Resting-State Functional Connectivity-Based Biomarkers and Functional MRI-Based Neurofeedback for Psychiatric Disorders: A Challenge for Developing Theranostic Biomarkers

Abstract

Psychiatric research has been hampered by an explanatory gap between psychiatric symptoms and their neural underpinnings, which has resulted in poor treatment outcomes. This situation has prompted us to shift from symptom-based diagnosis to data-driven diagnosis, aiming to redefine psychiatric disorders as disorders of neural circuitry. Promising candidates for data-driven diagnosis include resting-state functional connectivity MRI (rs-fcMRI)-based biomarkers. Although biomarkers have been developed with the aim of diagnosing patients and predicting the efficacy of therapy, the focus has shifted to the identification of biomarkers that represent therapeutic targets, which would allow for more personalized treatment approaches. This type of biomarker (i.e., "theranostic biomarker") is expected to elucidate the disease mechanism of psychiatric conditions and to offer an individualized neural circuit-based therapeutic target based on the neural cause of a condition. To this end, researchers have developed rs-fcMRI-based biomarkers and investigated a causal relationship between potential biomarkers and disease-specific behavior using functional MRI (fMRI)-based neurofeedback on functional connectivity. In this review, we introduce a recent approach for creating a theranostic biomarker, which consists mainly of 2 parts: (1) developing an rs-fcMRI-based biomarker that can predict diagnosis and/or symptoms with high accuracy, and (2) the introduction of a proof-of-concept study investigating the relationship between normalizing the biomarker and symptom changes using fMRI-based neurofeedback. In parallel with the introduction of recent studies, we review rs-fcMRI-based biomarker and fMRI-based neurofeedback, focusing on the technological improvements and limitations associated with clinical use.

Keywords: neurofeedback; psychiatric disorder; resting-state functional connectivity; theranostic biomarker.

Figure 1.

Distribution of weighted linear summations (WLS) calculated by functional connections. (a) The white and black bars denote the number of typically developing (TD) and autism spectrum disorder (ASD) individuals in the Japanese dataset, respectively. A horizontal axis denotes WLS score. If the WLS score is positive, an individual is classified as having ASD, while a negative WLS score indicates TD. (b) A histogram shows the distribution of WLS scores for the US ABIDE dataset. (c) The density distribution of WLS when applying the ASD classifier to various psychiatric conditions, such as ASD, schizophrenia (SCZ), ADHD, and major depressive disorder (MDD). In each panel, TD/HC distribution is gray and ASD distribution is red. The distribution of other psychiatric conditions (i.e., SCZ, ADHD, and MDD) is colored with blue, green, and yellow, respectively. Area under the curve (AUC) values are based on the classification between each psychiatric condition and TD/HC. P values are obtained by the Benjamini-Hochberg-corrected Kolmogorov-Smirnov test. The TD distribution of WLS at each panel is adjusted to have the same median and SD for the visualization purpose. Adapted, with permission, from Figures 1 and 5 in Yahata et al. A small number of abnormal connections predicts adult autism spectrum disorder. Nature Communications, DOI: 10.1038/ncomms11254 (2016).

Figure 2.

The procedure of decoded neurofeedback (DecNef). During training, participants were instructed to self-regulate brain activity to maximize the feedback score. This was represented by, for example, the size of a green disc, which corresponded to the participant’s success in inducing a current brain activity pattern as similar as possible to the target brain activity pattern.

Figure 3.

The procedure of functional connectivity (FC)-based neurofeedback (FCNef). During training, participants were instructed to self-regulate brain activity to maximize the feedback score. This was represented by, for example, the size of a green disc, which reflected the degree of success in achieving target FC.

Figure 4.

Results from 3 individuals with depression and 7 subclinical participants. (a) and (b) show the results of participants with depression. (c) and (d) show the results of participants with subclinical depression. (a) Neurofeedback scores across the 4 training days. Red bar denotes the mean of neurofeedback scores for all trials. Error bar denotes SEM. Asterisk shows the statistical significance (P < .001). (b) Hamilton Depression Rating Scale scores at pre- and postfunctional connectivity-based neurofeedback (FCNef). Red bar denotes the mean of Hamilton Depression Rating Scale scores and error bar shows SEM. (c) Neurofeedback scores in the same format as a. Asterisk shows the statistical significance (P < .01). (d) Scatter plot of the change in the Beck Depression Inventory (BDI) score vs the change of the target resting-state functional connectivity (FC) MRI (rs-fcMRI) between post- and preneurofeedback. Each dot represents individual data. The line denotes the linear regression of the change of BDI score from the change of the target rs-fcMRI.

Figure 5.

Neurofeedback-induced change of functional connectivity (FC) toward the neurotypical pattern in a case of adult high-functioning autism spectrum disorder (ASD). The graph shows the feedback scores during the training sessions (blank squares and error bars) and the outputs of the ASD biomarker (Yahata et al., 2016) using the resting-state FC data collected before (i.e., RS-1 and RS-2) and after (i.e., RS-3) the neurofeedback training (x signs). The open circle denotes the mean output of the ASD biomarker across the three rs-fcMRI sessions conducted in a single day. Although the linear weighted summation of FCs in the ASD biomarker ranged between 0 (neurotypical pattern) and 1 (typical ASD pattern), the value was fed into a mathematical transformation involving a sigmoid function, such that the output of the ASD biomarker ranged between 0 (typical ASD) to 100 (neurotypical). In each training day, there were 6 runs (filled squares and error bars; except for three runs in the final day), each of which had 10 trials. Note that, whereas the outputs of the biomarker had remained close to 0 before the training, the resting state FCs exhibited the neurotypical pattern at least twice out of 3 scans in the posttraining, which was acquired 3 weeks after the training.