Classification of borderline personality disorder based on spectral power of resting-state fMRI

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.

Aberrant resting-state functional connectivity in incarcerated women with elevated psychopathic traits

Previous work in incarcerated men suggests that individuals scoring high on psychopathy exhibit aberrant resting-state paralimbic functional network connectivity (FNC). However, it is unclear whether similar results extend to women scoring high on psychopathy. This study examined whether psychopathic traits [assessed via the Hare Psychopathy Checklist - Revised (PCL-R)] were associated with aberrant inter-network connectivity, intra-network connectivity (i.e., functional coherence within a network), and amplitude of fluctuations across limbic and surrounding paralimbic regions among incarcerated women (n = 297). Resting-state networks were identified by applying group Independent Component Analysis to resting-state fMRI scans. We tested the association of psychopathic traits (PCL-R Factor 1 measuring interpersonal/affective psychopathic traits and PCL-R Factor 2 assessing lifestyle/antisocial psychopathic traits) to the three FNC measures. PCL-R Factor 1 scores were associated with increased low-frequency fluctuations in executive control and attentional networks, decreased high-frequency fluctuations in executive control and visual networks, and decreased intra-network FNC in default mode network. PCL-R Factor 2 scores were associated with decreased high-frequency fluctuations and default mode networks, and both increased and decreased intra-network functional connectivity in visual networks. Similar to previous analyses in incarcerated men, our results suggest that psychopathic traits among incarcerated women are associated with aberrant intra-network amplitude fluctuations and connectivity across multiple networks including limbic and surrounding paralimbic regions.

Borderline personality disorder classification based on brain network measures during emotion regulation

Borderline Personality Disorder (BPD) is characterized by an increased emotional sensitivity and dysfunctional capacity to regulate emotions. While amygdala and prefrontal cortex interactions are regarded as the critical neural mechanisms underlying these problems, the empirical evidence hereof is inconsistent. In the current study, we aimed to systematically test different properties of brain connectivity and evaluate the predictive power to detect borderline personality disorder. Patients with borderline personality disorder (n = 51), cluster C personality disorder (n = 26) and non-patient controls (n = 44), performed an fMRI emotion regulation task. Brain network analyses focused on two properties of task-related connectivity: phasic refers to task-event dependent changes in connectivity, while tonic was defined as task-stable background connectivity. Three different network measures were estimated (strength, local efficiency, and participation coefficient) and entered as separate models in a nested cross-validated linear support vector machine classification analysis. Borderline personality disorder vs. non-patient controls classification showed a balanced accuracy of 55%, which was not significant under a permutation null-model, p = 0.23. Exploratory analyses did indicate that the tonic strength model was the highest performing model (balanced accuracy 62%), and the amygdala was one of the most important features. Despite being one of the largest data-sets in the field of BPD fMRI research, the sample size may have been limited for this type of classification analysis. The results and analytic procedures do provide starting points for future research, focusing on network measures of tonic connectivity, and potentially focusing on subgroups of BPD.

Brain structure and symptom dimensions in borderline personality disorder

BACKGROUND

Borderline personality disorder (BPD) presents with symptoms across different domains, whose neurobiology is poorly understood.

METHODS

We applied voxel-based morphometry on high-resolution magnetic resonance imaging scans of 19 female BPD patients and 50 matched female controls.

RESULTS

Group comparison showed bilateral orbitofrontal gray matter loss in patients, but no significant changes in the hippocampus. Voxel-wise correlation of gray matter with symptom severity scores from the Borderline Symptom List (BSL-95) showed overall negative correlation in bilateral prefrontal, right inferior temporal/fusiform and occipital cortices, and left thalamus. Significant (negative) correlations with BSL-95 subscores within the patient cohort linked autoaggression to left lateral prefrontal and insular cortices, right inferior temporal/temporal pole, and right orbital cortex; dysthymia/dysphoria to right orbitofrontal cortex; self-perception to left postcentral, bilateral inferior/middle temporal, right orbitofrontal, and occipital cortices. Schema therapy-based Young Schema Questionnaire (YSQ-S2) scores of early maladaptive schemas on emotional deprivation were linked to left medial temporal lobe gray matter reductions.

CONCLUSIONS

Our results confirm orbitofrontal structural deficits in BPD, while providing a framework and preliminary findings on identifying structural correlates of symptom dimensions in BPD, especially with dorsolateral and orbitofrontal cortices.

Ranking Regions, Edges and Classifying Tasks in Functional Brain Graphs by Sub-Graph Entropy

This paper considers analysis of human brain networks or graphs constructed from time-series collected from functional magnetic resonance imaging (fMRI). In the network of time-series, the nodes describe the regions and the edge weights correspond to the absolute values of correlation coefficients of the time-series of the two nodes associated with the edges. The paper introduces a novel information-theoretic metric, referred as sub-graph entropy, to measure uncertainty associated with a sub-graph. Nodes and edges constitute two special cases of sub-graph structures. Node and edge entropies are used in this paper to rank regions and edges in a functional brain network. The paper analyzes task-fMRI data collected from 475 subjects in the Human Connectome Project (HCP) study for gambling and emotion tasks. The proposed approach is used to rank regions and edges associated with these tasks. The differential node (edge) entropy metric is defined as the difference of the node (edge) entropy corresponding to two different networks belonging to two different classes. Differential entropy of nodes and edges are used to rank top regions and edges associated with the two classes of data. Using top node and edge entropy features separately, two-class classifiers are designed using support vector machine (SVM) with radial basis function (RBF) kernel and leave-one-out method to classify time-series for emotion task vs. no-task, gambling task vs. no-task and emotion task vs. gambling task. Using node entropies, the SVM classifier achieves classification accuracies of 0.96, 0.97 and 0.98, respectively. Using edge entropies, the classifier achieves classification accuracies of 0.91, 0.96 and 0.94, respectively.

A signature-based machine learning model for distinguishing bipolar disorder and borderline personality disorder

Mobile technologies offer new opportunities for prospective, high resolution monitoring of long-term health conditions. The opportunities seem of particular promise in psychiatry where diagnoses often rely on retrospective and subjective recall of mood states. However, deriving clinically meaningful information from the complex time series data these technologies present is challenging, and the current implications for patient care are uncertain. In this study, 130 participants with bipolar disorder (n = 48) or borderline personality disorder (n = 31) and healthy volunteers (n = 51) completed daily mood ratings using a bespoke smartphone app for up to 1 year. A signature-based learning method was used to capture the evolving interrelationships between the different elements of mood and exploit this information to classify participants' diagnosis and to predict subsequent mood. The three participant groups could be distinguished from one another on the basis of self-reported mood using the signature methodology. The methodology classified 75% of participants into the correct diagnostic group compared with 54% using standard approaches. Subsequent mood ratings were correctly predicted with >70% accuracy. Prediction of mood was most accurate in healthy volunteers (89-98%) compared to bipolar disorder (82-90%) and borderline personality disorder (70-78%). The signature method provided an effective approach to the analysis of mood data both in terms of diagnostic classification and prediction of future mood. It also highlighted the differing predictability and the overlap inherent within disorders. The three cohorts offered internally consistent but distinct patterns of mood interaction in their reporting which have the potential to enable more efficient and accurate diagnoses and thus earlier treatment.

Function-specific and Enhanced Brain Structural Connectivity Mapping via Joint Modeling of Diffusion and Functional MRI

A joint structural-functional brain network model is presented, which enables the discovery of function-specific brain circuits, and recovers structural connections that are under-estimated by diffusion MRI (dMRI). Incorporating information from functional MRI (fMRI) into diffusion MRI to estimate brain circuits is a challenging task. Usually, seed regions for tractography are selected from fMRI activation maps to extract the white matter pathways of interest. The proposed method jointly analyzes whole brain dMRI and fMRI data, allowing the estimation of complete function-specific structural networks instead of interactively investigating the connectivity of individual cortical/sub-cortical areas. Additionally, tractography techniques are prone to limitations, which can result in erroneous pathways. The proposed framework explicitly models the interactions between structural and functional connectivity measures thereby improving anatomical circuit estimation. Results on Human Connectome Project (HCP) data demonstrate the benefits of the approach by successfully identifying function-specific anatomical circuits, such as the language and resting-state networks. In contrast to correlation-based or independent component analysis (ICA) functional connectivity mapping, detailed anatomical connectivity patterns are revealed for each functional module. Results on a phantom (Fibercup) also indicate improvements in structural connectivity mapping by rejecting false-positive connections with insufficient support from fMRI, and enhancing under-estimated connectivity with strong functional correlation.

Discriminability of personality profiles in isolated and Co-morbid marijuana and nicotine users

Specific personality traits have been linked with substance use disorders (SUDs), genetic mechanisms, and brain systems. Thus, determining the specificity of personality traits to types of SUD can advance the field towards defining SUD endophenotypes as well as understanding the brain systems involved for the development of novel treatments. Disentangling these factors is particularly important in highly co morbid SUDs, such as marijuana and nicotine use, so treatment can occur effectively for both. This study evaluated personality traits that distinguish isolated and co-morbid use of marijuana and nicotine. To that end, we collected the NEO Five Factor Inventory in participants who used marijuana-only (n=59), nicotine-only (n=27), both marijuana and nicotine (n=28), and in non-using controls (n=28). We used factor analyses to identify personality profiles, which are linear combinations of the five NEO Factors. We then conducted Receiver Operating Characteristics (ROC) curve analysis to test accuracy of the personality factors in discriminating isolated and co-morbid marijuana and nicotine users from each other. ROC curve analysis distinguished the four groups based on their NEO personality patterns. Results showed that NEO Factor 2 (openness, extraversion, agreeableness) discriminated marijuana and marijuana+nicotine users from controls and nicotine-only users with high predictability. Additional ANOVA results showed that the openness dimension discriminated marijuana users from nicotine users. These findings suggest that personality dimensions distinguish marijuana users from nicotine users and should be considered in prevention strategies.