Functional brain network alterations linked to cancer-related cognitive impairment in breast cancer patients Authors: Alberto Benelli, Julianne Reilly, Jorge Ortega-Márquez, Michael Parsons, Jorg Dietrich, Giulia Sprugnoli, Felipe Fregni, Emiliano Santarnecchi Published: 2026 ABSTRACT AbstractBackgroundLife expectancy for cancer patients has improved due to new therapies, but many survivors experience long-lasting side effects, including cancer-related cognitive impairment (CRCI). CRCI symptoms—such as difficulties with attention, memory, processing speed, fatigue, and concentration—can persist for years post-treatment. While clinical evidence of CRCI exists, its underlying mechanisms and neural correlates remain unclear. This study aimed to identify functional brain networks associated with CRCI in breast cancer (BC) patients using structural and functional neuroimaging data.MethodsWe reviewed studies on CRCI-related brain changes assessed via diffusion tensor imaging (DTI, n = 282), voxel-based morphometry (VBM, n = 177), and functional MRI (fMRI, n = 328). Using the Yeo 7-network resting-state functional atlas, we evaluated the overlap between CRCI-related brain alterations and known functional networks. Overlapping percentages were calculated, and statistical independence tests assessed associations between CRCI and specific networks.ResultsBased on inclusion criteria, six DTI, six VBM, and ten fMRI studies were selected. Chi-square analyses revealed significant associations between CRCI-related changes and several networks: the Dorsal Attention Network (DAN, χ2 = 17.71, P = .0001), Default Mode Network (DMN, χ2 = 21.08, P = .0000002), Frontoparietal Network (FPN, χ2 = 110.63, P = .000002), and Ventral Attention Network (VAN, χ2 = 8.203, P = .01).ConclusionsFindings highlight network-specific alterations linked to CRCI symptoms in breast cancer survivors. These results enhance understanding of CRCI and may guide targeted therapeutic strategies.Keywords: cancer-related cognitive impairment (CRCI), breast cancer survivors (BCS), chemotherapy, functional brain networks, DTI, fMRI, VBM With the growing success of oncological therapies and increasing survivorship, there is a critical need to understand the long-term effects of treatment on cognitive health. Our study addresses this gap by systematically reviewing the structural and functional neuroimaging literature on cancer-related cognitive impairment (CRCI) in breast cancer patients and mapping the associated brain changes onto functional networks. Using a robust selection of diffusion tensor imaging, voxel-based morphometry, and fMRI studies, we found significant associations between CRCI and changes in specific functional brain networks. These findings highlight potential neural substrates underlying CRCI symptoms such as attention deficits, memory impairment, and reduced processing speed and provide a neurobiological framework that may inform future diagnostic and therapeutic strategies. Introduction Breast cancer (BC) is the most frequently diagnosed cancer in women worldwide and poses a significant global health challenge.1 Each year, approximately 2.26 million new cases are recorded, accounting for 11.7% of all new cancer diagnoses.1,2 Recent advancements in oncology therapeutics have significantly improved survival rates, with up to a 91.2% five-year survival rate reported.3,4 However, despite these advancements, the adverse effects of treatments, especially after chemotherapy, continue to significantly impact the quality of life of BC survivors (BCS).5 Notably, cognitive impairments—primarily in learning and memory, processing speed, executive function, and word retrieval—are prominent among the side effects experienced by BCS.6 Several studies indicate that these cognitive impairments affect between 40%-70% of BCS after chemotherapy7,8 and that those treated with chemotherapy are five times more likely to report cognitive impairment compared to those who underwent other treatments, such as radiation or surgery.8 This evidence underscores the critical need to better understand and address this issue. Since the first reports of this phenomenon in the 1990s, substantial efforts have been made to understand the links among BC, chemotherapeutic agents, and cognitive impairment.9 Cognitive changes and impairments related to cancer have been observed in patients with non-central nervous system (non-CNS) cancers even before undergoing treatment. Multiple studies have highlighted cognitive deficits in newly diagnosed cancer patients across various domains, such as verbal memory, language, visual-spatial abilities, executive function, and psychomotor skills.10 Preclinical studies have illuminated potential mechanisms connecting chemotherapy drugs to CRCI. For example, oxidative stress and inflammation have been observed in the brains of rodent models using doxorubicin and cyclophosphamide.1 Other mechanisms, such as reduced neurogenesis and neurotransmitters,11–15 impaired neuronal dendrites,16 disrupted myelin generation,17 and changes in cytokine levels,18 have also been associated with cognitive decline following chemotherapy. Additionally, chemotherapy in postmenopausal breast cancer patients has been shown to interact with and influence hormone levels of luteinizing hormone (LH), follicle-stimulating hormone (FSH), and progesterone, which have also been linked to cognitive deficits.19,20 Despite significant advancements at the mechanistic level, the structural and functional understanding of CRCI remains in infant stages. Neuroimaging has been used to study the functional and structural aspects of cognitive impairment in BCS.21,22 Structural imaging techniques like diffusion-tensor imaging (DTI) and voxel-based morphometry (VBM) can provide a detailed understanding of changes in physical synaptic connections and white matter (WM) fibers that connect gray matter (GM) regions.23 Studies using VBM in BC patients have revealed diffuse cortical and subcortical WM, as well as bilateral neocortical GM volume reductions,24 and reductions in volumes of the right superior frontal gyrus (SFG) GM and left precuneus WM correlated with declines in attention and visual memory index scores.25 In contrast, functional techniques such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) are sensitive to various neurophysiological processes and can detect differences in brain function even without visible structural damage.26 fMRI studies in BC populations undergoing chemotherapy have reported greater levels of activation in the right inferior frontal gyrus27 and increased activation in the left inferior frontal gyrus (IFG), left insula, bilateral thalamus, and right midbrain during cognitive tasks28,29 However, additional insight on the nature and dynamics of cognitive alterations might be offered by additional imaging techniques like resting-state fMRI (rs-fMRI), especially with respect to resting-state functional networks (RSNs).30,31 The importance of RSNs is that their topography closely corresponds to the topography of responses elicited by a wide variety of sensory, motor, and cognitive tasks.30,31 Indeed, rs-fMRI has become a key component of investigations into neurocognitive32–34 and psychiatric behaviour,35 as brain activity in the resting state has been suggested as a way of investigating the underlying pathophysiological characteristics of mental disorders.36 Specifically, several studies have demonstrated the potential clinical application of rs-fMRI in identifying biomarkers of Alzheimer’s disease and as a method of distinguishing between different types of dementia.37 Despite advancements in neuroimaging providing a better understanding of the functional and structural aspects related to CRCI, there remains a lack of granularity in defining the specific brain networks related to the cognitive decline in BCS. This study aims to identify the specific functional networks associated with CRCI in breast cancer patients. We conducted a literature review of both structural and functional brain imaging studies regarding comparisons between patients with CRCI and a control group. Studies reporting altered anatomical and functional features of CRCI (eg, MNI coordinates/labels) by specified neuroimaging techniques (VBM, DTI, and fMRI) were included. The results of this review were used to create a map of brain networks associated with CRCI, hoping to provide an integrative understanding of this condition by highlighting the importance of its functional profile. Methods We included studies reporting CRCI in patients with BC. The comparisons presented across the selected papers showed differences not only between patients undergoing chemotherapy and healthy subjects (HC) but also between patients undergoing chemotherapy and those without any ongoing treatment (BC−). Therefore, the control groups across studies consisted of either (1) healthy control individuals without a history of cancer or chemotherapy or (2) breast cancer patients not receiving chemotherapy at the time of neuroimaging. This dual-control approach allowed for differentiation between the effects of cancer itself and the specific effects of chemotherapy. We primarily considered the literature that reported altered anatomical and functional features of CRCI as seen through the specified neuroimaging techniques, VBM, DTI, and fMRI. Data extraction was performed for a total of six DTI, ten fMRI, and two VBM studies. We did not select papers based on sustained therapy, as this was not our primary focus. Outlined in Tables 1-3 are the included studies’ protocols, chemotherapy regimens, sample size, Montreal Neurological Institute (MNI) coordinates/labels, and results. Table 1.Summary of DTI (FA) neuroimaging findings in CRCI.ReferencesStudy protocolChemotherapy regimenMNI coordinate (x y z)/labelResults Billet et al.38 34 (BC+)16 (BC−)19 (HCs) FEC (6 cycles) 10FEC (3 cycles) + paclitaxel (3 cycles) 15 Parietal part of corona radiata corpus callosumFrontal part of superior longitudinal fasciculusParietal part of superior longitudinal fasciculusPart of forceps major (BC+) < (BC−) Tong et al.39 24 (BC+)20 (BC−) Common standard-dosePCG Bilaterally(BC+) < (BC−) Deprez et al.40 14 (BC+)10 (BC−)15 (HCs) FEC (6 cycles) 3FEC (3 cycles) + paclitaxel (3 cycles) 11 12 −26 34−39 −42 210 −26 36−33 −38 1439 −44 18−45 −27 −814 −32 38 (BC+) < (BC−, HCs) Abraham et al.41 10 (BC+)9 (HCs) 5(BC+): 4 cycles of doxorubicin andcyclophosphamide (AC)5(BC+): 4 cycles of ACfollowed by 4 cycles of a taxane Genu of corpus callosum(BC+) < (HCs) Mzayek et al.42 26 (BC+)23 (BC−)30 (HCs) 3(BC+): AC17(BC+): AC− docetaxel3(BC+): AC− paclitaxel3(BC+): FEC SLFBilateral posterior thalamic radiationBilateral sagittal StriatumBilateral CingulumBilateral Corticospinal tractBilateral Corona radiataSpleniumBody of the corpus callosum. (BC+) .05). Abbreviations: DAN, Dorsal Attention Network; DMN, Default Mode Network; DTI, diffusion tensor imaging; fMRI, functional magnetic resonance imaging; FPN, Frontoparietal Network; LN, Limbic Network; SMN, Sensorimotor Network; VAN, Ventral Attentional Network; VBM, voxel-based morphometry; VN, Visual Network. Network node parcellation Figure 4 illustrates the percentages of each node most affected by CRCI, based on the 100 node parcellation, through an anatomical (Figure 4A) and graphical (Figure 4B) representation. Results show within the DMN a greater involvement of the right and left Posterior Cingulate Cortex (PCC) as well as the right Parietal Lobule (RHIPL1). As far as the FPN, the nodes most affected are left PCC and LHIPS. In contrast, in the DAN there was a greater involvement of both the right Superior Parietal Lobule (LHSPL2) and the right perirhinal cortex (RHPrC). Finally, for the VAN, an involvement of the left Frontal Medial region (LHFrMed) and right Frontal Medial region (RHFrMed) was observed. A (from top to bottom). graphical representation of overlapping percentages with DMN, FPN, DAN, and VAN nodes, according to a color-graded bar: dark color side indicates low overlapping percentages and bright color side high overlapping percentages. B (from top to bottom). plotting of overlapping percentages with DMN, FPN, DAN, and VAN nodes. Abbreviations: DAN, Dorsal Attention Network; DMN, Default Mode Network; FPN, Frontoparietal Network; VAN, Ventral Attentional Network. Discussion Our study aimed to synthesize existing imaging literature to better define the specific neural networks associated with CRCI in BCS, using outcomes from both functional and structural neuroimaging studies. The analyses showed the highest percentage overlap for DMN, FPN, DAN, and VAN nodes, which also demonstrated a statistically significant association between overall neuroimaging techniques outcomes using the Chi-square test for independence. Furthermore, a more granular overview provided by Fisher’s exact test revealed rs-fMRI as the most consistent brain imaging technique in detecting alterations across brain networks associated with CRCI in the analyzed BCS. Functional and structural neural networks associated with CRCI in BC survivors Attentional difficulties are well documented in CRCI, and observing the link with the VAN and DAN, which are two critical sensory-orienting systems in the human brain58 is consistent with these symptoms. The VAN responds to unexpected or unattended stimuli and is involved in stimulus-driven reorientation and sustained attention. It includes the dorsomedial, mid, and ventrolateral prefrontal cortex; anterior insula; and parietal areas, such as the intraparietal sulcus and temporoparietal junction, along with the midbrain, thalamus, basal ganglia, and cerebellum.59 In contrast, the DAN primarily manages visuospatial attention and primarily consists of the intraparietal sulcus (IPS) and frontal eye fields (FEF).60,61 Despite these networks involving distinct anatomical regions, they are interconnected, and both are crucial for flexible attention control.62 A study using rs-fMRI reported functional alterations in the DAN associated with BC and chemotherapy.44 Moreover, changes in the VAN related to CRCI have been demonstrated using quantitative electroencephalography (qEEG).63 Therefore, findings from this study highlight the potential involvement of the VAN and DAN in the impairment of attention flexibility in BC survivors following chemotherapy. Notably, alterations in the Dorsal Attention Network (DAN) were primarily reported by fMRI rather than structural techniques. This likely reflects the greater sensitivity of fMRI to capture functional connectivity and activation changes within attentional networks, which may not always correspond to detectable structural abnormalities. In addition to attentional difficulties, impairments in planning and processing responses that fall under the direct control of the FPN are well reported. Indeed, the FPN is involved in working memory tasks, cognitive control, and episodic memory and comprises parts of the lateral prefrontal cortex and the posterior parietal cortex.64 Taya et al. (2017), while studying compensation mechanisms for performance decline induced by mental fatigue, showed that an increase in connections within the FPN correlated with cognitive decline.65 Previous findings using rs-fMRI in patients after carboplatin-based chemotherapy for CRCI have reported functional changes in the prefrontal cortex, insula, and caudate.66 A study using Voxel-based Morphometry (T1-VBM) conducted on BC survivors treated with chemotherapy showed reductions in the volume of the right middle and superior frontal gyrus (rMFG and rSFG), both areas known to be involved in executive function and spatially oriented processing,25 cognitive functions previously associated with impairments in cancer patients.67 Hence, our findings in this study underscore the potential role of the FPN in characterizing cognitive decline in CRCI in BCS, specifically in the executive function and spatially oriented processing cognitive domains. The DMN, whose functional roles include memory consolidation, self-referential thought, mind-wandering, and autobiographical memory,68,69 was also implicated as associated with CRCI in our analysis. The DMN comprises various regions, including the ventral medial prefrontal cortex, posterior cingulate cortex, inferior parietal lobule, lateral temporal cortex, and dorsal medial prefrontal cortex.68 Previous neuroimaging studies have suggested DMN impairments following BC therapies, such as reduced gray matter volumes in the precuneus, cingulate cortex, lateral parietal cortex, medial frontal gyrus, and hippocampus.49,70 The association observed in this study between DMN and CRCI, reinforces previous evidence of a potential link between DMN disruption and cognitive impairments observed in BC patients,71 thus suggesting that the DMN may play a role in mediating the symptomatology associated with various aspects of working memory. The cognitive symptoms related to CRCI partially overlap those found in aging, so it is not surprising that there is also the involvement of the DMN, which has already been shown to be vulnerable to aging and sensitive to toxicity and disease states (eg, Alzheimer’s disease).72 Biomarkers of DMN connectivity could contribute to the identification and diagnosis of CRCI, as well as potential progression.49 Further longitudinal research is needed to explore this application. Integration of functional and structural neuroimaging in characterizing CRCI VBM showed a broad range of moderate associations across multiple networks, including the LN, FPN, and DMN, suggesting VBM’s efficacy in capturing structural changes, corroborated by previous studies that noted VBM’s utility in identifying local volumetric and density alterations post-chemotherapy-treatment.24,25,74 DTI presented a mixed pattern, notable for its significant association with VAN, LN, and DMN. These findings reflect the strength of structural brain imaging in depicting white matter integrity loss, a finding consistent with the conclusions drawn by previous studies on brain structural disruption in cognition.40,43,75 Functional neuroimaging, characterized by facilitating the drawing of inferences about the connection between specific brain systems and cognitive processes,76 demonstrated the highest consistency across studies and findings using fMRI, with significant associations observed in networks for the DAN, VAN, LN, and FPN. These findings align with previous evidence suggesting fMRI’s superior sensitivity in detecting functional changes within specific brain networks following cognitive impairment after chemotherapy.22,27,77,78 Furthermore, the inclusion of both non-resting or task-related and resting-state within the functional techniques as part of our analysis allowed for the integration of brain networks associated with CRCI under varying circumstances. While task-related fMRI reflects the changes in the brain’s blood-oxygen-level-dependent (BOLD) signal induced in different networks by certain events while performing cognitive tasks, rs-fMRI measures the BOLD in subjects while they are not engaged in any specific cognitive task. This allows for the integrative mapping of local activity in individual brain regions and relevant networks.79 The individual network characterization by rs-fMRI techniques has provided valuable insight into similar disorders, such as Parkinson’s disease dementia (PDD). The sensitive nature of the technique has allowed for researchers to map individual disruptions in different networks such as the DMN, the visual, and the sensorimotor networks, even with subtle cognitive symptoms, although not all studies have consistently detected such alterations.30,80 Thus, by integrating both resting and non-resting state functional brain imaging techniques, this study shows the potential integration of disruption of networks during cognitive activity as well as the networks affected in the resting state. Potential implications and future directions Our findings highlight several potential implications and directions for neuroimaging in characterizing CRCI in BCS. First, the inclusion of both structural and functional imaging techniques revealed networks potentially associated with cognitive impairment in the BC population. This not only enhances our understanding of CRCI but also generates potential biomarkers for early detection and diagnosis, much like clinical populations such as PPD, Alzheimer’s disease, or FTD, where techniques such as rs-fMRI have shown to be a sensitive modality detecting even subtle cognitive symptoms.80–82 However, caution is warranted when considering MRI as a potential screening tool for CRCI. At present, there are limited data linking early MRI changes to long-term cognitive outcomes, and simple clinical screening questions regarding cognitive complaints may be equally effective in practice. The current value of MRI lies in demonstrating the neural bases of CRCI and guiding future research, rather than routine screening. The current value of MRI lies in demonstrating the neural bases of CRCI and guiding future research, rather than routine screening. In particular, the analysis across structural and functional imaging data suggests a hierarchy of functional networks involved in CRCI, with the DMN and FPN representing the top candidates for both disease tracking and potential therapeutic applications. Further neuroimaging studies are warranted to investigate the neurobiological mechanisms underlying interventions in CRCI. For instance, most studies included in Table 1 reported reduced activity in specific brain areas in patients undergoing chemotherapy compared to both the control group of patients without therapy and healthy controls. These findings highlight the potential to intervene with therapeutic methods that can modulate impaired cortical activity, such as non-invasive brain stimulation (NiBS) techniques, which have previously shown positive effects in mitigating symptoms of cognitive decline by modulating functional connectivity of targeted neural networks.76,83,84 In terms of other methods for improving CRCI symptomatology, it is pertinent to refer to the review by Merceur and colleagues (2024), which reports that cognitive rehabilitation programs were particularly effective, showing improvements in memory and attention for most participants. Additional examples include physical activity programs that have a positive impact on quality of life compared to pharmacological interventions.85 Therefore, the application of neuroimaging is essential in advancing our understanding of CRCI and enhancing therapeutic strategies. This comprehensive approach could significantly improve the quality of life for BC survivors by providing targeted and effective interventions based on individual neurobiological profiles. Limitations Our study has some limitations. First, the number of studies included in the review remained relatively small, which may limit the generalizability of the findings; however, it might represent a solid starting point for future investigations on CRCI effects in BC patients. Second, there was a high variation in the pharmacological treatments within the population of the included studies. Given the limited number of studies, subgroup analysis using similar therapeutic schemes could not be conducted, which may represent a potential confounding source depending on the established therapeutic scheme and the neuroimaging findings. Another important factor to consider was that despite considering an integrative approach within the neural imaging techniques, the individual CRCI effects by different techniques, such as task-related and rs-fMRI, made it difficult to distinguish between resting-state and task performance networks. Finally, additional potential biases and confounding variables at the individual study level must be considered. These include variability in patients’ characteristics, chemotherapy regimens (which may differently affect neurocognitive outcomes), timing of post-treatment neuroimaging (which ranged from weeks to years across studies and may reflect different stages of cognitive recovery or compensation), and differences in imaging timepoints, methodology, and data analysis. Importantly, most studies did not systematically assess or control for cognitive or psychological comorbidities, such as depression, anxiety, fatigue, or menopausal status, which are known to influence both cognitive function and neuroimaging findings. Small sample sizes and variability in cognitive assessments across studies further contribute to heterogeneity and may limit comparability. In addition, our review primarily focused on studies reporting positive neuroimaging findings of CRCI, consistent with our aim of mapping the brain regions and networks implicated in this condition. However, we recognize that negative or null findings can be equally informative, particularly when they highlight methodological challenges or sources of variability. Future research that systematically integrates both positive and negative results will be essential for a more comprehensive understanding of CRCI. In addition, standardized reporting of key patient characteristics such as age and menopausal status will be crucial to improve comparability across studies and to clarify their influence on the onset, severity, and reversibility of CRCI. Therefore, while this study offers a thorough integrative characterization of CRCI in BCS, several factors restrict the broader generalization of the findings. Conclusions Our analysis of neuroimaging studies related to BCS revealed that the main cortical networks associated with CRCI are the DMN, FPN, DAN, and VAN. The findings suggest a potential multi-network involvement in this phenomenon, which may explain the varied cognitive symptoms previously reported across different cognitive subdomains86 and patient reports of inability to solve demanding daily tasks.73 Despite the limited number of studies, this research highlights neural networks that may serve as early diagnostic tools or therapeutic targets for CRCI, underscoring the critical need for further investigation to identify biomarkers and interventions that could alleviate cognitive decline and improve the quality of life for BC survivors. Contributor Information Alberto Benelli, Siena Brain Investigation & Neuromodulation Lab (Si-BIN Lab), Unit of Neurology and Clinical Neurophysiology, Department of Medicine, Surgery and Neuroscience, University of Siena, Siena, 53100, Italy; Precision Neuroscience & Neuromodulation Program, Gordon Center for Medical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Julianne Reilly, Precision Neuroscience & Neuromodulation Program, Gordon Center for Medical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Jorge Ortega-Márquez, Neuromodulation Center and Center for Clinical Research Learning, Spaulding Rehabilitation Hospital and Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Michael Parsons, Pappas Center for Neuro-Oncology, Department of Psychiatry, Massachusetts General Hospital, Boston, MA 02114, United States; Division of Neuro-Oncology, Mass General Cancer Center, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Jorg Dietrich, Division of Neuro-Oncology, Mass General Cancer Center, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Giulia Sprugnoli, Precision Neuroscience & Neuromodulation Program, Gordon Center for Medical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Felipe Fregni, Neuromodulation Center and Center for Clinical Research Learning, Spaulding Rehabilitation Hospital and Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Emiliano Santarnecchi, Precision Neuroscience & Neuromodulation Program, Gordon Center for Medical Imaging, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114, United States. Author contributions Alberto Benelli (Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Julianne Reilly (Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Jorge Ortega-Márquez (Conceptualization, Formal analysis, Methodology, Writing—original draft, Writing—review & editing), Michael Parsons (Methodology, Writing—original draft, Writing—review & editing), Jorg Dietrich (Writing—original draft, Writing—review & editing), Giulia Sprugnoli (Writing—original draft, Writing—review & editing), Felipe Fregni (Conceptualization, Methodology, Writing—original draft, Writing—review & editing), and Emiliano Santarnecchi (Conceptualization, Methodology, Writing—original draft, Writing—review & editing) Funding E.S. is supported by the Alzheimer’s Drug Discovery Foundation (ADDF), the Association for Frontotemporal Dementia (AFTD), and the Treat FTD Fund via GA—201902—2017902; F.F. is funded by NIH (5R21NS131635-02). Conflicts of interest None declared. Data availability Data collected and used in this meta-analysis can be requested from the corresponding author. References Associated Data Data Availability Statement Data collected and used in this meta-analysis can be requested from the corresponding author.