Machine-learning-based diagnosis of drug-naive adult patients with attention-deficit hyperactivity disorder using mismatch negativity

Machine Learning-Based Diagnosis of Chronic Subjective Tinnitus With Altered Cognitive Function: An Event-Related Potential Study

OBJECTIVES

Due to the absence of objective diagnostic criteria, tinnitus diagnosis primarily relies on subjective assessments. However, its neuropathological features can be objectively quantified using electroencephalography (EEG). Despite the existing research, the pathophysiology of tinnitus remains unclear. The objective of this study was to gain a deeper comprehension of the neural mechanisms underlying tinnitus through the comparison of cognitive event-related potentials in patients with tinnitus and healthy controls (HCs). Furthermore, we explored the potential of EEG-derived features as biomarkers for tinnitus using machine learning techniques.

DESIGN

Forty-eight participants (24 patients with tinnitus and 24 HCs) underwent comprehensive audiological assessments and EEG recordings. We extracted N2 and P3 components of the midline electrodes using an auditory oddball paradigm, to explore the relationship between tinnitus and cognitive function. In addition, the current source density for N2- and P3-related regions of interest was computed. A linear support vector machine classifier was used to distinguish patients with tinnitus from HCs.

RESULTS

The P3 peak amplitudes were significantly diminished in patients with tinnitus at the AFz, Fz, Cz, and Pz electrodes, whereas the N2 peak latencies were significantly delayed at Cz electrode. Source analysis revealed notably reduced N2 activities in bilateral fusiform gyrus, bilateral cuneus, bilateral temporal gyrus, and bilateral insula of patients with tinnitus. Correlation analysis revealed significant associations between the Hospital Anxiety and Depression Scale-Depression scores and N2 source activities at left insula, right insula, and left inferior temporal gyrus. The best classification performance showed a validation accuracy of 85.42%, validation sensitivity of 87.50%, and validation specificity of 83.33% in distinguishing between patients with tinnitus and HCs by using a total of 18 features in both sensor- and source-level.

CONCLUSIONS

This study demonstrated that patients with tinnitus exhibited significantly altered neural processing during the cognitive-related oddball paradigm, including lower P3 amplitudes, delayed N2 latency, and reduced source activities in specific brain regions in cognitive-related oddball paradigm. The correlations between N2 source activities and Hospital Anxiety and Depression Scale-Depression scores suggest a potential link between the physiological symptoms of tinnitus and their neural impact on patients with tinnitus. Such findings underscore the potential diagnostic relevance of N2- and P3-related features in tinnitus, while also highlighting the interplay between the temporal lobe and occipital lobe in tinnitus. Furthermore, the application of machine learning techniques has shown reliable results in distinguishing tinnitus patients from HCs, reinforcing the viability of N2 and P3 features as biomarkers for tinnitus.

A systematic literature review of machine learning techniques for the detection of attention-deficit/hyperactivity disorder using MRI and/or EEG data

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental condition common in teenagers across the globe. Neuroimaging and Machine Learning (ML) advancements have revolutionized its diagnosis and treatment approaches. Although, the researchers are continuously developing automated ADHD diagnostic tools, there is no reliable ML-based diagnostic system for clinicians. Thus, the study aims to systematically review ML and DL-based approaches for ADHD diagnosis, leveraging brain data from magnetic resonance imaging (MRI) and electroencephalogram (EEG) data. A methodical review for the period 2016 to 2022 is conducted by following the PRISMA guidelines. Four reputable repositories, namely PubMed, IEEE, ScienceDirect, and Springer are searched for the related literature on ADHD diagnosis using MRI/EEG data. 87 studies are selected after screening abstracts of the papers. We critically conducted an analysis of these studies by examining various aspects related to training ML/DL-models, including diverse datasets, hyperparameter tuning, overfitting, and interpretability. The quality and risk assessment is conducted using the QUADAS2 tool to determine the bias due to patient selection, index test, reference standard, and flow and timing. Our rigours analysis observed significant diversity in dataset acquisition and its size, feature extraction and selection techniques, validation strategies and classifier choices. Our findings emphasize the need for generalizability, transparency, interpretability, and reproducibility in future research. The challenges and potential solutions associated with integrating diagnostic models into clinical settings are also discussed. The identified research gaps will guide researchers in developing a reliable ADHD diagnostic system that addresses the associated challenges.

Retinal fundus imaging as biomarker for ADHD using machine learning for screening and visual attention stratification

Attention-deficit/hyperactivity disorder (ADHD), characterized by diagnostic complexity and symptom heterogeneity, is a prevalent neurodevelopmental disorder. Here, we explored the machine learning (ML) analysis of retinal fundus photographs as a noninvasive biomarker for ADHD screening and stratification of executive function (EF) deficits. From April to October 2022, 323 children and adolescents with ADHD were recruited from two tertiary South Korean hospitals, and the age- and sex-matched individuals with typical development were retrospectively collected. We used the AutoMorph pipeline to extract retinal features and used four types of ML models for ADHD screening and EF subdomain prediction, and we adopted the Shapely additive explanation method. ADHD screening models achieved 95.5%-96.9% AUROC. For EF function stratification, the visual and auditory subdomains showed strong (AUROC > 85%) and poor performances, respectively. Our analysis of retinal fundus photographs demonstrated potential as a noninvasive biomarker for ADHD screening and EF deficit stratification in the visual attention domain.

Association between auditory mismatch negativity and visual working memory in school-age children with attention deficit/hyperactivity disorder

BACKGROUND

Attention-deficit/hyperactivity disorder (ADHD) patients exhibit characteristics of impaired working memory (WM) and diminished sensory processing function. This study aimed to identify the neurophysiologic basis underlying the association between visual WM and auditory processing function in children with ADHD.

METHODS

The participants included 86 children with ADHD (aged 6-15 years, mean age 9.66 years, 70 boys, and 16 girls) and 90 typically developing (TD) children (aged 7-16 years, mean age 10.30 years, 66 boys, and 24 girls). Electroencephalograms were recorded from all participants while they performed an auditory discrimination task (oddball task). The visual WM capacity and ADHD symptom severity were measured for all participants.

RESULTS

Compared with TD children, children with ADHD presented a poorer visual WM capacity and a smaller mismatch negativity (MMN) amplitude. Notably, the smaller MMN amplitude in children with ADHD predicted a less impaired WM capacity and milder inattention symptom severity. In contrast, the larger MMN amplitude in TD children predicted a better visual WM capacity.

CONCLUSIONS

Our results suggest an intimate relationship and potential shared mechanism between visual WM and auditory processing function. We liken this shared mechanism to a total cognitive resource limit that varies between groups of children, which could drive correlated individual differences in auditory processing function and visual WM. Our findings provide a neurophysiological correlate for reports of WM deficits in ADHD patients and indicate potential effective markers for clinical intervention.

Virtual reality-assisted prediction of adult ADHD based on eye tracking, EEG, actigraphy and behavioral indices: a machine learning analysis of independent training and test samples

Given the heterogeneous nature of attention-deficit/hyperactivity disorder (ADHD) and the absence of established biomarkers, accurate diagnosis and effective treatment remain a challenge in clinical practice. This study investigates the predictive utility of multimodal data, including eye tracking, EEG, actigraphy, and behavioral indices, in differentiating adults with ADHD from healthy individuals. Using a support vector machine model, we analyzed independent training (n = 50) and test (n = 36) samples from two clinically controlled studies. In both studies, participants performed an attention task (continuous performance task) in a virtual reality seminar room while encountering virtual distractions. Task performance, head movements, gaze behavior, EEG, and current self-reported inattention, hyperactivity, and impulsivity were simultaneously recorded and used for model training. Our final model based on the optimal number of features (maximal relevance minimal redundancy criterion) achieved a promising classification accuracy of 81% in the independent test set. Notably, the extracted EEG-based features had no significant contribution to this prediction and therefore were not included in the final model. Our results suggest the potential of applying ecologically valid virtual reality environments and integrating different data modalities for enhancing robustness of ADHD diagnosis.

Automatic sensory change processing in adults with attention deficit and hyperactivity disorder: a visual mismatch negativity study

In addition to higher-order executive functions, underlying sensory processing ability is also thought to play an important role in Attention-Deficit/Hyperactivity Disorder (AD/HD). An event-related potential feature, the mismatch negativity, reflects the ability of automatic sensory change processing and may be correlated with AD/HD symptoms and executive functions. This study aims to investigate the characteristics of visual mismatch negativity (vMMN) in adults with AD/HD. Twenty eight adults with AD/HD and 31 healthy controls were included in this study. These two groups were matched in age, IQ and sex. In addition, both groups completed psychiatric evaluations, a visual ERP task used to elicit vMMN, and psychological measures about AD/HD symptoms and day-to-day executive functions. Compared to trols, the late vMMN (230-330 ms) was significantly reduced in the AD/HD group. Correlation analyses showed that late vMMN was correlated with executive functions but not AD/HD symptoms. However, further mediation analyses showed that different executive functions had mediated the relationships between late vMMN and AD/HD symptoms. Our findings indicate that the late vMMN, reflecting automatic sensory change processing ability, was impaired in adults with AD/HD. This impairment could have negative impact on AD/HD symptoms via affecting day-to-day executive functions.

From neurons to brain networks, pharmacodynamics of stimulant medication for ADHD

Stimulants represent the first line pharmacological treatment for attention-deficit/hyperactivity disorder (ADHD) and are among the most prescribed psychopharmacological treatments. Their mechanism of action at synaptic level has been extensively studied. However, it is less clear how their mechanism of action determines clinically observed benefits. To help bridge this gap, we provide a comprehensive review of stimulant effects, with an emphasis on nuclear medicine and magnetic resonance imaging (MRI) findings. There is evidence that stimulant-induced modulation of dopamine and norepinephrine neurotransmission optimizes engagement of task-related brain networks, increases perceived saliency, and reduces interference from the default mode network. An acute administration of stimulants may reduce brain alterations observed in untreated individuals in fronto-striato-parieto-cerebellar networks during tasks or at rest. Potential effects of prolonged treatment remain controversial. Overall, neuroimaging has fostered understanding on stimulant mechanism of action. However, studies are often limited by small samples, short or no follow-up, and methodological heterogeneity. Future studies should address age-related and longer-term effects, potential differences among stimulants, and predictors of treatment response.

Differentiation between suicide attempt and suicidal ideation in patients with major depressive disorder using cortical functional network

Studies exploring the neurophysiology of suicide are scarce and the neuropathology of related disorders is poorly understood. This study investigated source-level cortical functional networks using resting-state electroencephalography (EEG) in drug-naïve depressed patients with suicide attempt (SA) and suicidal ideation (SI). EEG was recorded in 55 patients with SA and in 54 patients with SI. Particularly, all patients with SA were evaluated using EEG immediately after their SA (within 7 days). Graph-theory-based source-level weighted functional networks were assessed using strength, clustering coefficient (CC), and path length (PL) in seven frequency bands. Finally, we applied machine learning to differentiate between the two groups using source-level network features. At the global level, patients with SA showed lower strength and CC and higher PL in the high alpha band than those with SI. At the nodal level, compared with patients with SI, patients with SA showed lower high alpha band nodal CCs in most brain regions. The best classification performances for SA and SI showed an accuracy of 73.39%, a sensitivity of 76.36%, and a specificity of 70.37% based on high alpha band network features. Our findings suggest that abnormal high alpha band functional network may reflect the pathophysiological characteristics of suicide and serve as a clinical biomarker for suicide.

Gabor filter-based statistical features for ADHD detection

Attention deficit/hyperactivity disorder (ADHD) is a neuropsychological disorder that occurs in children and is characterized by inattention, impulsivity, and hyperactivity. Early and accurate diagnosis of ADHD is very important for effective intervention. The aim of this study is to develop a computer-aided approach to detecting ADHD using electroencephalogram (EEG) signals. Specifically, we explore a Gabor filter-based statistical features approach for the classification of EEG signals into ADHD and healthy control (HC). The EEG signal is processed by a bank of Gabor filters to obtain narrow-band signals. Subsequently, a set of statistical features is extracted. The computed features are then subjected to feature selection. Finally, the obtained feature vector is given to a classifier to detect ADHD and HC. Our approach achieves the highest classification accuracy of 96.4% on a publicly available dataset. Furthermore, our approach demonstrates better classification accuracy than the existing methods.

A state-of-the-art overview of candidate diagnostic biomarkers for Attention-deficit/hyperactivity disorder (ADHD)

INTRODUCTION

Attention-deficit/hyperactivity disorder (ADHD) is one of the most common neurodevelopmental conditions and is highly heterogeneous in terms of symptom profile, associated cognitive deficits, comorbidities, and outcomes. Heterogeneity may also affect the ability to recognize and diagnose this condition. The diagnosis of ADHD is primarily clinical but there are increasing research efforts aiming at identifying biomarkers that can aid the diagnosis.

AREAS COVERED

We first discuss the definition of biomarkers and the necessary research steps from discovery to implementation. We then provide a broad overview of research studies on candidate diagnostic biomarkers in ADHD encompassing genetic/epigenetic, biochemical, neuroimaging, neurophysiological and neuropsychological techniques. Finally, we critically appraise current limitations in the field and suggest possible ways forward.

EXPERT OPINION

Despite the large number of studies and variety of techniques used, no promising biomarkers have been identified so far. Clinical and biological heterogeneity as well as methodological limitations, including small sample size, lack of standardization, confounding factors, and poor replicability, have hampered progress in the field. Going forward, increased international collaborative efforts are warranted to support larger and more robustly designed studies, develop multimodal datasets to combine biomarkers and improve diagnostic accuracy, and ensure reproducibility and meaningful clinical translation.

Gray matter volume associations in youth with ADHD features of inattention and hyperactivity/impulsivity

BACKGROUND

Prior research has shown smaller cortical and subcortical gray matter volumes among individuals with attention-deficit/hyperactivity disorder (ADHD). However, neuroimaging studies often do not differentiate between inattention and hyperactivity/impulsivity, which are distinct core features of ADHD. The present study uses an approach to disentangle overlapping variance to examine the neurostructural heterogeneity of inattention and hyperactivity/impulsivity dimensions.

METHODS

We analyzed data from 10,692 9- to 10-year-old children from the Adolescent Brain Cognitive Development (ABCD) Study. Confirmatory factor analysis was used to derive factors representing inattentive and hyperactive/impulsive traits. We employed structural equation modeling to examine these factors' associations with gray matter volume while controlling for the shared variance between factors.

RESULTS

Greater endorsement of inattentive traits was associated with smaller bilateral caudal anterior cingulate and left parahippocampal volumes. Greater endorsement of hyperactivity/impulsivity traits was associated with smaller bilateral caudate and left parahippocampal volumes. The results were similar when accounting for socioeconomic status, medication, and in-scanner motion. The magnitude of these findings increased when accounting for overall volume and intracranial volume, supporting a focal effect in our results.

CONCLUSIONS

Inattentive and hyperactivity/impulsivity traits show common volume deficits in regions associated with visuospatial processing and memory while at the same time showing dissociable differences, with inattention showing differences in areas associated with attention and emotion regulation and hyperactivity/impulsivity associated with volume differences in motor activity regions. Uncovering such biological underpinnings within the broader disorder of ADHD allows us to refine our understanding of ADHD presentations.

Acoustic and Text Features Analysis for Adult ADHD Screening: A Data-Driven Approach Utilizing DIVA Interview

Attention Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder commonly seen in childhood that leads to behavioural changes in social development and communication patterns, often continues into undiagnosed adulthood due to a global shortage of psychiatrists, resulting in delayed diagnoses with lasting consequences on individual's well-being and the societal impact. Recently, machine learning methodologies have been incorporated into healthcare systems to facilitate the diagnosis and enhance the potential prediction of treatment outcomes for mental health conditions. In ADHD detection, the previous research focused on utilizing functional magnetic resonance imaging (fMRI) or Electroencephalography (EEG) signals, which require costly equipment and trained personnel for data collection. In recent years, speech and text modalities have garnered increasing attention due to their cost-effectiveness and non-wearable sensing in data collection. In this research, conducted in collaboration with the Cumbria, Northumberland, Tyne and Wear NHS Foundation Trust, we gathered audio data from both ADHD patients and normal controls based on the clinically popular Diagnostic Interview for ADHD in adults (DIVA). Subsequently, we transformed the speech data into text modalities through the utilization of the Google Cloud Speech API. We extracted both acoustic and text features from the data, encompassing traditional acoustic features (e.g., MFCC), specialized feature sets (e.g., eGeMAPS), as well as deep-learned linguistic and semantic features derived from pre-trained deep learning models. These features are employed in conjunction with a support vector machine for ADHD classification, yielding promising outcomes in the utilization of audio and text data for effective adult ADHD screening. Clinical impact: This research introduces a transformative approach in ADHD diagnosis, employing speech and text analysis to facilitate early and more accessible detection, particularly beneficial in areas with limited psychiatric resources. Clinical and Translational Impact Statement: The successful application of machine learning techniques in analyzing audio and text data for ADHD screening represents a significant advancement in mental health diagnostics, paving the way for its integration into clinical settings and potentially improving patient outcomes on a broader scale.

Short-term neuronal and synaptic plasticity act in synergy for deviance detection in spiking networks

Sensory areas of cortex respond more strongly to infrequent stimuli when these violate previously established regularities, a phenomenon known as deviance detection (DD). Previous modeling work has mainly attempted to explain DD on the basis of synaptic plasticity. However, a large fraction of cortical neurons also exhibit firing rate adaptation, an underexplored potential mechanism. Here, we investigate DD in a spiking neuronal network model with two types of short-term plasticity, fast synaptic short-term depression (STD) and slower threshold adaptation (TA). We probe the model with an oddball stimulation paradigm and assess DD by evaluating the network responses. We find that TA is sufficient to elicit DD. It achieves this by habituating neurons near the stimulation site that respond earliest to the frequently presented standard stimulus (local fatigue), which diminishes the response and promotes the recovery (global fatigue) of the wider network. Further, we find a synergy effect between STD and TA, where they interact with each other to achieve greater DD than the sum of their individual effects. We show that this synergy is caused by the local fatigue added by STD, which inhibits the global response to the frequently presented stimulus, allowing greater recovery of TA-mediated global fatigue and making the network more responsive to the deviant stimulus. Finally, we show that the magnitude of DD strongly depends on the timescale of stimulation. We conclude that highly predictable information can be encoded in strong local fatigue, which allows greater global recovery and subsequent heightened sensitivity for DD.

Machine learning in attention-deficit/hyperactivity disorder: new approaches toward understanding the neural mechanisms

Attention-deficit/hyperactivity disorder (ADHD) is a highly prevalent and heterogeneous neurodevelopmental disorder in children and has a high chance of persisting in adulthood. The development of individualized, efficient, and reliable treatment strategies is limited by the lack of understanding of the underlying neural mechanisms. Diverging and inconsistent findings from existing studies suggest that ADHD may be simultaneously associated with multivariate factors across cognitive, genetic, and biological domains. Machine learning algorithms are more capable of detecting complex interactions between multiple variables than conventional statistical methods. Here we present a narrative review of the existing machine learning studies that have contributed to understanding mechanisms underlying ADHD with a focus on behavioral and neurocognitive problems, neurobiological measures including genetic data, structural magnetic resonance imaging (MRI), task-based and resting-state functional MRI (fMRI), electroencephalogram, and functional near-infrared spectroscopy, and prevention and treatment strategies. Implications of machine learning models in ADHD research are discussed. Although increasing evidence suggests that machine learning has potential in studying ADHD, extra precautions are still required when designing machine learning strategies considering the limitations of interpretability and generalization.

Adolescent ADHD and electrophysiological reward responsiveness: A machine learning approach to evaluate classification accuracy and prognosis

We evaluated event-related potential (ERP) indices of reinforcement sensitivity as ADHD biomarkers by examining, in N=306 adolescents (M_age=15.78, SD=1.08), the extent to which ERP amplitude and latency variables measuring reward anticipation and response (1) differentiate, in age- and sex-matched subsamples, (i) youth with vs. without ADHD, (ii) youth at-risk for vs. not at-risk for ADHD, and, in the with ADHD subsample, (iii) youth with the inattentive vs. the hyperactive/impulsive (H/I) and combined presentations. We further examined the extent to which ERP variables (2) predict, in the ADHD subsample, substance use (i) concurrently and (ii) prospectively at 18-month follow-up. Linear support vector machine analyses indicated ERPs weakly differentiate youth with/without (65%) - and at-risk for/not at-risk for (63%) - ADHD but better differentiate ADHD presentations (78%). Regression analyses showed in adolescents with ADHD, ERPs explain a considerable proportion of variance (50%) in concurrent alcohol use and, controlling for concurrent marijuana and tobacco use, explain a considerable proportion of variance (87 and 87%) in, and predict later marijuana and tobacco use. Findings are consistent with the dual-pathway model of ADHD. Results also highlight limitations of a dichotomous, syndromic classification and indicate differences in neural reinforcement sensitivity are a promising ADHD prognostic biomarker.

Technologies to support the diagnosis and/or treatment of neurodevelopmental disorders: A systematic review

In recent years, there has been a great interest in utilizing technology in mental health research. The rapid technological development has encouraged researchers to apply technology as a part of a diagnostic process or treatment of Neurodevelopmental Disorders (NDDs). With the large number of studies being published comes an urgent need to inform clinicians and researchers about the latest advances in this field. Here, we methodically explore and summarize findings from studies published between August 2019 and February 2022. A search strategy led to the identification of 4108 records from PubMed and APA PsycInfo databases. 221 quantitative studies were included, covering a wide range of technologies used for diagnosis and/or treatment of NDDs, with the biggest focus on Autism Spectrum Disorder (ASD). The most popular technologies included machine learning, functional magnetic resonance imaging, electroencephalogram, magnetic resonance imaging, and neurofeedback. The results of the review indicate that technology-based diagnosis and intervention for NDD population is promising. However, given a high risk of bias of many studies, more high-quality research is needed.

Cognitive deficit in adults with ADHD lies in the cognitive state disorder rather than the working memory deficit: A functional near-infrared spectroscopy study

This study tested whether cognitive deficit in patients with adult attention deficit hyperactivity disorder (ADHD) is a working memory deficit or cognitive state disorder during the N-back task. Twenty-two adults with ADHD and twenty-four healthy controls participated in the N-back task. The functional near-infrared spectroscopy (fNIRS) was combined with three perspectives from behavioral and spatial and temporal activation characteristics of blood oxygen levels in the prefrontal cortex to examine the psychological and neuroprocessing characteristics of adult ADHD. Data were acquired using a block design during an N-back task with three memory loads. Visual stimuli were presented on a computer monitor. Behaviorally, response time and accuracy showed no significant differences between the two groups. Spatially, in the left orbitofrontal area and the left frontopolar area (Channels 4 and 11), adult ADHD had significantly higher activation levels of oxyHb in the 2-back task and lower activation levels of deoxyHb in the 3-back task than healthy controls (corrected p < 0.05). Therefore, Channel 4 in the 2-back condition and Channel 11 in the 3-back condition were used as the regions of interest (ROI). Temporally, adults with ADHD peaked earlier in the ROIs than healthy controls. Furthermore, working memory deficit was not found directly from the behavioral performance in adult ADHD. However, adult ADHD can be affected by memory load, task duration, and novelty stimulus. Our findings suggest that patients with adult ADHD have cognitive state disorder instead of working memory deficit.