優秀學生論文
競賽決選

The ability to extract a sound target from competing background sounds is essential to speech and music communication, which often occurs in sub-optimal listening context. Recent evidence suggests that absolute pitch (AP) proficiency, the ability to identify musical tones without a referent, is associated with hearing-in-noise benefits. The veridical mapping account postulates a link between better segregation ability in AP individuals and regional hyper-connectivity at the neural level. However, how AP proficiency modulates the neural mechanisms and functional connectivity underlying hearing-in-noise perception remains unclear. Here we use functional magnetic resonance imaging to contrast brain responses to speech and melody targets masked in noise under three signal-to-noise ratios (SNR: no noise, 0, －9 dB) in AP musicians, non-AP musicians, and non-musicians. Results showed that AP ability modulated music-evoked activations extensively in bilateral temporal and parietal regions associated with auditory-motor integration, most notably in the right superior and middle temporal gyri (STG, MTG), precentral gyrus, and right superior parietal lobule. In comparison, AP musicians showed greater activations in left auditory regions (STG, MTG), the left superior and inferior frontal gyri and insula for speech perception in noise, with increasing activation at lower SNRs. An enhanced functional connectivity between the right auditory seed (STG) and Heschl’s gyrus and the superior marginal gyrus paralleled an AP advantage in music-in-noise perception. For speech perception in noise, AP musicians relies on stronger functional connectivity between the left auditory seed and left inferior frontal gyrus and insula, areas implicated in speech-motor circuit. Additionally, an increased interhemispheric functional connectivity between the superior temporal gyrus and angular region was revealed for musicians over non-musicians for speech and melody perception. Our findings suggest that AP benefits hearing-in-noise perception by enhancing bottom-up neural representations in auditory regions with distinct frontal-parietal modulations for music and speech streams.

Negative expectations (i.e., expecting increased pain) and positive expectations (i.e., expecting decreased pain) toward noxious stimulations respectively exacerbate and alleviate human pain perceptions; however, underlying psychological factors remain unclear. By applying the functional magnetic resonance imaging technology, we aim to investigate emotional mechanisms underlying pain expectancy modulations. Thirty-one participants were instructed to use emotion regulation strategies to down-regulate expectation-related emotions in a cue-based expectancy paradigm. As indicated by subjective emotional ratings and skin conductance responses, participants successfully reduced their anxiety toward negative expectations and pleasantness toward positive expectations when applying emotion regulation strategies. Importantly, when participants actively down-regulated their expectation-related emotions, both negative and positive expectancy effects on pain were diminished. Furthermore, the reduction in emotional ratings predicted the expectancy effects, supporting the presence of the emotional component in the expectancy effects. At the neural level, negative expectations recruited the amygdala in anxiety-related processing and positive expectations involved the medial orbitofrontal cortex (mOFC) tracking the pleasantness. We proposed a framework to describe pain perceptions as the output of an emotion-dependent integration between the prior experience (i.e., expectation) and incoming sensory input. The observed individual expectancy effects significantly correlated with hypothetical effects predicted by our algorithm, with negative expectations engaging connectivity between the amygdala and rostral anterior cingulate cortex, a region encoding pain intensity, and positive expectations engaging the functional coupling between the mOFC and anterior hippocampus, a region implicated in regulations on the anxiety during pain relief. Additionally, expectancy-related emotions modulated the encoding of aversive prediction error, a mismatch between the expected and perceived pain intensity, in the periaqueductal grey for negative expectations and mOFC for positive expectations. Taken together, the current study identifies that emotion plays an important role in not only the integration but also the mismatch detection between expectations and the nociception for human pain modulations.

Reward and punishment act as important factors that modulate human behavior through learning to maximize the former, and minimizing the latter. Although much is known about how reward and punishment contribute to guiding our behavior independently, how these two types of learning interact with each other remains largely unclear. At the neural level, whether reward learning and punishment learning involve common or distinct neural substrates is still under debate. The aim of the current study was to investigate (1) the neural mechanisms underlying reward learning and punishment learning by comparing between these two types learning in a single experiment, and (2) the interaction between reward learning and punishment learning at both behavioral and neural levels. By using a probabilistic instrumental learning task in combination with functional magnetic resonance imaging, healthy participants were required to try their best to earn money and to avoid losing money or avoid painful stimulation at the same time. We found that different brain regions contributed to the encoding of prediction error signals for reward versus punishment learning. Moreover, when an option was simultaneously associated with rewarding and punishing outcomes, only the performance of reward learning was interfered by punishment learning, but not vice versa. When reward learning was interfered by punishment learning, we discovered the engagement of punishment-related prediction error signals during this process. Taken together, these findings provide novel insights into the understanding about the difference between reward learning and punishment learning in humans.

With the intentional and implicit messages provided by mutual gaze, both parties can share attention and become more engaged. However, as interactions increasingly take place remotely, establishing a mutual gaze seems challenging. This study leveraged eye trackers to create a pseudo–mutual gaze channel, allowing each interacting dyad’s gazes to be mirrored on each remote screen. We used hyperscanning electroencephalography (EEG) technique to simultaneously record the brain activity of interacting dyads involved in a joint attention task to demonstrate fluctuations in interpersonal interactions. Results showed that the mutual gaze could facilitate remote partners to perform cooperative and cooperative activities efficiently. The low-frequency interbrain synchrony (IBS) estimated using the phase locking value (PLV) involving frontal and temporal areas varied with interaction modes. The current evidence supports that the shared gaze is a promising nonverbal channel to facilitate effective interaction and joint attention between remote parties.

Processing uncertain events is associated with the P300 component. It has been shown that there is a significant relationship between P300 and event uncertainty using a quantity known as surprise, which can be used to quantify the subjective probability of an event. It has been shown, however, that in sequence learning humans learn probabilities of event transitions that take into account both event repetitions and event alterations. How the event probability or event transition probability in a dynamic sequence influences P300 has not been addressed. In this study, we investigate whether fluctuations of the P300 induced by surprise are better explained by event probability or by event transition probability. A hidden Markov model (HMM) was employed in our experiment for generating a binary sequence based on a two-choice response task. Different event transitions changed people's response times. According to the model-based analysis, trial-by-trial P300 is better explained by surprise based on event transition probabilities. Our results support that human uses event transition knowledge in a changing environment.

Understanding how the human brain processes has always been a fundamental challenge. The free-energy principle proposes a unified theory that the brain constantly minimizes its free energy as the entropy of the environment increases. Self-organizing systems tend to reduce surprise; however, previous studies also suggest that humans actively seek novelty in recreational activities such as aesthetic experiences and have higher preferences for information with intermediate complexity. The present study proposes that self-organizing systems seek novelty according to the encountered uncertainty level to maintain homeostasis. To validate this, 33 young adults took part in the functional magnetic resonance imaging (fMRI) experiment in which they listened to monophonic melodies with various uncertainty (entropy) levels in the probe phase, decided whether they wanted to listen to it again, and listened to the monophonic melodies in the outcome phase which were probabilistically determined by their own decisions. Behaviorally, our results suggest reliable preferences for information with intermediate uncertainty. Neurally, we observed a positive linear effect in the bilateral caudate as well as a positive quadratic effect in the bilateral frontal regions, putamen, and the mid cingulate. In addition, we found a stronger neural activation in the default mode network (DMN) regions when the brain is satisfied with the uncertainty level of the information, and a stronger activation in the task regions when the brain detects insufficient or excessive uncertainty and activates the exploration state. Overall, our findings indicate that the brain is constantly balancing between its default state and exploration state in order to reach equilibrium.

Spatial navigation (SN) is one of the cognitive functions compromised by aging. Older adults (OA) require more time and make more errors to construct a cognitive map or reach target locations. This age-related SN change can be attributed to many factors, such as perception alterations. Previous studies indicate that inaccurate distance judgement (DJ) may be another cause. Healthy OAs tend to underestimate actual distances more than their younger counterparts as actual distances increase. Such DJ underestimation may imply that their mental representations of the world are more distorted than younger adults’ (YA). Contemporary imaging research also suggests different SN neural responses between OAs and YAs. However, how older brains process these distortions under the frameworks of SN remains unclear. To assay neural representations of age-related distance distortion, we investigated the neural representations of distances in YAs and OAs. Ergo, a SN and DJ paradigm within a virtual environment was utilized in the current study. Participants freely navigated the virtual environment and encoded the map to criterion. At testing, participants were asked to deliberate distances between start and target landmarks, and travel to targets inside a 3T scanner. The behavioral results demonstrated that OAs tended to underestimate further and overestimate nearer distances more than YAs. To identify distance-related areas, representational similarity analysis (RSA) was applied to functional images during DJs. We discovered YAs and OAs tended to employee different strategies and corresponding networks during judgements to represent distances cooccurring with age-related distortions. YAs recruited frontal, fronto-parietal, and occipital regions to adapt elaboration, whereas OAs recruited frontal, striatal, fronto-parietal, temporal, and thalamic regions to adapt constructions. Eventually, information would be jointly projected to a cluster of the left anterior cingulate cortex for gating final DJ outputs. Such functional and strategic shifts further result in age-related distance distortions during judgement under SN frameworks.

Motor sequence learning is the ability to acquire information of a series of actions and to improve performance in executing them. Though the nature of motor sequence learning has been intensively studied with the Serial Reaction Time Task (SRTT) paradigm, how do spatial and temporal variations of the sequence contribute to learning remains unclear and is the focus of the current study. Here we have adopted a SRTT with four spatially distinct choices that prompt responses in a fixed series of 8 elements. Four different temporal intervals segregated each of the sequential elements, also in a fixed sequential manner. In the training phase, participants performed the SRTT on the repetitive spatiotemporal sequence intermixing the aforementioned spatial and temporal constraints. In the post-training phase, participants were tested on various blocks of sequence, retaining only the original spatial, temporal, or none of the sequential information. The outcomes demonstrated significant learning, as indicated by the RT difference between each post-training spatial or temporal blocks vs. The random block, for both temporal and spatial variations, suggesting the respective contributions from temporal and spatial variation in the sequence. Furthermore, the summated “learning effect” of spatial and temporal sequences did not differ from that of the spatiotemporal sequence, while the correlations among the learning effects were only significant between the spatial and the spatiotemporal sequences. Taken together, our findings suggest that spatial and temporal information processing independently contributed to the learning of the integrated spatiotemporal sequence, where the degree of learning is more reliably determined by the spatial information, reflecting the salience of the spatial information in motor sequence learning.

Introduction. The insula cortex is involved in processing both external stimuli and stimuli from within the body. Due to its’ complexity and involvement in a variety of different functions, parcellation of the insular cortex became an important methodological question. Previous studies using the gradient approach showed that the functional connectivity diversity of the insula could be considered a continuum with a gradual change rather than parcellate it into distinct subregions. The study aimed to apply the gradient approach to investigate the structural connectivity diversity of the insula cortex and its correspondence with activity in functional magnetic resonance imaging (fMRI).

Methods. Sixteenth healthy subjects underwent fMRI with interoceptive and exteroceptive conditions. During the first one, participants silently counted their heartbeats. For the latter one, participants silently counted the tones played through headphones attached to the scanner. The same participants underwent diffusion magnetic resonance imaging (dMRI). The data was used for tractography using probabilistic tracking with crossing fibres (PROBTRACKX) from each insula voxel to a set of regions of interest (ROI). Gradients were computed using BrainSpace toolbox based on probability matrices as an output of PROBTRACKX. 

Results. Analysis of fMRI revealed differential activity within insula between two conditions. Specifically, a higher positive BOLD response was seen during the interoceptive condition than exteroceptive. The primary gradient in both hemispheres aligned with the superior-inferior axis. This structural connectivity gradient was positively correlated with insula activation in fMRI for the interoceptive condition compared to the exteroceptive one.

Conclusion. The primary gradient of structural connectivity diversity in the insula cortex follows the superior-inferior axis of the region. This gradient correlates with differential insula activity in interoceptive against exteroceptive condition. These results suggest that the difference in brain responses to internal and external stimuli is guided by structural connectivity across the cortex.

Aims: Integrating Holo-Hilbert Spectral Analysis (HHSA) and machine learning algorithms in rsEEG analysis to discriminate mild cognitive decline (MCI) and mild Alzheimer's disease (AD1) from cognitively normal elderly and to predict the progression of MCI to AD within 3-year longitudinal cohort. 

Methods: We recruited 154 participants from three hospitals, which included cognitive normal elderly (CN; n=51, MMSE>26), MCI (n=42, CDR=0.5, MMSE>25, and AD1 (n=61, CDR=1, MMSE<25) with rsEEG recordings. Besides, seventy-two patients with MCI (CDR= 0.5) were longitudinally followed with two rsEEG recordings within three years. Two subgroups were defined according to the clinical results, namely MCI-stable (MCI-S) and MCI-converted (MCI-C) subgroups. We compared the performance of three analytic methods (HHSA, Hilbert-Huang transform, and windowed fast Fourier transform) for feature engineering with subsequent machine learning algorithms in analyzing the rsEEG.

Results: (a) At the group-level analysis, the HHSA contrast of MCI and AD1 compared with CN shows that increasing amplitude modulation (AM) power of lower frequency oscillations (LFO; delta and theta bands) became prominent with disease progression while decreasing AM power of higher frequency oscillations (HFO; beta and gamma bands). The alpha frequency oscillations show a dual AM pattern revealing an increasing AM power in posterior brain regions but decreasing AM power in anterior brain regions in MCI and AD1. (b) At the individual-level analysis, HHSA-based feature extraction outperforms other methods in discriminating MCI and AD1 from CN with sensitivity and specificity of 0.82 and 0.80, 0.94 and 0.80, respectively. (c) In the MCI longitudinal cohort, the baseline HHSA contrast between MCI-C and MCI-S shows a significantly decreasing AM power of alpha and beta band oscillations. 

Conclusions: Integrating HHSA-based feature extraction and machine learning algorithms in analyzing rsEEG surpass other analytic methods for classifying and predicting patients with early cognitive decline.