Fig. 3.1
The topological map of non-drug users (a) and chronic heroin users (b) (Reprint with permission from Liu et al. 2009). The colors represented medial temporal (yellow), subcortical (purple), temporal (pink), parietal-(pre)motor (brown), frontal (red), and occipital (green). Black triangles represent the most remarkable increases in connectivity strength in chronic heroin users versus non-drug users (P < 0.05, two-sample t-test ) (Liu et al. 2009)
As shown in Fig. 3.1, dysfunctional integrations were detected among the heroin-addicted individuals during the resting state. Repeated drug administration might lead to significant alterations in dendritic branching and spine density resultant and pathway irregularities in chronic heroin user, which subsequently led to irregular brain network connectivity and the abovementioned cognitive impairments.
3.2.1.1 Alterations in Functional Network Characteristics According to Duration of Heroin Use
Several recent studies investigated the human brain’s resting-state networks (Greicius et al. 2003; Fox and Raichle 2007; Buckner et al. 2008) and found that the correlated spontaneous activity occured during task-free conditions (Greicius et al. 2003; Ma et al. 2010). It was also found that the BOLD (Blood Oxygenation Level Dependent) signal fluctuations occur at low frequencies (0.01–0.08 Hz) (Fox and Raichle 2007). In addition, the resting-state networks in brain disorders were assessed at the level of systems integration during task-free conditions (Wang et al. 2006; Garrity et al. 2007). Further, resting-state fMRI is a validated approach for the study of disease. In previous studies, it was found that the decreased connectivity between the hippocampus and visual cortices has been observed in Alzheimer’s disease (AD) and abnormal resting-state bilateral frontoparietal, fronto-cingulate, and fronto-thalamic connectivity was observed in schizophrenia patients. The findings in previous studies revealed the pathological alterations in the integrity and efficiency of disease-related networks (Wang et al. 2006; Garrity et al. 2007). These findings help to understand the disease states and provide potential diagnostic information or treatment strategies. All of these studies suggest that resting-state fMRI might be a useful approach to the study of heroin addiction.
In previous studies, Liu et al. (2009) and Ma et al. (2010) observed addiction-related alterations in resting-state brain connectivity, while it is still unclear whether these resting-state connectivity alterations are specifically to chronic heroin use (Yuan et al. 2010a). In contrast, Yuan et al. explored the topological properties of resting-state brain networks in heroin-dependent individuals and hypothesized that network disruptions would occur in brain regions related to addictive behavior and stress regulation. In Yuan’s work, the brain functional connections of the heroin-dependent individuals and control subjects were compared with a graph theory analysis (GTA). This analysis thoroughly assessed the topological properties of networks by evaluating network strength as well as temporal and spatial interaction patterns. They also conducted correlation analyses between the statistical parameters (the degree, D, and shortest absolute path length, L) and the duration of heroin use in heroin-dependent individuals. They found that in heroin-dependent individuals, D values were higher in the bilateral orbitofrontal cortex (OFC), dorsolateral prefrontal cortex (PFC), rostral ACC (rACC) , precuneus, parahippocampal gyrus, putamen, and cerebellum; the left inferior frontal cortex, medial PFC, caudate, thalamus, and posterior cingulate cortex (PCC) ; and finally, the right middle temporal gyrus relative to control subjects (Fig. 3.2). Moreover, they observed positive correlation between the duration of heroin and the D value in the right parahippocampal gyrus, left putamen, and bilateral cerebellum. It was also observed that the duration of heroin use was negatively correlated with the L parameter in the same regions (P < 0.05) (Fig. 3.3).
Fig. 3.2
The degrees of chronic heroin users were different from the control subjects (Reprint with permission from Yuan et al. 2010a)
Fig. 3.3
The duration of heroin use was correlated with the regional topological properties (D value and L parameter) (Reprint with permission from Yuan et al. 2010a)
The areas of elevated D values in heroin-dependent patients were distributed in addiction-related circuits: the caudate and putamen of the reward circuit; the OFC and dorsolateral PFC of motivation/drive circuitry; the parahippocampal gyrus involved in memory; the rACC of the control circuit, consistent with a previous study (Fu et al. 2008); and the medial PFC representing stress regulation and emotional modulation. The findings of the abnormal topological properties might help to understand the underlying mechanism of heroin addiction.
The dorsolateral PFC integrates cognitive and motivational information from the OFC and contributes to regulatory processing (Groenewegen and Uylings 2000). In addition, there are studies which found that the OFC and dorsolateral PFC were important in the motivation-relevant processes of drug addiction (Goldstein and Volkow 2002; Goldstein et al. 2007). The parahippocampal gyrus is very important for memory encoding and retrieval. Memory is likely to influence the effects of drug use during intoxication (Shalev et al. 2002; Volkow et al. 2003). In addition, this area may be related to cue reactivity as mentioned earlier, where places, people, or other cues trigger an intense desire for drug use (Shalev et al. 2002). In previous heroin studies, hypoactivation of the rACC has been reported (control circuit) in heroin users (Fu et al. 2008) and cocaine users (Kaufman et al. 2003) during a GO/NOGO task, which indicated a role for the rACC in inhibitory control. Accordingly, the alterations observed in the work of Yuan et al. are consistent with what is known regarding the neurobiological basis of addiction and further inform the interconnectivity of relevant regions.
Yuan also found that the duration of heroin use is related to the topological properties (D and L) of several brain regions, including the bilateral cerebellum, right parahippocampal gyrus, and left putamen. The results suggest that heroin use has a cumulative effect on brain topology. An interesting interpretation of this result is that the information transfer of reward and memory becomes progressively more complicated with prolonged heroin use and probably leads to poor control and decision-making. These results agreed with a previous behavioral study which observed that the duration of heroin use was correlated to the performance in a stop-signal task (Monterosso et al. 2005). It can be concluded that the early intervention was crucial for the treatment of heroin addiction.
3.2.1.2 Spatial and Temporal Alterations in Resting-State Networks Related to Heroin Addiction
In Yuan’s work, spatial and temporal information was combined to investigate the resting-state networks changes in heroin-dependent individuals (Yuan et al. 2010b). In their work, the relationship between resting-state functional connectivity changes and duration of heroin use was investigated with correlation analysis. Yuan et al. employed the discrete cosine transform (DCT) (Fransson 2005) to assess the spatial distribution of low-frequency BOLD oscillations during resting state between the heroin-dependent individuals and healthy subjects. In addition, the temporal characteristics of brain regions that exhibited similar spatial patterns in DCT analysis were also investigated between the heroin-dependent individuals and healthy subjects. It was assumed that the resting-state functional connectivity of heroin-dependent individuals was changed and was correlated with duration of heroin use.
In comparison with healthy subjects, widespread spontaneous fluctuations in the resting state were detected in heroin-dependent individuals. Spontaneous signal changes were mainly found in the regions including posterior cingulate cortex (PCC)/precuneus , ACC, inferior parietal lobe, supplementary motor area (SMA), middle temporal lobe, and occipital lobe in heroin-dependent individuals and healthy subjects (Fig. 3.4).
Fig. 3.4
Discrete cosine transform analysis result (Reprint with permission from Yuan et al. 2010b). The regions of interest for the functional connectivity analysis (PCC/precuneus and rACC) were chosen from overlapping regions between heroin-dependent individuals and healthy subjects
In a later study, Fransson and Marrelec applied a partial correlation analysis and found that out of all the regions in the default mode network (DMN) , the PCC/precuneus was the only region connected with all other regions (Fransson and Marrelec 2008). The findings of Fransson and Marrelec might suggest that the PCC/precuneus played an important role in the DMN. Accordingly, in the work of Yuan, the PCC/precuneus was selected as the ROI to detect alterations in the PCC/precuneus network among the heroin-dependent individuals. In comparison, significantly reduced functional connectivity between the PCC/precuneus and right cerebellum (P < 0.05, corrected) was found among the heroin-dependent individuals. They also found that the resting-state functional connectivity between the PCC/precuneus and right cerebellum was negatively correlated with duration of heroin use in heroin-dependent individuals (Fig. 3.4). In conclusion, the DMN was thought to facilitate the retrieval and manipulation of past events for decision-making (Greicius et al. 2003), and it might be inferred that the decreased functional connectivity between the PCC/precuneus and right cerebellum in the DMN may have an effect on the decision-making in heroin-dependent patients.
3.2.1.3 Gray Matter Loss and Resting-State Abnormalities in Abstinent Heroin-Dependent Individuals
There have been only few studies investigating the gray matter density and the resting-state functional connectivity changes of heroin-dependent individuals. In the previous study, Yuan employed the voxel-based morphometry (VBM) to identify the decreased gray matter density in heroin-dependent individuals (Yuan et al. 2010c). In addition, Yuan also investigated the altered functional connectivity on resting-state fMRI among the heroin-dependent individuals (Yuan et al. 2010c). In comparison with the healthy control, significant reductions in gray matter density in the right dorsolateral PFC, left IPL, right fusiform gyrus, and left middle cingulate cortex (MCC) were observed among the heroin-dependent individuals (Fig. 3.5a). No significant increases in gray matter density were found. Moreover, it was found that the gray matter density of the right dorsolateral PFC was negatively correlated (r = −0.8359, P = 0.0013) with the duration of heroin use in heroin-dependent individuals (Fig. 3.5b).
Fig. 3.5
Voxel-based morphometry (VBM) analysis results (Reprint with permission from Yuan et al. 2010c). (a) The VBM analysis (b) correlation analysis results (r, correlation coefficient; p, P-value). DLPFC dorsolateral prefrontal cortex, IPL inferior parietal lobe, L left, MCC middle cingulate cortex, R right
In healthy subjects, activity in several brain regions was positively correlated with that in the right dorsolateral PFC, including the bilateral insula, putamen, caudate, MCC, IPL, ACC, and OFC, and the right thalamus and fusiform gyrus (P < 0.05, family-wise error rate [FWER] corrected). However, in heroin-dependent individuals, activity in the right insula, MCC, fusiform gyrus, and IPL was positively correlated with that in the right dorsolateral PFC (P < 0.05, FWER corrected). In comparison with the healthy subjects, significantly reduced functional connectivity between the right dorsolateral PFC and bilateral IPL (P < 0.05, corrected) was observed among the heroin-dependent individuals (P < 0.05, corrected) (Fig. 3.6a). It was also found that the functional connectivity between the right dorsolateral PFC and left IPL was significantly negatively correlated with the duration of heroin use (r = −0.7676, P = 0.0058) (Fig. 3.6b).
Fig. 3.6
Functional connectivity network results (Reprint with permission from Yuan et al. 2010c). (a) The connectivities of right dorsolateral prefrontal cortex and (b) the connectivities of heroin-dependent individuals were different from healthy subjects (P < 0.05, corrected). The correlation analysis was shown in the lower panel (r, correlation coefficient; p, P-value). ACC anterior cingulate cortex, DLPFC dorsolateral prefrontal cortex, IPL inferior parietal lobe, L left, MCC middle cingulate cortex, OFC orbital frontal cortex, P posterior, R right
The findings in the abovementioned study were consistent with previous studies (Lyoo et al. 2006; Yuan et al. 2009). Furthermore, a negative relationship between the gray matter density of the right dorsolateral PFC and the duration of heroin use further confirmed a cumulative effect of heroin use on the brain in an addiction-related region. Numerous studies have indicated that drug-related cues elicit significant activation of the dorsolateral PFC in users (Garavan et al. 2000; Bonson et al. 2002; Due et al. 2002). As indicated in previous study, the neurons in the dorsolateral PFC encoded reward expectancy during a delay, which was very important for the subsequent behavioral responses in rewarded tasks (Wallis and Miller 2003). It was also found that the impairments of the rat prelimbic cortex would hamper the acquisition and modification of contingency-guided behavior, indicating that this region is crucial for the cognitive control of goal-directed behavior (Balleine and Dickinson 1998). In the previous study, the dorsolateral PFC was found crucial for the decision-making tasks requiring the integration of cognitive and motivationally relevant information (Wilson et al. 2004). Thus, reduced gray matter intensity in the dorsolateral PFC was likely in part associated with impaired decision-making and goal-directed behavior dysfunction in heroin dependence (Lyoo et al. 2006; Xiao et al. 2006; Ma et al. 2010).
Several studies have reported that the IPL also integrated information from different sensory modalities to participate in higher cognitive functions (Caspers et al. 2006). For both humans and animals, the bilateral IPL is activated during working memory paradigms (Greicius et al. 2003). In addition, Park et al. reported significant activation in the bilateral IPL and right fusiform gyrus in response to alcohol-related cues among subjects with alcohol use disorders (Park et al. 2007). In the work by Yuan, the functional connectivity between the right dorsolateral PFC and bilateral IPL was found significantly decreased when compared with healthy control. In addition, the functional connectivity between the right dorsolateral PFC and left IPL was significantly negatively correlated with the duration of heroin use. These results suggested that persistent heroin use progressively degrades communication between the right dorsolateral PFC and left IPL in a manner possibly related to reported functional impairments in decision-making and cognitive control in these individuals (Xiao et al. 2006; Fu et al. 2008). It was also observed that the gray matter changes in the right dorsolateral PFC agreed with the resting-state functional connectivity alterations among the abstinent heroin-dependent individuals, which might indicate that the alteration in functional connectivity was of systems level (Xiao et al. 2006; Fu et al. 2008). While, it was noteworthy that the relationship of gray matter density changes to functional connectivity was still unclear. In the future, more work is needed to explore this issue.
3.2.2 Microstructural Abnormalities in Adolescents with Internet Addiction Disorder (IAD)
Worldwide use of the Internet has expanded incredibly in the last decade. The Internet provides remote access to other individuals and abundant information in all areas of interest. The IAD has been more and more prevalent and becomes serious society problem, which has attracted the attention of psychiatrists, educators, and the public. Due to the reason that the cognitive control and the impulse control of adolescents are relatively immature, they are at a high risk for addiction problems such as IAD. Though there have been studies suggesting that the IAD was related to gray matter changes, there have been few studies investigating the microstructural integrity of major neuronal fiber pathways in IAD, and almost no studies to date have assessed the temporal course of these changes.
In a more recent study, Yuan and colleagues used an optimized VBM method to investigate brain morphology in adolescents with IAD according to disease duration (Yuan et al. 2011). From previous studies, it was known that IAD subjects showed impaired cognitive control, and accordingly the authors hypothesized that long-term IAD would result in structural alterations in the brain associated with cognitive functional impairment. Assessing regional gray matter volume changes nonparametrically with an optimized VBM method and correcting for multiple comparisons using a cluster-based thresholding method, it was determined that IAD subjects had decreased gray matter volume in several clusters relative to matched control subjects; these regions were the bilateral dorsolateral PFC, SMA, OFC, cerebellum, and left rACC. Moreover, gray matter volumes of the right dorsolateral PFC, left rACC, and right SMA showed significant negative correlations with disease duration (r 1 = 20.73, P 1 < 0.005; r 2 = 20.74, P 2 < 0.005; r 3 = 20.65, P 3 = 0.005). No brain regions showed higher gray matter volumes with respect to matched control subjects (Fig. 3.7).
Fig. 3.7
Voxel-based morphometry results (Reprint from Yuan et al. 2011). (a) The gray matter volume of subjects with Internet addiction disorder (IAD) was reduced when compared to the control subjects. (b) Gray matter volumes of the dorsolateral prefrontal cortex (DLPFC), rostral anterior cingulate cortex (rACC), and supplementary motor area (SMA) were negatively correlated with duration of Internet addiction
As indicated in previous studies, long-term substance abuse (Kaufman et al. 2003; Li and Sinha 2008) and Internet addiction (Zhou et al. 2011) would bring about the cognitive control impairments. Numerous functional neuroimaging studies have revealed that the dorsolateral PFC and rACC were centrally involved in cognitive control as components of a specific cortico-subcortical circuit (MacDonald et al. 2000; Krawczyk 2002; Wilson et al. 2004). It was hypothesized that the occurrence of response conflict was signaled by the rACC, which would recruit the dorsolateral PFC for improved cognitive control. This important role of the dorsolateral PFC has been identified as an important top-down regulatory process (Vanderhasselt et al. 2009). In support of these hypotheses, recent neuroimaging studies have reported deactivation of the rACC in heroin and cocaine users during a GO/NOGO task (Fu et al. 2008; Vanderhasselt et al. 2009).
The OFC played an important role in the assessment of motivational significance for a given stimulus and the selection of behavior to obtain a desired outcome and contributes to the cognitive control of goal-directed behavior. For the reason that the OFC was extensively connected with the striatum and limbic regions (such as the amygdala), the OFC was suitable to integrate the activity of several limbic and subcortical areas associated with motivational behavior and reward processing (Groenewegen and Uylings 2000). As indicated in previous animal studies, the damage of OFC or the rat prelimbic would impair contingency-guided behavior, which might indicate that these regions are crucial for the cognitive control of goal-directed behavior. The SMA was critical for the selection of an appropriate behavior or inhibition of an inappropriate response. Indeed, the reported role of the SMA in both simple and complex GO/NOGO tasks suggested its importance in mediating cognitive control (Li et al. 2006; Li and Sinha 2008).
3.3 Mechanisms of Migraine
Migraine is an idiopathic headache disorder that is a significant source of individual and social burden and can lead to disability, losses of productivity, and a decreased overall quality of life (Terwindt et al. 2000; Li et al. 2006; Schwedt and Dodick 2009). Migraine attacks furthermore increase the risk of developing brain lesions in some regions (Tietjen 2004; Zaidat 2004). Recently, research interest has been directed toward the CNS damage and dysregulation associated with headaches and migraine. With the development of neuroimaging technology, the mechanism of migraine has transformed from a vascular disorder to a neurovascular disorder and, finally, to a central nervous system disorder (Schwedt and Dodick 2009). Accordingly, advanced neuroimaging approaches have been employed to investigate the structural and functional changes occurring in migraine patients (May 2009a; Schwedt and Dodick 2009).
3.3.1 Regional Homogeneity Abnormalities in Patients with Interictal Migraine
As indicated in previous studies, migraine without aura would lead to structural and task-related functional changes in the brains. While resting-state brain studies in particular have the utility to inform the pathophysiology of migraine, there have been only few studies focusing on the resting-state abnormalities in patients with migraine without aura. In the work of Yu, the local features of spontaneous brain activity in patients with migraine without aura were analyzed by the regional homogeneity (ReHo) method (Yu et al. 2012). Firstly, one-sample t-test was used to extract the ReHo results across all the subjects. Then, two-sample t-test was applied to compare ReHo between the patients with migraine without aura group and the healthy control group (P < 0.05, FWE corrected). Yu observed significantly decreased ReHo values in the right rACC, PFC, OFC, and SMA in the patients with migraine without aura when compared with (P < 0.05, FWE corrected). In the correlation analysis, the duration of migraine was found significantly negatively correlated with the average ReHo values in the right PFC (r = −0.5032, P = 0.0088) and rACC (r = −0.4306, P = 0.0281) (Fig. 3.8). However, no relationships were found between resting-state properties and the average pain intensity or attack frequency in migraine patients.
Fig. 3.8
Altered regional homogeneity in migraine (Reprint with permission from Yu et al. 2012). (Left) Migraine-related changes in regional homogeneity (ReHo) shown as a comparison of Kendall’s coefficient of concordance (KCC) maps between patients with migraine without aura and control subjects (P < 0.05, corrected) during the resting state. (Right) The average ReHo values of the right prefrontal cortex (PFC) were significantly correlated to the disease duration
Both experimental and clinical studies have suggested that affective responses to pain such as unpleasantness and suffering were principally integrated in the rACC (Vogt 2005; Mechias et al. 2010; Shackman et al. 2011). In addition, the rACC was involved in pain control mediated by the endogenous opioid system (Petrovic et al. 2002; Wager et al. 2004). It might be inferred that the functional disruption in the rACC of patients with migraine without aura might be the underlying mechanism of the functional impairments in long-term pain affective responses and endogenous analgesia.
The PFC was another important region in opioid analgesia and other forms of pain modulation (Kupers et al. 2000; Petrovic et al. 2002). In particular, the PFC may play a role in suppressing pain perception via cognitive control mechanisms (Lieberman et al. 2004; Wiech et al. 2008). Aderjan et al. used fMRI to study the differences between the patients and healthy control subjects stimulated daily with trigeminal pain, 20 min per day for 8 consecutive days. The fMRI was obtained in day 1, day 8, and 3 months later. No difference on the behavioral pain ratings was observed between the two groups. While, they found opposing activity changes in several brain regions involved in endogenous pain control. The brain activity in PFC and rACC was found increased in healthy control subjects while decreased in the patients with migraine. The findings might indicate reduced efficiency in pain processing in patients with migraine without aura. In the correlation analysis, they also found that negative relationship between the ReHo values in the rACC and PFC and the duration of disease.
As indicated in previous studies, the OFC was involved in the behavior associated with sensitivity to reward and punishment, including sensory integration, decision-making, expectation, and planning (Bechara et al. 1994). Both positive (pleasure) and negative (pain) stimulation has been shown to elicit opioid release in the OFC (Phillips et al. 2003). Accordingly, all received sensations were likely to be modulated and assigned affective responses by the OFC (Vincent et al. 2003). In particular, the OFC seemed to be related to the responses based on the previous reward value need to be inhibited (Elliott et al. 2000). Similarly, the SMA seemed to be crucial for linking cognition to action (Picard and Strick 1996; Rushworth et al. 2004). The pre-SMA was involved in executive control (Curtis et al. 2005; Nachev et al. 2005; Aron and Poldrack 2006; Li et al. 2006), pain anticipation (Ploghaus et al. 1999; Wager et al. 2004; Roy et al. 2009), and the affective component of pain (Apkarian et al. 2005). The pre-SMA was also activated in the tasks requiring the inhibition of responses and switching between rules for action consequences (Isoda and Hikosaka 2007). Taken together, the observed decreased ReHo values in the OFC and SMA (particularly the pre-SMA) of patients with migraine without aura in the study by Yu et al. may have been related to deficits in affective pain modulation and affective pain response inhibition.
3.3.2 Gender-Related Differences in Resting-State Networks Dysfunction in Migraine
Migraine is known as gender specific, which is more prevalent in women than men. However, the underlying dysfunctional brain organization accounting for the gender-specific characteristics of migraine is little known. Liu et al. used resting-state fMRI to evaluate resting networks that were conceived as neurocognitive entities incorporating both local and global processes (Liu et al. 2011). Hypothesized differences in the topological properties of these networks were analyzed using a GTA, and functional brain networks were constructed to characterize interregional relationships, small worldness, network resilience, and node centrality in the brains of migraine patients and healthy control subjects. Importantly, networks were constructed using automated anatomically labeled template images (Tzourio-Mazoyer et al. 2002), which divided the whole brain into 90 regions of interest (ROIs). Based on the 90 ROIs as a set of nodes, the mean time series in each ROI were extracted and used to construct a 90 × 90 matrix of Pearson correlation coefficients for all possible node pair connections. The constructed network was used to calculate interregional connections, and the disruption in the topological properties of these networks was analyzed using a two-way ANOVA performed according to model gender (male vs. female) and disease state (patient vs. control) effects simultaneously. In the results, several pairs of connections were significantly altered in migraine patients with respect to control subjects (P < 0.05, false discovery rate corrected). In male migraineurs, connections were significantly increased in the PFC, supramarginal gyrus, amygdala, HIP, IPL, and temporal lobes (Fig. 3.9). The brain regions showing significant increased interregional correlations in female migraineurs included the orbital PFC, posterior cingulate gyrus, parahippocampal gyrus, cuneus, putamen, caudate, parietal lobule, temporal lobes, and occipital cortex (Fig. 3.10). No connections were found significantly decreased in migraineurs with respect to control subjects.
Fig. 3.9
Prospects of Acupuncture Research in the Future
Findings of Acupuncture Mechanisms Using EEG and MEG
Early fMRI Studies of Acupuncture
Temporospatial Encoding of Acupuncture Effects in the Brain
Minimally Invasive Posterior Surgical Approaches to the Lumbar Spine Through Tubular Retractors
Mild Traumatic Brain Injury: An Overview
Significant differences in resting-state functional connectivities in male migraineur patients (PM) versus healthy control subjects (HC) (Reprint from Liu et al. 2011). The red lines indicated increased intensity in male PMs (P < 0.05, corrected). AMYG, amygdala; HES, Heschl gyrus; HIP, hippocampus; IPL, inferior parietal lobe; ITG, Inferior temporal gyrus; MFG, middle frontal gyrus; MTG, middle temporal gyrus; OLF, Olfactory cortex; PCUN, precuneus; SMG, Supramarginal gyrus
Related posts:
Prospects of Acupuncture Research in the Future
Findings of Acupuncture Mechanisms Using EEG and MEG
Early fMRI Studies of Acupuncture
Temporospatial Encoding of Acupuncture Effects in the Brain
Minimally Invasive Posterior Surgical Approaches to the Lumbar Spine Through Tubular Retractors
Mild Traumatic Brain Injury: An Overview
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
