For many years, the method of understanding the function of a specific brain region was through the study of focal brain lesions that occurred as a result of strokes, tumors, and hemorrhages. If patients with similar symptoms had overlapping lesions in a specific brain region, we could pinpoint those neurological symptoms or behavioral deficits to that region (1, 2). Traditional lesion mapping has been the cornerstone of clinical neurology, providing valuable insights into the functional roles of different brain regions. However, most neurological and psychiatric symptoms cannot be traced back to a single region; instead, they involve a network of interconnected regions (3). Additionally, patients exhibiting similar symptoms may have lesions located in diverse areas, which poses a challenge to traditional lesion mapping in pinpointing those symptoms (4). Such symptom overlap could potentially be due to the disruption of an underlying, cohesive brain network (5, 6).

One strategy to address this challenge is through the implementation of lesion network mapping (LNM), a methodology that connects brain lesions to either functional or structural networks to identify common brain circuits tied to diverse lesion locations (4, 7).

This technique operates on the hypothesis that a lesion present at any location within a network mapped for a specific symptom has the potential to trigger that symptom (4, 8). The process of performing LNM overlap studies comprises four stages: Firstly, lesions are traced into a standardized brain atlas, establishing a foundation for connectivity analyses (9) (Figure 1). Secondly, an assessment is undertaken to determine the connectivity of each lesion location with other brain regions, using either structural or functional normative human connectome data. Thirdly, correlations between lesion locations and all other brain voxels are thresholded to delineate a map of interconnected regions. Finally, these maps from each patient are superimposed to pinpoint brain regions that are most frequently connected to lesion locations associated with the symptom in question (10). In addition, certain studies incorporate voxelwise statistical analyses, either by leveraging continuous outcomes or by contrasting patients with controls, with the aim of elucidating common networks associated with symptomatic lesions (11, 12). The availability of large-scale functional and anatomical normative maps, like those offered by the Human Connectome Project data, provide a robust foundation for correlating lesion locations with a shared network, thereby facilitating the study of neurological and psychiatric symptoms (13).

This methodology has been broadly utilized since its introduction in 2015 for the localization of an array of neuropsychiatric, behavioral, and movement disorders (8, 10, 14–18). Previous investigations have substantiated the validity of LNM; while several outcomes confirmed primary hypotheses concerning the neuroanatomical underpinnings of specific symptoms (19), some unexpected findings also emerged from LNM studies. For instance, a notable association between the putamen and hemichorea-hemiballism was identified, despite earlier evidence suggesting the involvement of the subthalamic nucleus in the genesis of hemichorea-hemiballism (20). As such, LNM may offer an innovative avenue to further investigate the functionality of brain networks.

The intent of this systematic review is to spotlight recent studies that have utilized LNM to map neurological and psychiatric symptoms, thereby providing insight into how this methodology enhances our comprehension of the function of distinct brain regions.

This systematic review was conducted in adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (21).

This review consolidates and critically assesses existing knowledge on the application of Lesion Network Mapping (LNM) methods in the diagnosis and treatment of neurological and psychiatric disorders. Our findings underscore the complementary role of functional and structural LNM, emphasizing their potential utility in improving therapeutic outcomes in neurological afflictions. Our analysis draws attention to certain brain regions and networks integral to specific neurological domains. For instance, the fronto-parieto-temporal network emerged as pivotal to cognitive functioning, while the dorsal anterior cingulate cortex and the left dorsolateral prefrontal cortex demonstrated considerable relevance for obsessive–compulsive disorder. These insights indicate that LNM may aid in pinpointing viable targets for neuromodulation interventions across various neurological disorders. The examined studies also underlined the indispensable role of LNM in deciphering the intricacies of motor and speech disorders. These studies highlighted the use of LNM in identifying brain regions and neural pathways implicated in diverse movement disorders, such as Parkinsonism, Essential Tremor, and Holmes Tremor. Consequently, we can anticipate LNM’s instrumental role in informing neuromodulation strategies tailored to specific motor and speech disorders, such as cervical dystonia and hemichorea hemiballismus. Our results suggest that the choice of LNM method should align with the specific type of functional deficit under consideration. Since the advent of LNM studies in 2015, numerous investigations have successfully employed this technique to elucidate how alterations in functional and structural networks can account for symptoms post focal brain lesions.

Moreover, our analysis identified several overlaps between different regions and networks implicated in diverse psychiatric and neurological disorders. Both structural (51) and functional LNM (24, 29) pinpointed the fronto-parieto-temporal network as integral to cognitive function, thereby asserting a potent correlation between this network and cognitive impairment. Furthermore, studies revealed that damage to the right ventrolateral prefrontal and right temporal cluster (47), coupled with alterations in the left retro splenial and right frontal cortex (7), could precipitate spatial perception deficits and engender delusions of space. Intriguingly, both functional LNM (54) and structural disconnectome-based analyses (57) associated the left dorsolateral prefrontal cortex with depression. This concurrence insinuates a common neural foundation for depressive symptoms, stressing the necessity of considering both structural and functional connectivity when attempting to comprehend the neurological underpinnings of varying disorders and enhancing treatment outcomes. The complex neural networks implicated in these disorders can guide the formulation of tailored treatment strategies and facilitate a deeper understanding of their neurobiological origins. By delineating these overlaps and parallels, we underscore the interwoven nature of the brain and its instrumental role across an array of disorders.

Our study unearthed a significant overlap within the thalamus for essential tremor relief, Holmes tremor, and executive function deficits. This intriguing finding challenges the conventional practice of symptom localization that ascribes particular symptoms to specific, isolated brain regions. Instead, it accentuates the importance of investigating the complex interactions among various brain regions and acknowledges the distributed nature of neural processing (4, 24). This shared neural substrate within the thalamus proposes potential unified therapeutic targets for these conditions, underscoring the imperative for a more holistic approach when studying neurological and psychiatric disorders. By harmonizing the strengths of both conventional localization and network-based perspectives, researchers can attain a comprehensive understanding of the brain’s structure and function, ultimately leading to more effective treatment strategies for a multitude of disorders.

The neurobiological mechanisms underlying brain disorders and the localization of brain function are traditionally explored by identifying abnormalities within the brain and contrasting them with matched controls, as exemplified by the study of hemiparesis following stroke lesions. However, this process becomes complex when the lesion is located in an unanticipated region or when numerous heterogeneous lesions are observed in disparate locations. For instance, hemiparkinsonism patients with varied causal lesions outside the nigrostriatal tract were found to map onto a common network (19). Moreover, most neurological and psychiatric symptoms cannot be ascribed to a single region; instead, they involve a network of interconnected areas. As such, LNM has proven to be a significant advance over traditional lesion analysis, enabling the localization of symptoms across different lesion locations, a task that was previously unachievable (60).

Focal brain lesions have been observed to modify resting functional connectivity and reduce the variability of neural states, thereby limiting the brain’s ability to process information (59, 61, 62). Empirical evidence from previous studies has indicated that alterations in functional connectivity following focal brain lesions are not restricted to a singular network but engage numerous regions (63). Strokes tend to affect white matter and subcortical regions more often than the cortex. Given that white matter contains numerous fiber pathways, strokes within these regions can result in widespread alterations (59, 64).

The pioneering study employing LNM was conducted by Boes et al. (4), investigating peduncular hallucinosis subsequent to subcortical lesions. These lesions were hypothesized to disrupt the extrastriate visual cortex, despite their heterogeneous locations. The researchers found that 22 out of 23 lesions demonstrated a negative correlation with the extrastriate visual cortex. Moreover, a study by Kim et al. discovered that lesions causing hallucinations localized to a shared brain network, encompassing the cerebellar vermis, inferior cerebellum, and the right superior temporal sulcus (5). Following the promising results of this inaugural LNM study, the method was applied to elucidate long-standing neurological enigmas characterized by heterogeneous lesions dispersed across different regions (6, 18, 31).

Considering the comparable effects of deep brain stimulation (DBS) and therapeutic lesions, a promising avenue for LNM involves leveraging the identified regions associated with disorders or common networks linked to beneficial brain lesions as therapeutic targets for DBS (65, 66). A pertinent example is the observed clinical improvement in patients where DBS electrode connectivity was in the claustrum, which was also identified as a shared network for lesions causing parkinsonism via LNM (19). Further, the exploration of beneficial brain lesions that alleviate symptoms to pinpoint optimal DBS targets should extend to various neurological or psychiatric disorders. However, it’s important to note that beneficial brain lesions are exceedingly rare. A study by Joutsa et al. demonstrated that varied lesion locations resulting in essential tremor relief overlapped in a common network within the cerebellum and thalamus, known targets for deep brain stimulation in the management of tremors (32).

Furthermore, a recent study by the same group delved into brain lesions associated with addiction improvement, identifying the paracingulate gyrus and the left frontal operculum. The medial frontopolar cortex emerged as the best-matching connectivity profile for addiction remission (39). These discoveries could pave the way to optimal treatment targets for addiction disorders, lending support to ongoing neuromodulation trials (67). However, it is clear that additional research is required to understand how to extrapolate and interpret LNM findings to identify the most suitable therapeutic targets.

When lesion connections coincide within a single brain network, it’s reasonable to infer that the network has a causative role in symptom production. Nevertheless, the regions at the network’s core may not be crucial in symptom development (68). The correlation between symptoms and the central region of the network is gleaned from brain connectivity patterns. Consequently, regional associations gleaned from LNM must be appraised in comparison with functional neuroimaging results.

A prominent question within LNM pertains to the neurobiological mechanism at play when a network is disrupted by lesions. Although lesions induce dysfunction at their locations, the remote impact of lesions on functional loss in interconnected regions—dependent on the type of connection—remains a topic of debate (69). Dysfunctions in excitatory connections could lead to decreased activity, while disruption in inhibitory connections could result in increased activity (26). Furthermore, it remains uncertain whether the labeling of functional connectivity as positive or negative can indicate decreases or increases in activity, respectively.

LNM’s focus has largely been on the spatial aspect of symptoms induced by lesions, often neglecting the temporal component. Investigating the temporal aspect of symptoms is just as important, given that symptoms evolve over time due to recovery processes and dynamic changes post-injury (3).

While the development of LNM has yielded intriguing results, complete elucidation of neurological enigmas via this method alone remains a considerable distance away. More robust studies are necessary, particularly those utilizing prospective data that assess pre-and post-lesion symptoms. Salvalaggio et al. (38) have previously suggested that direct and indirect measures of functional networks may not be as sensitive to behavioral deficits compared to using structural disconnections, possibly due to the coincidence of structural disconnections with structural damage following lesions. However, Cohen et al. contended that the poor outcome of LNM in predicting behavioral deficits in Salvalaggio et al.’s study could be attributed to methodological considerations (25). Ferguson et al. (40) presented contrasting results, affirming the value of LNM in localizing behavioral deficits. These discrepancies underscore the need for future studies to compare and enhance LNM methods. Furthermore, the efficacy of LNM in identifying therapeutic targets should be examined by strategically placing DBS electrodes in the proposed hub of a network (69).

Cotovio et al. (27) probed the effects of utilizing distinct connectomes on LNM outcomes for mania, concluding that the findings remained consistent and reliable, regardless of the specific connectome used for analysis. Future directions could involve refining normative connectome atlases using higher resolution imaging, integrating results from various connectomes obtained during different tasks, and employing age-and sex-matched datasets for each patient. As functional and structural normative data often involve young, healthy individuals, employing normative lifespan data could yield deeper insights into brain function and enable more accurate symptom mapping. While most studies use normative connectomes, the exploration of disease-specific connectomes may provide more precise results.

Moreover, Bonkhoff et al. (70) discovered that women tend to experience more severe strokes than men, and a recent large-scale study highlighted the effect of sex on neuroimaging metrics over time (71). Thus, incorporating sex-specific normative data in LNM could enhance the precision of this approach.

While we have primarily considered functional and structural networks as relatively stable entities in this review, research increasingly shows that these networks are dynamically changing over time. These temporal fluctuations in connectivity patterns are believed to be crucial for flexible cognitive function, and disturbances in these dynamics have been linked to various neurological and psychiatric conditions. Incorporating these dynamic changes into LNM analyses could therefore potentially increase the sensitivity of the method to detect functional abnormalities linked to behavioral deficits. This integration of dynamics would be a novel addition to the methodology and potentially offers a new avenue for understanding brain function and dysfunction.

Another important discussion point to add is the development and application of machine learning techniques in the context of LNM. Recent advancements in machine learning and artificial intelligence (AI) present exciting opportunities for further refining and enhancing the predictive power of LNM. Machine learning models could be trained to predict the likelihood of specific deficits or symptom severity based on the observed lesion distribution and connectivity disruptions. Such models could provide clinicians with additional tools for prognosis and treatment planning, complementing the traditional, more qualitative approaches.

Lastly, the inter-individual variability in brain connectivity and anatomy should not be overlooked. Even among healthy individuals, there can be substantial differences in the structure and functional connectivity of the brain. This variability might influence the impact of a lesion on cognitive and behavioral function and might partly explain the heterogeneity observed in clinical outcomes following brain damage. Future studies could consider incorporating measures of inter-individual variability into the LNM framework to provide more personalized predictions about treatment outcomes. This could be especially relevant when considering DBS targets, as the optimal target might differ slightly between individuals due to this variability.

These proposed additions aim to enrich the current findings and provide a more comprehensive view of the applications and potential advancements in lesion network mapping.

There are several important methodological considerations that should be addressed in LNM.

LNM offers solid findings in localizing a wide range of neuropsychiatric, behavioral, and movement disorders. Furthermore, LNM is anticipated to identify new treatment targets through symptom mapping. However, the veracity of these methodologies must be validated through more comprehensive prospective studies.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

FN and MA contributed to searching and writing and revising the manuscript. All authors contributed to the article and approved the submitted version.

MA is funded by the EU-project euSNN European School of Network Neuroscience (MSCA-ITN-ETN H2020-860563).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.