed by four machine learning methods based on 813 inhibitors. Model 6 J, the best QSAR model, was developed by fully connected neural networks with 50 RDKit descriptors. It resulted in a coefficient of determination (r ²) of 0.770 and a root mean squared error (RMSE) of 0.482 on the test set. The AD of Model 6 J was visualized by Williams plot. The models built in this study can be obtained from the authors.

Entropy: A Promising EEG Biomarker Dichotomizing Subjects With Opioid Use Disorder and Healthy Controls

Electroencephalography (EEG) signals are known to be nonstationary and often multicomponential signals containing information about the condition of the brain. Since the EEG signal has complex, nonlinear, nonstationary, and highly random behaviour, numerous linear feature extraction methods related to the short-time windowing technique do not satisfy higher classification accuracy. Since biosignals are highly subjective, the symptoms may appear at random in the time scale and very small variations in EEG signals may depict a definite type of brain abnormality it is valuable and vital to extract and analyze the EEG signal parameters using computers. The challenge is to design and develop signal processing algorithms that extract this subtle information and use it for diagnosis, monitoring, and treatment of subjects suffering from psychiatric disorders. For this purpose, finite impulse response-based filtering process was employed rather than traditional time and frequency domain methods. Finite impulse response subbands were analyzed further to obtain feature vectors of different entropy markers and these features were fed into a classifier namely multilayer perceptron. The performances of the classifiers were finally compared considering overall classification accuracies, area under receiver operating characteristic curve scores. Our results underline the potential benefit of the introduced methodology is promising and is to be treated as a clinical interface in dichotomizing substance use disorders subjects and for other medical data analysis studies. The results also indicate that entropy estimators can distinguish normal and opioid use disorder subjects. EEG data and theta frequency band have distinctive capability for almost all types of entropies while nonextensive Tsallis entropy outperforms compared with other types of entropies.

EEG Signal Reconstruction Using a Generative Adversarial Network With Wasserstein Distance and Temporal-Spatial-Frequency Loss

Applications based on electroencephalography (EEG) signals suffer from the mutual contradiction of high classification performance vs. low cost. The nature of this contradiction makes EEG signal reconstruction with high sampling rates and sensitivity challenging. Conventional reconstruction algorithms lead to loss of the representative details of brain activity and suffer from remaining artifacts because such algorithms only aim to minimize the temporal mean-squared-error (MSE) under generic penalties. Instead of using temporal MSE according to conventional mathematical models, this paper introduces a novel reconstruction algorithm based on generative adversarial networks with the Wasserstein distance (WGAN) and a temporal-spatial-frequency (TSF-MSE) loss function. The carefully designed TSF-MSE-based loss function reconstructs signals by computing the MSE from time-series features, common spatial pattern features, and power spectral density features. Promising reconstruction and classification results are obtained from three motor-related EEG signal datasets with different sampling rates and sensitivities. Our proposed method significantly improves classification performances of EEG signals reconstructions with the same sensitivity and the average classification accuracy improvements of EEG signals reconstruction with different sensitivities. By introducing the WGAN reconstruction model with TSF-MSE loss function, the proposed method is beneficial for the requirements of high classification performance and low cost and is convenient for the design of high-performance brain computer interface systems.

Support vector machine-based multivariate pattern classification of methamphetamine dependence using arterial spin labeling

Arterial spin labeling (ASL) magnetic resonance imaging has been widely applied to identify cerebral blood flow (CBF) abnormalities in a number of brain disorders. To evaluate its significance in detecting methamphetamine (MA) dependence, this study used a multivariate pattern classification algorithm, ie, a support vector machine (SVM), to construct classifiers for discriminating MA-dependent subjects from normal controls. Forty-five MA-dependent subjects, 45 normal controls, and 36 heroin-dependent subjects were enrolled. Classifiers trained with ASL-CBF data from the left or right cerebrum showed significant hemispheric asymmetry in their cross-validated prediction performance (P < 0.001 for accuracy, sensitivity, specificity, kappa, and area under the curve [AUC] of the receiver operating characteristics [ROC] curve). A classifier trained with ASL-CBF data from all cerebral regions (bilateral hemispheres and corpus callosum) was able to differentiate MA-dependent subjects from normal controls with a cross-validated prediction accuracy, sensitivity, specificity, kappa, and AUC of 89%, 94%, 84%, 0.78, and 0.95, respectively. The discrimination map extracted from this classifier covered multiple brain circuits that either constitute a network related to drug abuse and addiction or could be impaired in MA-dependence. The cerebral regions contribute most to classification include occipital lobe, insular cortex, postcentral gyrus, corpus callosum, and inferior frontal cortex. This classifier was also specific to MA-dependence rather than substance use disorders in general (ie, 55.56% accuracy for heroin dependence). These results support the future utilization of ASL with an SVM-based classifier for the diagnosis of MA-dependence and could help improve the understanding of MA-related neuropathology.

Applications of machine learning in addiction studies: A systematic review

This study aims to provide a systematic review of the applications of machine learning methods in addiction research. In this study, multiple searches on MEDLINE, Embase and the Cochrane Database of Systematic Reviews were performed. 23 full-text articles were assessed and 17 articles met the inclusion criteria for the final review. The selected studies covered mainly substance addiction (N = 14, 82.4%), including smoking (N = 4), alcohol drinking (N = 3), as well as uses of cocaine (N = 4), opioids (N = 1), and multiple substances (N = 2). Other studies were non-substance addiction (N = 3, 17.6%), including gambling (N = 2) and internet gaming (N = 1). There were eight cross-sectional, seven cohort, one non-randomized controlled, and one crossover trial studies. Majority of the studies employed supervised learning (N = 13), and others employed unsupervised learning (N = 2) and reinforcement learning (N = 2). Among the supervised learning studies, five studies used ensemble learning methods or multiple algorithm comparisons, six used regression, and two used classification. The two included reinforcement learning studies used the direct methods. These results suggest that machine learning methods, particularly supervised learning are increasingly used in addiction psychiatry for informing medical decisions.