Based on an economical assumption, our new Unified Structural and Functional Connectivity (USFC) modeling represents the first effort to build a brain’s effective “traffic map” highlighting the brain’s major structural pathways that are most heavily used for efficient functional signal transferring. Based on this model, we revealed a highly skewed brain traffic system featuring the subcortical, the default-mode, and the salience network housing some of the brain’s most traversed nodes and a medial frontal-caudate-thalamus-posterior cingulate-visual cortex midline “backbone” corridor as the mostly heavily used structural pathways. Moreover, the finding that stronger structural connections are underlying stronger negative functional connections further supports the functional roles of negative FC and provides a fresh perspective on the dynamic interactions among brain regions. Finally, the significantly higher efficiency, modularity, and betweenness centrality demonstrated in the USFC map when compared with structural and functional connectomes may support the superiority of this “traffic map” in potentially revealing the true working mechanism of the human brain. Overall, the proposed USFC model opens a new window for brain connectome modeling and provides a considerable leap forward in brain mapping efforts by offering a more intricate depiction of the brain's connectivity landscape.
Our analysis uncovered an striking pattern within the brain's USFC blueprint: the DMN regions collectively possess the third highest nodal USFC values while more strikingly, seven of the top ten most heavily trafficked pathways involve either the PCG or medial prefrontal cortex, the two hub regions of the DMN44. Centrally located and occupy a large portion of the brain, the DMN is known for being “active” during rest and its versatile roles in self-reference, social cognition, episodic and autobiographical memory, language, sematic memory, among others45–47. All these functions involve complex communications within and between DMN and other brain regions which likely underlies our finding of its central role in the newly defined USFC system. Specifically, the prominent inter-network connections between the DMN hubs and subcortical/visual regions as shown in the top ten USFC pathways likely underscore the DMN’s potential integrative role across different domains, which is highly in line with findings demonstrating DMN’s active and dynamic reorganization of its connectivity patterns across a range of cognitive and socioemotional tasks48–51. This finding provides another critical piece of evidence from a global brain “traffic map” perspective that the DMN's role likely goes beyond a passive default state but rather globally contributes to the brain’s efficient signal processing across task domains49,50. Overall, our finding of the central role of the DMN in the newly defined USFC system provides new support/explanation for its established importance in development50,52, normal adult functioning48–51,53−55, aging56,57 and various brain disorders58–61.
The importance of subcortical/salience networks in USFC and midline “backbone” corridor
Beyond DMN connections, six of the top-ten most heavily trafficked segments involve the thalamus/caudate while at a network level, the subcortical and salience network regions collectively rank as the two mostly traversed networks in the whole brain “traffic map” ranking (Fig. 2). Regarding the salience network, although not highlighted in the top ten mostly heavily used pathways, its regions collectively rank second in the whole brain traffic map system and the middle cingulate cortex was detected as one of the “outliers” with the highest USFC loadings. These findings are consistent with its reported role of lying on the apex of the brain’s global coordination system by performing a “switching” role among large scale functional networks, especially between the DMN and dorsal attention networks48,51,62,63.
The subcortical regions, in particular the thalamus's prominence in this traffic system is consistent with not only its known role as an “relay center” connecting peripheral neural system with the brain cortices but also its versatile involvement in modulating and refining sensory data, shaping consciousness, and enhancing cognitive functions64–66. Its highly utilized connectivity with the PCG may be particularly indicative of a sophisticated mechanism that merges external sensory inputs with internal states, an essential process for coherent cognitive function67. Similarly, the caudate nucleus not only plays a critical role in movement planning and execution but also serves in a multitude of essential brain functions, including learning, memory, reward, motivation, emotional regulation, and aspects of romantic interaction68,69. Structurally, frontal regions are known to be connected to the caudate, which in turn is connected to the thalamus, and subsequently projecting to PCG, providing SC support for the observed medial frontal-caudate-thalamus-posterior cingulate -visual pathway that leads the most heavily USFC segments. The finding of a clearly defined midline corridor connecting frontal to caudate to thalamus to posterior cingulate and finally to visual cortices supporting the most “traffic” in the brain through USFC modeling is striking and opens up new windows for better understanding of the “backbone” structure of the brain’s global communication system. Consistent with our findings, Hagman et al have previously delineated the SC hubs of the human brain and similarly detected a midline “structural core” linking precuneus to posterior, middle, anterior cingulate cortex and finally to medial orbital frontal cortices4. However, their examinations exclude subcortical areas so the potential “bridging”/ “disseminating” (e.g., the thalamus) role of subcortical regions were not counted for. With combined consideration of both functional and SC and including both cortical and subcortical regions, the midline corridor delineated in this study featuring frontal-subcortical-parietal-occipital links may have better captured the “backbone” of the brain’s global communication system and deserves more attention in future search of its relevance in health and disease.
The intriguing finding of strong structural underpinnings of negative FCs.
The finding of moderate but significant positive correlations between SC and FC strengths associated with positive FCs is in line with previous reports24,70. However, the finding that routes underpinning negative FCs show a robust negative relationship between SC and FC strengths across one-to-three step connections is more intriguing. Ongoing debate regarding global signal regression and the consequent observation of negative correlations (anti-correlations), underscores the lack of consensus on a singular method for processing resting state data to uncover the 'true' nature of brain functionality71. Contrary to the notion of negative FC as a mere byproduct of signal processing, emerging research posits it as a salient aspect of the brain's functional architecture defining modularity of the resting-state fMRI connectome, deeply linked with its structural framework12,72–77. Our findings add to the evidence supporting the functional significance of negative FCs after global signal regression and suggest that the brain utilizes a delicate traffic system to choose the best routes (i.e., composed of segments with stronger SC) for negative interactions across different brain regions. Notably, Skudlarski et al. indicated that regions with negative functional FC are not necessarily disconnected structurally78. Instead, there is an implication of a complex relationship where structurally close regions can exhibit negative FC, suggesting an intricate coordination of brain dynamics. However, we have to point out that the “one-step” route delineated in this study should not be confused with “direct SC” or “connected by a single white matter bundle” give the limitation of diffusion-weighted imaging-based tractography. In other words, the one-step SC used in this study was derived based on probabilistic tractography and as long as there is a “connected structural route” connecting two brain regions, we define these two regions are “structurally connected” and treat them as “one-step” connections. It is possible that multiple white matter fiber bundles are underlying each of these “one-step” structural connection and the accumulated phase lag across the multiple structural connections may have contributed to the observed negative FC79. Compared with the relationships associated with negative FCs, where all three step groups (i.e., 1–3) show significant negative correlations, the relationships associated with positive FCs only show positive relationships for 1-step route. One potential explanation could be that choices for multiple-step positive FCs are more abundant than those for negative FCs and SC is not necessarily a limiting factor, and the choices are not as tightly regulated, resulting in weaker SC-FC correlations. Regardless, the finding that stronger structural routes are underlying stronger negative FCs provides further support for the importance of negative FCs in the brain's efficient/effective communication and functioning.
The USFC-based connectome demonstrates significantly higher communication performance than both the FC and SC systems.
For all three measures of the brain system communication effectiveness, namely global efficiency, modularity, and betweenness centrality, the USFC-based connectome demonstrates significantly higher performance than both the FC and SC systems. These findings support the potential superiority of the USFC system in depicting the brain’s signal transferring efficiency. Essentially, only looking at the “road system” (i.e., equivalent to the brain’s SC system) or the final “number of people traveling between any two cities” (i.e., equivalent to the brain’s FC system) could not provide a clear picture of the brain’s “traffic patterns” while it is this traffic pattern that directly unveils how the road system effectively work to support the between-city travelling (i.e., signal transferring). The much higher global efficiency and betweenness centrality is likely supported by the highlighted most heavily utilized routes between major functional works while the higher modularity may result from the more densely connected local systems within USFC.
Although this work provides a new perspective on brain connectome modeling, there are several major limitations associated with the current version of USFC that deserve future improvements. First, we made the economic assumption (i.e., shorter distance and stronger SC) for route selection but the “real-time traffic” is not considered in this formula. In other words, future improvement could further consider the current “traffic” along each route (i.e., real-time modeling of the “dynamic” FC80) in determining the optimal route between two brain regions. Second, as mentioned above, direct structural connection in this study might not represent one single fiber bundle the 1-step routes may consist of multiple white matter fiber bundles, which bears critical implications on the understanding of SC-FC relationships, particular those with the negative FCs. Finally, we used average FA along the tracts to index SC strength but there are other metrics too (e.g., number of fibers) worth further consideration.
Overall, the USFC model presents a compelling new framework to model the brains “effective connectome” and opens a new window for future research aimed at deciphering the enigmatic principles that govern the brain's efficient communication system. By highlighting the “most-heavily-used brain pathways/networks” in its current version and pursuing continued efforts to refine/navigate this complex "traffic" in both normal and diseased populations, the implications from this new model may reach far into the realms of neuroscience, with the potential to transform both theoretical models and clinical/intervention approaches.