Pnas: amygdala individual specific functional connectivity: Fundamentals of precision psychiatry

  The amygdala plays a central role in the pathophysiology of many mental disorders . However , The understanding of how the amygdala fits into the larger network of the human brain is not accurate , It limits our ability to create dysfunction models in individual patients to guide personalized treatment . therefore , We are 10 Highly sampled individuals ( Everyone 5 Hours of fMRI data ) The location of amygdala and its functional subdivision in brain network organization were studied . We subdivided the amygdala of each person into three functions . We found that , A partition priority is related to the default mode network ; The second priority is related to dorsal attention and frontoparietal neural network ; The third sub partition does not have any network associated with the other two sub partitions . These three branches are positively correlated with ventral attention and somatic motor network , It was negatively correlated with significance and buckle cap network . These observations are in an independent 120 Individual group datasets are replicated . We also found that , The distribution and size of the functional connections between the amygdala and the cerebral cortex are related to the individual differences in the amygdala subdivision and the stereotactic position of the cortical functional brain network . Last , Using lag analysis , We found a subdivision relative to the amygdala , The time sequence of fMRI signals in the cortex is consistent . All in all , This work provides a detailed amygdala - A framework for cortical interactions , It can be used as the basis of a model linking the abnormal amygdala junction of individual patients with mental symptoms .

1.  brief introduction

      Mental illness is the leading cause of incidence rate and mortality worldwide . In the past few decades , Little progress has been made in reducing this burden , Part of the reason is that we lack personalized mental illness Modeling Technology , It can be used for individual patient diagnosis and guiding treatment . The amygdala is a structure of the medial temporal lobe , Personalized models of any mental illness are important . Functional magnetic resonance imaging (fMRI) Measuring the functional connectivity of the amygdala has been widely associated with the symptoms of many different mental disorders 、 Longitudinal course of disease is related to treatment response . However , A major obstacle in translating these research results into clinical biomarkers is , The role of the amygdala in the organization of the larger functional networks of the individual brain is poorly understood . result , Our ability to create individual patient function and dysfunction models to guide personalized treatment is limited .

      A basic organizing principle of the human brain is , It can be divided into 10 To 20 A large-scale 、 Distributed functional brain network . Functional magnetic resonance imaging (fMRI) Research has established the connection characteristics between and within these large-scale brain networks in the adult brain , This is through slow and low activity between brain regions (ISA, <0.1 Hz) To assess the relevance of . Besides , Information about the time direction of connections within and between networks , By calculating the difference between two brain regions ISA Time offset to provide , These two regions maximize their relevance .

      All in all , These tools have built an increasingly complex model of the human brain network , among ISA Propagate within and between networks in a specific directional pattern . Changes in the intensity and timing of these associations are associated with the risk of mental illness .

      The biology and function of the amygdala is determined by its position in the network hierarchy . A detailed description of the basic amygdala circuit leads to a model , In the model , The lateral amygdala integrates sensory input with current physiological status , The central amygdala sends a wide range of outputs , To guide appropriate behavior and physiological responses . Especially in rodents , Specific cortical connections that drive the amygdala to respond to external stimuli have been identified , And up and down regulation in these responses . Besides , Studies of rodents and non-human primates have shown that , The changes of individual behavior response to threat stimuli are related to the changes of amygdala biology and cortical connections .

The interruption of functional connections in the amygdala may be the core of mental illness . However , Little is known about the functional connections of the individual amygdala , This limits our ability to understand and treat individual amygdala junction disorders . ad locum , We will 10 The individual amygdala is divided into three parts , And use everyone 5 Hours of fMRI data to define the connection mode . We prove , In the individual , Each of the three amygdala divisions occupies roughly the same position , And it shows consistent functional connection with specific cortical functional network : A network to default mode , The other goes to the dorsal side to pay attention to the network , The third one has no priority connection .

2. result

2.1 Individualized amygdala zoning

      Standard group based studies usually use probabilistic structural templates to divide the amygdala into three partitions , The medial center corresponding to the mean anatomical position of the group 、 Basolateral and superficial amygdala divisions .

      We are in the picture 1A This publicly available template is described in . Please note that , In the initial Amunts In the study of et al , The size and spatial extent of these partitions are highly variable . therefore , These probability templates may not accurately define the location of all individual amygdala subdivisions .

To solve this problem , We developed a method , Connect through functions , stay 10 Three amygdala subdivisions are defined in each individual . The subareas are marked according to the network with the highest positive functional connectivity relative to the other two subareas , As follows .

      Empirically defined individual amygdala subdivisions ( chart 1B) It is similar to the publicly available amygdala partition in mean position and inter individual variation ( chart 1A). Please note that , Despite the experience 、 Functionally defined subdivisions are on average related to probability 、 Anatomically defined subdivisions overlap , But there were significant variations among subjects ( chart 1B and SI appendix , chart S8). This cross subject variation is related to Amunts The cross subject variation of cell structure definition partition in the study of et al ( chart 1A).

      One potential consequence of this difference is , Apply a common template to all individuals , The amygdala subdivision will be wrongly labeled in many individuals . therefore , The subdivision network relationship of the cortex may be covered up , This possibility will be explored in detail below .

      We use the same empirical procedure to define WU Three subdivisions in the population average dataset , And three similar subdivisions are obtained ( chart 1C). Please note that , The spatial layout of these three subdivisions is similar to Amunts The probability partition of et al , But it's not exactly the same .

chart 1 The amygdala division defined in an individual according to the connection pattern with the cortex is similar to the amygdala division defined in an individual according to the cell structure

2.2 Group average functional connectivity for probabilistic amygdala partitioning

        We started with 120 A group average data set of individuals (WU Data sets ) in , Yes amts Each of the three probabilistic amygdala partitions derived by et al. Performs the standard method of calculating cortical connectivity . The connection pattern produced by the standard template method is similar to the previous investigation ( chart 2). The connections between amygdala and cortex in the three probability zones are highly similar , A pattern of cortical connections between paired partitions Pearson Relevance in 0.88 To 0.92 Between .

chart 2 The three amygdala probability zones are highly similar to the cortical connection patterns .

2.3 Functional connections of individualized amygdala partitions

        Next , We calculated each MSC The cortical functional connectivity of three experiences and separately defined amygdala subdivisions in an individual ( chart 3A and SI appendix , chart S3). Each empirically defined subdivision has a unique cortical network connection pattern ( chart 3B). Name the subdivision area according to the network , Compared with the other two subdivisions , The subdivision region has the highest positive connectivity ( chart 3C). The default mode is subdivided with DMN The degree of positive correlation is higher than the other two . On the back side, pay attention to the partition and DAN It has a high positive connection with the frontal parietal network . An unspecified subdivision region has no unique positive correlation with the specific network of other subdivisions , It only has the connectivity attribute shared by three subdivision regions . Relative to the default mode ( Average pairing correlation = 0.45) And dorsal attention ( Average pairing correlation = 0.35) To subdivide , The whole brain connection pattern without specified subdivision had the lowest similarity among individuals ( Average pairing correlation = 0.29).

      In addition to each subdivision having unique features connected to the cortex , All three subdivisions share many features connected to the cortex ( chart 3B). for example , The activities of all three branches are related to ventral attention (VAN) And body movement (SMN) The activity of the network is positively correlated , And with the buckle - Operculum (CON) And significance (SN) The activity of the network is negatively correlated . These patterns in individuals MSC participants ( chart 3B and SI appendix , chart S8) And group average WU Data sets (SI appendix , chart S5) It is obvious in .

chart 3 A separate subdivision of the amygdala indicates , The amygdala has a fine partition selectively associated with DMN Functional connection , And another fine partition is selectively associated with DAN Functional connection .

2.4 The functional connections of the amygdala respect the boundaries of the brain's functional network

        Next , We tested whether group average functional boundaries and individual specific functional boundaries can better capture the network specificity of amygdala subdivision functional connections . In this analysis , We chose to evaluate amygdala default pattern segmentation with medial prefrontal cortex (mPFC) Region of interest defined in (ROI) Functional connection between , Because a lot of previous work emphasized the amygdala mPFC The role of connectivity in mental illness . chart 4A The default pattern of amygdala of each individual is described and subdivided into Talairach In space mPFC[0 33 0] Define for the central public group ROI Functional connectivity of . such ROI From the amygdala mPFC Meta analysis linking studies related to internalization symptoms . Between individuals , Even with personalized amygdala default mode partitions , Derived from the group mPFC There was also no significant difference in regional connectivity .

      Next , We measured the individual specific amygdala default pattern subdivision function connection and mPFC Individual specific function definition location in ( chart 4B). In all 10 Of individuals , Amygdala default mode subdivides connectivity with mPFC Of DMN Partial positive correlation , And mPFC Of SN Partial positive correlation . although mPFC Inside DMN There is a moderate overlap between individuals ( Average Dice The coefficient is 0.61), but SN The position of is highly variable among individuals ( Average Dice The coefficient is 0.07). Pictured 4B, In a large area of positive connectivity ( Mainly corresponding to DMN Individual specific location of ), Most individuals have negatively connected islands ( Mainly corresponding to SN Individual specific location of ). Because of everyone's SN The location is different , The default mode of amygdala subdividing the connection to any specific stereotactic position is highly variable , And depends on the individual specific mPFC Network layout . These analyses show that , stay mPFC in , The connectivity of a particular stereotactic location is variable among participants , Because different people have different networks in these locations . This phenomenon is in SI appendix ( chart S9) Has been studied more extensively .

chart 4 The functional connection between amygdala and cortex is related to the specific functional network boundary of individuals

2.5 The temporal relationship between the amygdala and the cortex

        Lag analysis was used to explore the most robust cortical amygdala functional connections ( chart 5A; For statistical standards, see SI appendix ). stay SI appendix S3 and S4 A complete list of all detected hysteresis relationships is provided in ; ad locum , We summarize the results through segmentation and functional brain networks .

        adopt MSC and WU Data sets , The subdivision of the amygdala occupies a consistent time position in the larger network organization of the human brain . say concretely , The default mode of amygdala segmentation fMRI Activities precede VAN and DMN Of mPFC Some activities . contrary , The breakdown of the default amygdala pattern lags behind DMN The rest of , Such as lateral parietal cortex (LPC). stay FPN in , The default pattern of the amygdala is subdivided before fMRI Activities , But in SN、CON And parietal lobe memory (PMN) In the network , The breakdown of the default amygdala pattern lags behind fMRI Activities ; but fMRI Activity is negatively correlated between the default mode subdivision and each network . The dorsal amygdala attention zone lags behind DAN, It also lags behind the anti - related SN and PMN. Last , Unspecified amygdala partition lags behind CON、SN and PMN, These partitions are all negatively related to this partition . The connectivity and lag relationship between each amygdala partition and the cortical network are shown in the figure 5 and SI Appendix, Fig. S10 Shown .

chart 5 The delay analysis revealed that relative to the cortical network , In each amygdala division ISA Time sequence of

3.  Discuss

      In this study, a personalized functional connectivity estimation method was used to characterize the amygdala and its subdivision as a part of the overall functional network of the brain . We found that , The subdivisions of the three amygdala occupy roughly the same position in the subjects , And it shows consistent functional connection with specific cortical functional network . say concretely , We describe a division of the amygdala in a superior position for most individuals , And with DMN Have priority function connection ; The second amygdala is located on the medial side of most people , And DAN Have priority function connection ; The third amygdala branch is located on the ventral side of most people without any network .fMRI Activity and VAN and SMN There is a positive correlation between the activity of , And in CON and SN There is a negative correlation between the activity of . In two separate datasets (WU and MSC) in , We detected a consistent temporal relationship between each amygdala subdivision and the cortical functional network . It is worth noting that , The stereotactic positions of amygdala subarea and cortical functional network were different in different subjects . therefore , The connectivity of amygdala was measured by the method based on standard template , Different functional connections are often captured in different individuals . In addition to informing the basic biology of the amygdala , This work provides a framework , Development mechanism , Biologically plausible models , Individual patients with amygdala function and dysfunction .