Identifying ADHD Individuals From Resting-State Functional Connectivity Using Subspace Clustering and Binary Hypothesis Testing

A review of ADHD detection studies with machine learning methods using rsfMRI data

Attention deficit hyperactivity disorder (ADHD) is a common mental health condition that significantly affects school-age children, causing difficulties with learning and daily functioning. Early identification is crucial, and reliable and objective diagnostic tools are necessary. However, current clinical evaluations of behavioral symptoms can be inconsistent and subjective. Functional magnetic resonance imaging (fMRI) is a non-invasive technique that has proven effective in detecting brain abnormalities in individuals with ADHD. Recent studies have shown promising outcomes in using resting state fMRI (rsfMRI)-based brain functional networks to diagnose various brain disorders, including ADHD. Several review papers have examined the detection of other diseases using fMRI data and machine learning or deep learning methods. However, no review paper has specifically addressed ADHD. Therefore, this study aims to contribute to the literature by reviewing the use of rsfMRI data and machine learning methods for detection of ADHD. The study provides general information about fMRI databases and detailed knowledge of the ADHD-200 database, which is commonly used for ADHD detection. It also emphasizes the importance of examining all stages of the process, including network and atlas selection, feature extraction, and feature selection, before the classification stage. The study compares the performance, advantages, and disadvantages of previous studies in detail. This comprehensive approach may be a useful starting point for new researchers in this area.

Unveiling critical ADHD biomarkers in limbic system and cerebellum using a binary hypothesis testing approach

Attention deficit hyperactivity disorder (ADHD) is a common childhood developmental disorder. In recent years, pattern recognition methods have been increasingly applied to neuroimaging studies of ADHD. However, these methods often suffer from limited accuracy and interpretability, impeding their contribution to the identification of ADHD-related biomarkers. To address these limitations, we applied the amplitude of low-frequency fluctuation (ALFF) results for the limbic system and cerebellar network as input data and conducted a binary hypothesis testing framework for ADHD biomarker detection. Our study on the ADHD-200 dataset at multiple sites resulted in an average classification accuracy of 93%, indicating strong discriminative power of the input brain regions between the ADHD and control groups. Moreover, our approach identified critical brain regions, including the thalamus, hippocampal gyrus, and cerebellum Crus 2, as biomarkers. Overall, this investigation uncovered potential ADHD biomarkers in the limbic system and cerebellar network through the use of ALFF realizing highly credible results, which can provide new insights for ADHD diagnosis and treatment.

Identifying non-adult attention-deficit/hyperactivity disorder individuals using a stacked machine learning algorithm using administrative data population registers in a universal healthcare system

Background

This research project aims to build a Machine Learning algorithm (ML) to predict first-time ADHD diagnosis, given that it is the most frequent mental disorder for the non-adult population.

Methods

We used a stacked model combining 4 ML approaches to predict the presence of ADHD. The dataset contains data from population health care administrative registers in Catalonia comprising 1,225,406 non-adult individuals for 2013-2017, linked to socioeconomic characteristics and dispensed drug consumption. We defined a measure of proper ADHD diagnoses based on medical factors.

Results

We obtained an AUC of 79.6% with the stacked model. Significant variables that explain the ADHD presence are the dispersion across patients' visits to healthcare providers; the number of visits, diagnoses related to other mental disorders and drug consumption; age, and sex.

Conclusions

ML techniques can help predict ADHD early diagnosis using administrative registers. We must continuously investigate the potential use of ADHD early detection strategies and intervention in the health system.

Subtype classification of attention deficit hyperactivity disorder with hierarchical binary hypothesis testing framework

Objective. The diagnosis of attention deficit hyperactivity disorder (ADHD) subtypes is important for the refined treatment of ADHD children. Although automated diagnosis methods based on machine learning are performed with structural and functional magnetic resonance imaging (sMRI and fMRI) data which have full observation of brains, they are not satisfactory with the accuracy of less than80%for the ADHD subtype diagnosis.Approach. To improve the accuracy and obtain the biomarker of ADHD subtypes, we proposed a hierarchical binary hypothesis testing (H-BHT) framework by using brain functional connectivity (FC) as input bio-signals. The framework includes a two-stage procedure with a decision tree strategy and thus becomes suitable for the subtype classification. Also, typical FC is extracted in both two stages of identifying ADHD subtypes. That means the important FC is found out for the subtype recognition.Main results. We apply the proposed H-BHT framework to resting state fMRI datasets from ADHD-200 consortium. The results are achieved with the average accuracy97.1%and an average kappa score 0.947. Discriminative FC between ADHD subtypes is found by comparing the P-values of typical FC.Significance. The proposed framework not only is an effective structure for ADHD subtype classification, but also provides useful reference for multiclass classification of mental disease subtypes.

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.

Identifying individuals with attention‐deficit/hyperactivity disorder based on multisite resting‐state functional magnetic resonance imaging: A radiomics analysis

Attention-deficit/hyperactivity disorder (ADHD) is one of the most common neurodevelopmental disorders, characterized by symptoms of age-inappropriate inattention, hyperactivity, and impulsivity. Apart from behavioral symptoms investigated by psychiatric methods, there is no standard biological test to diagnose ADHD. This study aimed to explore whether the radiomics features based on resting-state functional magnetic resonance (rs-fMRI) have more discriminative power for the diagnosis of ADHD. The rs-fMRI of 187 subjects with ADHD and 187 healthy controls were collected from 5 sites of ADHD-200 Consortium. A total of four preprocessed rs-fMRI images including regional homogeneity (ReHo), amplitude of low-frequency fluctuation (ALFF), voxel-mirrored homotopic connectivity (VMHC) and network degree centrality (DC) were used in this study. From each of the four images, we extracted 93 radiomics features within each of 116 automated anatomical labeling brain areas, resulting in a total of 43,152 features for each subject. After dimension reduction and feature selection, 19 radiomics features were retained (5 from ALFF, 9 from ReHo, 3 from VMHC and 2 from DC). By training and optimizing a support vector machine model using the retained features of training dataset, we achieved the accuracy of 76.3% and 77.0% (areas under curve = 0.811 and 0.797) in the training and testing datasets, respectively. Our findings demonstrate that radiomics can be a novel strategy for fully utilizing rs-fMRI information to distinguish ADHD from healthy controls. The rs-fMRI-based radiomics features have the potential to be neuroimaging biomarkers for ADHD.

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.

Machine Learning and MRI-based Diagnostic Models for ADHD: Are We There Yet?

OBJECTIVE

Machine learning (ML) has been applied to develop magnetic resonance imaging (MRI)-based diagnostic classifiers for attention-deficit/hyperactivity disorder (ADHD). This systematic review examines this literature to clarify its clinical significance and to assess the implications of the various analytic methods applied.

METHODS

A comprehensive literature search on MRI-based diagnostic classifiers for ADHD was performed and data regarding the utilized models and samples were gathered.

RESULTS

We found that, although most studies reported the classification accuracies, they varied in choice of MRI modalities, ML models, cross-validation and testing methods, and sample sizes. We found that the accuracies of cross-validation methods inflated the performance estimation compared with those of a held-out test, compromising the model generalizability. Test accuracies have increased with publication year but were not associated with training sample sizes. Improved test accuracy over time was likely due to the use of better ML methods along with strategies to deal with data imbalances.

CONCLUSION

Ultimately, large multi-modal imaging datasets, and potentially the combination with other types of data, like cognitive data and/or genetics, will be essential to achieve the goal of developing clinically useful imaging classification tools for ADHD in the future.

ADHD classification using auto-encoding neural network and binary hypothesis testing

Attention Deficit Hyperactivity Disorder (ADHD) is a highly prevalent neurodevelopmental disease of school-age children. Early diagnosis is crucial for ADHD treatment, wherein its neurobiological diagnosis (or classification) is helpful and provides the objective evidence to clinicians. The existing ADHD classification methods suffer two problems, i.e., insufficient data and feature noise disturbance from other associated disorders. As an attempt to overcome these difficulties, a novel deep-learning classification architecture based on a binary hypothesis testing framework and a modified auto-encoding (AE) network is proposed in this paper. The binary hypothesis testing framework is introduced to cope with insufficient data of ADHD database. Brain functional connectivities (FCs) of test data (without seeing their labels) are incorporated during feature selection along with those of training data and affect the sequential deep learning procedure under binary hypotheses. On the other hand, the modified AE network is developed to capture more effective features from training data, such that the difference of inter- and intra-class variability scores between binary hypotheses can be enlarged and effectively alleviate the disturbance of feature noise. On the test of ADHD-200 database, our method significantly outperforms the existing classification methods. The average accuracy reaches 99.6% with the leave-one-out cross validation. Our method is also more robust and practically convenient for ADHD classification due to its uniform parameter setting across various datasets.

Predicting disease severity in children with combined attention deficit hyperactivity disorder using quantitative features from structural MRI of amygdaloid and hippocampal subfields

OBJECTIVE

Volumetric changes in the amygdaloid and hippocampal subfields have been observed in children with combined attention deficit hyperactivity disorder (ADHD-C). The purpose of this study was to investigate whether volumetric changes in the amygdaloid and hippocampal subfields could be used to predict disease severity in children with ADHD-C.

APPROACH

The data used in this study was from ADHD-200 datasets, a total of 76 ADHD-C patients were included in this study. T1 structural MRI data were used and 64 structural features from the amygdala and hippocampus were extracted. Three ADHD rating scales were used as indicators of ADHD severity. Sequential backward elimination (SBE) algorithm was used for feature selection. A linear support vector regression (SVR) was configured to predict disease severity in children with ADHD-C.

MAIN RESULTS

The three ADHD rating scales could be accurately predicted with the use of SBE-SVR. SBE-SVR achieved the highest accuracy in predicting ADHD index with a correlation of 0.7164 (p < 0.001, tested with 1000-time permutation test). Mean squared error of the SVR was 43.6868, normalized mean squared error was 0.0086, mean absolute error was 3.2893. Several amygdaloid and hippocampal subregions were significantly related to ADHD severity, as revealed by the absolute weight from the SVR model.

SIGNIFICANCE

The proposed SBE-SVR could accurately predict the severity of patients with ADHD-C based on quantitative features extracted from the amygdaloid and hippocampal structures. The results also demonstrated that the two subcortical nuclei could be used as potential biomarkers in the progression and evaluation of ADHD.

Separated Channel Attention Convolutional Neural Network (SC-CNN-Attention) to Identify ADHD in Multi-Site Rs-fMRI Dataset

The accurate identification of an attention deficit hyperactivity disorder (ADHD) subject has remained a challenge for both neuroscience research and clinical diagnosis. Unfortunately, the traditional methods concerning the classification model and feature extraction usually depend on the single-channel model and static measurements (i.e., functional connectivity, FC) in the small, homogenous single-site dataset, which is limited and may cause the loss of intrinsic information in functional MRI (fMRI). In this study, we proposed a new two-stage network structure by combing a separated channel convolutional neural network (SC-CNN) with an attention-based network (SC-CNN-attention) to discriminate ADHD and healthy controls on a large-scale multi-site database (5 sites and n = 1019). To utilize both intrinsic temporal feature and the interactions of temporal dependent in whole-brain resting-state fMRI, in the first stage of our proposed network structure, a SC- CNN is used to learn the temporal feature of each brain region, and an attention network in the second stage is adopted to capture temporal dependent features among regions and extract fusion features. Using a “leave-one-site-out” cross-validation framework, our proposed method obtained a mean classification accuracy of 68.6% on five different sites, which is higher than those reported in previous studies. The classification results demonstrate that our proposed network is robust to data variants and is also replicated across sites. The combination of the SC-CNN with the attention network is powerful to capture the intrinsic fMRI information to discriminate ADHD across multi-site resting-state fMRI data.