Exploring memory function in earthquake trauma survivors with resting-state fMRI and machine learning

Abstract

Background: Traumatized earthquake survivors may develop poor memory function. Resting-state functional magnetic resonance imaging (rs-fMRI) and machine learning techniques may one day aid the clinical assessment of individual psychiatric patients. This study aims to use machine learning with Rs-fMRI from the perspectives of neurophysiology and neuroimaging to explore the association between it and the individual memory function of trauma survivors. Methods: Rs-fMRI data was acquired for eighty-nine survivors (male (33%), average age (SD):45.18(6.31) years) of Wenchuan earthquakes in 2008 each of whom was screened by experienced psychiatrists based on the clinicianadministered post-traumatic stress disorder (PTSD) scale (CAPS), and their memory function scores were determined by the Wechsler Memory Scale-IV (WMS-IV). We explored which memory function scores were significantly associated with CAPS scores. Using simple multiple kernel learning (MKL), Rs-fMRI was used to predict the memory function scores that were associated with CAPS scores. A support vector machine (SVM) was also used to make classifications in trauma survivors with or without PTSD. Results: Spatial addition (SA), which is defined by spatial working memory function, was negatively correlated with the total CAPS score (r = − 0.22, P = 0.04). The use of simple MKL allowed quantitative association of SA scores with statistically significant accuracy (correlation = 0.28, P = 0.03; mean squared error = 8.36; P = 0.04). The left middle frontal gyrus and the left precuneus contributed the largest proportion to the simple MKL association frame. The SVM could not make a quantitative classification of diagnosis with statistically significant accuracy. Limitations: The use of the cross-sectional study design after exposure to an earthquake and the leave-one-out cross-validation (LOOCV) increases the risk of overfitting. Conclusion: Spontaneous brain activity of the left middle frontal gyrus and the left precuneus acquired by rs-fMRI may be a brain mechanism of visual working memory that is related to PTSD symptoms. Machine learning may be a useful tool in the identification of brain mechanisms of memory impairment in trauma survivors.

Multimodal Imaging-Based Classification of PTSD Using Data-Driven Computational Approaches: A Multisite Big Data Study from the ENIGMA-PGC PTSD Consortium

“…Good performance on training data, with poor performance on test data, suggests overfitting, as most machine-learning studies are evaluated only on the basis of cross validation. Therefore, while our accuracy is relatively low (13), the strength of our methods and sample size support the importance of our findings. Conversely, the present results are comparable with machine-learning studies using large scale imaging datasets in other psychiatric disorders based on s-MRI data – 65% accuracy when classifying MDD from HC (9); 65.2% accuracy in differentiating patients with bipolar disorder and from controls (45); and a CV AUC of 0.57-0.61 when classifying cases as OCD or controls (52).…”

“…Specifically, we rigorously tested the classification performance on both cross-validation AUC and test AUC, in which a fully independent portion of the data was left out when selecting the model (both architectures and parameters). We found relatively poor classification performance in classifying PTSD vs. controls (60% test AUC for s-MRI, 59% for rs-fMRI and 56% for d-MRI using SVM), which is lower than top-performing studies conducted at a single site, ranging between 55.56% (13) and 97.1% (14) for rs-fMRI, and between 73% (26) and 80% (29) for studies focusing on multimodal biomarkers. Yet, single-site studies show poor generalization to independent datasets (51), suggesting that performance might be adversely affected by small sample sizes, high-dimensional features, and use of complex models with a large number of parameters.…”

“…First, most extant studies are single-site studies of small homogeneous samples. Although efforts have been made to deal with overfitting (11,12), single-site studies still tend to yield better performance than studies of larger samples, due to overfitting in the latter (13)(14)(15). Second, methodological differences across these studies (e.g., machine learning approaches, scanners, acquisition parameters, and data processing pipelines) limit the ability to directly compare their results.…”

Flights and Perchings of the BrainMind: A Temporospatial Approach to Psychotherapy

“… Cracco et al (2020) found preliminary evidence of dysfunctional spontaneous activity in adult women with experience of childhood abuse as participants showed hypoactivation of the right temporo-parietal junction during a spontaneous cognitive mentalizing task, a region commonly implicated in the individual’s “theory of mind” network (i.e., attributing states, beliefs, or perceptions to others). Using resting-state mean amplitude of spontaneous low-frequency fluctuation, Li et al (2020) found that visual working memory was negatively correlated with severity of PTSD symptoms of earthquake survivors. For individuals with PTSD, visual information processing may be impaired and declarative memory of visual information subsequently.…”

The Contributing Factors of Delayed-Onset Post-traumatic Stress Disorder Symptoms: A Nested Case-Control Study Conducted After the 2008 Wenchuan Earthquake

“…Another study also confirmed that lifetime PTSD affected about 10% of women and 5% of men in the general population (20,21). Moreover, a study focused on the nervous system and revealed the survivors' cognitive might decline, which might lead to loss of well-being in later life (22,23). In addition, according to the stress sensitization hypothesis, that individuals who have experienced previous PTSD are more susceptible to developing PTSD following subsequent traumas (24,25).…”