Glioma-white matter tract interactions: A diffusion magnetic resonance imaging-based 3-tier classification and its clinical relevance

Abstract

Background

This study proposed a classification system for the interaction between gliomas and white matter tracts, exploring its potential associations with clinical characteristics, tumor pathological subtypes, and patient outcomes.

Methods

Clinical data and diffusion magnetic resonance imageing (dMRI) from 360 glioma patients who underwent craniotomy were analyzed. Using automatic fiber tractography, glioma-tract relationships were categorized into 3 types: displacement, infiltration, and disruption. Double immunohistochemical staining for isocitrate dehydrogenase (IDH) and myelin basic protein was performed on neuronavigation-guided tissue samples to validate the imaging-based classifications. The clinical implications of these classifications on the extent of tumor resection, postoperative motor function, and survival outcomes were evaluated.

Results

Among the patients, 35 (9.7%) were categorized as displacement type, 283 (78.6%) as infiltration type, and 42 (11.7%) as disruption type. Disruption-type tracts were predominantly associated with IDH wild-type gliomas (87.2%), significantly higher than infiltration (28.5%) and displacement types (23.5%) (P < .001). Displacement and infiltration types were more common in IDH-mutant gliomas (P < .001). Displacement-type tracts were significantly associated with higher rates of gross tumor resection compared to infiltration types (P = .015). In corticospinal tract involved cases, displacement-type tumors demonstrated no significant postoperative motor strength changes, whereas infiltration (P < .001) and disruption types (P = .013) were highly associated with postoperative motor deficits. Histological results aligned with dMRI-based classifications.

Conclusions

This dMRI-based classification of glioma-tract interactions is significantly associated with tumor pathology, resection outcomes, functional prognosis, and survival, providing a valuable tool for personalized and precise surgical planning.

Key Points.

• Proposed a diffusion magnetic resonance imaging-based 3-tier classification of glioma-tract interactions.

• Glioma-tract types correlate with pathology, resection extent, and outcomes.

Importance of the Study.

This study introduces a novel diffusion magnetic resonance imaging-based 3-tier classification for glioma-white matter tract interactions: displacement, infiltration, and disruption. This classification was validated through double immunohistochemical staining (isocitrate dehydrogenase [IDH] and myelin basic protein [MBP]), and it reveals significant correlations with tumor pathology, extent of resection, postoperative motor function, and survival. Displacement-type tracts are associated with higher rates of complete resection and preserved motor strength, while disruption-type tracts correlate with IDH wild-type tumors, significant motor deficits, and poorer survival. These findings underscore the classification’s critical role in preoperative planning, facilitating personalized surgical strategies, and optimizing functional outcomes. By integrating advanced imaging techniques with clinical decision-making, this study establishes a clinically relevant framework for precision neurosurgery and deepens the understanding of glioma-tract interactions in neuro-oncology.

Gliomas are highly invasive malignancies that frequently infiltrate along white matter tracts.^1,2 White matter tracts connect different brain regions and serve as the neurobiological substrate for higher cognitive functions, including motor coordination, language, and emotion, by facilitating neural signal transmission.^3–5 When gliomas involve white matter tracts, a subset of patients exhibit significant neurological dysfunction, whereas others remain functionally intact.^6,7 Similarly, postoperative outcomes vary widely; a proportion of patients develop new or worsening deficits after tumor resection, while others maintain their preoperative functional status. Based on clinical observations, these variations in functional outcomes are hypothesized to stem from differences in the type and degree of tumor-tract interactions, though robust evidence has been lacking.⁸

Visualization of white matter tracts is crucial for surgical planning.^9,10 Diffusion magnetic resonance imaging (dMRI) has become the preferred noninvasive method for reconstructing and visualizing brain white matter tracts. Previous studies using the diffusion tensor imaging process and fiber tractography categorized tumor-induced white matter changes into 4 types: displacement, edema, infiltration, and disruption. While this imaging-based classification aligns with clinical intuition, its pathological accuracy has not been validated. Furthermore, the predictive value of this classification for functional outcomes remains uncertain.

This study aims to address these gaps by investigating the classification of glioma-tract interactions and their clinical significance. Using advanced tractography techniques, we systematically classified tumor-tract relationships and validated these classifications through neuronavigation-guided biopsies and double immunohistochemical staining. By integrating preoperative and postoperative functional evaluations with follow-up data, we examined the associations between tumor-tract types, glioma pathology, and patient outcomes. This study provides direct evidence for tumor-tract classification, enhancing our understanding of the interactions between glioma growth and invasion with white matter tracts. It also offers novel insights and a solid foundation for developing personalized therapeutic strategies based on tumor-tract interaction types.

Materials and Methods

Clinical Data Collection

This retrospective study included patients who underwent their primary glioma resection at the Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, between January 2016 and December 2021. The inclusion criteria were as follows: (a) patients aged ≥18 years; (b) histologically confirmed primary gliomas; (c) availability of preoperative dMRI data; (d) patients received no chemotherapy or radiotherapy before they underwent their first operation; and (e) informed consent provided by the patients or their families. Exclusion criteria included: (a) poor-quality dMRI data or other factors preventing fiber tractography analysis; (b) history of brain injury or other neurological diseases. A total of 360 patients met the criteria and were included in this study. Ethical approval was obtained from the institutional review board of Beijing Tiantan Hospital (KY2014-002-02).

Comprehensive clinical data were systematically collected, including preoperative clinical presentations, positive neurological signs, seizure history, and postoperative clinical outcomes, including complications. Tumor characteristics, such as location, preoperative volume, and extent of resection, were derived from imaging data. Pathological information, including tumor subtype, histological grade, immunohistochemical staining, and key molecular markers, was recorded. Pathological diagnoses adhered to the WHO 2021 classification criteria.¹¹

MRI Acquisition

Preoperative MRI was performed on a Prisma 3.0T scanner (Siemens) with standard imaging protocols. Anatomical images included T1-weighted images (TR = 2300 ms, TE = 2.3 ms, flip angle = 8°, FOV = 240 × 240 mm², voxel size = 0.9 × 0.9375 × 0.9375 mm³, and slices = 192) and T2-weighted images (TR = 5000 ms, TE = 105 ms, flip angle = 150°, FOV = 215 × 215 mm², voxel size = 0.36 × 0.36 × 4.4 mm³, and slices = 30). Diffusion magnetic resonance imaging data were acquired using a single-shot echo-planar imaging sequence (TR = 8000 ms, TE = 60 ms, FOV = 282 × 282 mm², voxel size = 2.2 × 2.2 × 2.0 mm³, directions = 30, b-values = 0/1000 s/mm², and slices = 75).

dMRI Processing

Diffusion magnetic resonance imaging preprocessing and fiber tractography were conducted using the QSIPrep platform (https://github.com/MPIB/qsiprep, version 0.20.0),¹² which integrates multiple diffusion image processing tools. Preprocessing steps included denoising, motion correction, eddy current correction, distortion correction, coregistration with T1-weighted images, spatial normalization, resampling, and gradient direction correction, resulting in preprocessed structural and diffusion imaging data. These preprocessed files served as input for the reconstruction process. Fiber tractography employed probabilistic algorithms provided by MRtrix3 software (https://qsiprep.readthedocs.io/en/latest/reconstruction.html) to generate probabilistic fiber tracking maps.¹³

White Matter Tract Segmentation

Individual white matter tracts were segmented using the TractSeg software (https://github.com/jasonbian97/TractSeg),¹⁴ which employs a deep neural network model for tract segmentation. Preprocessed dMRI data from the prior step served as input. Using constrained spherical deconvolution, the fiber orientation distribution function for each voxel was estimated. The maximum of this function (referred to as the “peak”) represents the most likely fiber orientation within the voxel and was used as input to a fully convolutional neural network trained to segment tracts. This process resulted in the identification of 72 distinct white matter tracts in individual space. This method has been applied in multiple studies and has been proven to be accurate and effective.^15–17

Tumor Segmentation and Volume Calculation

For patients with functional area gliomas, we combine intraoperative awakening and direct electrical stimulation to determine the distance between the tumor and the functional area, so that we can remove the tumor to the maximum extent possible while ensuring patient function.¹⁸

Tumor volumes were calculated using a volumetric approach based on MRI scans. Tumor masks, both preoperative and postoperative, were manually delineated using MRIcron software (https://www.nitrc.org/projects/mricron) by 2 experienced neurosurgeons. If the overlap between the 2 masks exceeded 95%, the segmentation from the more senior neurosurgeon was used. For discrepancies greater than 5%, a third neurosurgeon with over 10 years of experience determined the final segmentation. Volumetric analyses of tumor size and extent of resection of tumors were performed on T2-weighted fluid-attenuation inversion recovery images and T2-weighted images (for WHO grade 2 and grade 3 gliomas), and T1-weighted contrast-enhanced images (for WHO grade 4 gliomas).

The extent of resection (EOR) was calculated using the following formula:

EOR = 1 − $\frac{T\mathrm{u}\mathrm{m}\mathrm{o}\mathrm{r}volume\left(postoperative\right)}{T\mathrm{u}\mathrm{m}\mathrm{o}\mathrm{r}volume\left(preoperative\right)}$

Tumor resection was classified into 3 scales: (a) gross total resection (GTR): EOR ≥ 0.9; (b) subtotal resection (STR): 0.8 ≤ EOR < 0.9; and (c) partial resection (PR): EOR < 0.8.

Classification of White Matter Tracts

The classification of white matter tract types for each patient was conducted in 2 steps. First, white matter tract masks obtained from TractSeg and tumor masks were co-registered to individual T1-weighted structural images using the linear registration tool in FMRIB Software Library (FSL). Second, white matter tracts were classified into 3 types based on 2 features: the presence of tracts within the tumor and the continuity of the tracts (Figure 1). The classifications were defined as follows: displacement type: no evident fiber tracts present within the tumor-infiltrated region, and the surrounding fiber tracts are displaced by the tumor but structurally intact. Infiltration type: distinct fiber tracts are observed traversing the tumor-infiltrated region. Disruption type: no evident fiber tracts are present within the tumor-infiltrated region, and the surrounding fiber tracts exhibit significant disruption in both morphology and continuity. The predominant pattern for all cases included in this study was independently classified by 2 neurosurgeons. When consensus was reached, the agreed-upon classification was assigned to the case. In cases of disagreement, a senior neurosurgeon with over 10 years of experience provided the final classification decision.

Figure 1.

Overview of the study design. (A) dMRI preprocessing and probabilistic fiber tracking. (B) Tumor segmentation and 3D reconstruction. (C) Spatial mapping of tumor lesions and fiber tracts. (D) Histopathological sectioning and immunohistochemical staining. (E) Statistical analysis.

Perioperative Functional Assessment

The motor functional status of patients was evaluated using the Medical Research Council (MRC) muscle strength grading system, ranging from 0 to 5, with higher grades indicating better motor function. Medical Research Council scores were recorded 1 day before surgery and 1 week postoperatively to compare preoperative and postoperative motor function.¹⁹

Tissue Sampling and Double Immunohistochemical Staining

During tumor resection in 3 patients, neuronavigation (Brainlab, kick2) was used to precisely define the location of resection (registration error <1 mm). Tissue samples from tumor regions near the margin were paraffin-embedded, sectioned, and subjected to double immunohistochemical staining. Tumor-tract interaction types identified on dMRI were compared with histological findings from 18 sampled sites.

Two experienced pathologists independently reviewed hematoxylin-eosin and immunohistochemistry staining results to ensure diagnostic consistency. Double immunohistochemical staining involved the simultaneous use of 2 antibodies on a single slide: isocitrate dehydrogenase 1 (IDH1) to label tumor cells and myelin basic protein (MBP) to identify the myelinated structures of white matter tracts. This approach illustrated the spatial distribution of tumor cells and white matter tracts. The antibodies used were mouse-derived anti-IDH1 R132H (1:100, Dianova, H09-w/o BSA) and rabbit-derived recombinant anti-MBP (1:5000, Abcam, ab218011). IDH-positive cells appeared brown and MBP-positive regions appeared red. Additionally, routine IHC staining for P53 and ATRX was performed.

Statistical Analysis

Categorical data, such as tract types, extent of resection, pathological grades, and subtypes, were analyzed using the chi-square test or Fisher’s exact test (for expected frequencies <5). For continuous variables, including patient age and preoperative tumor volume, comparisons were made using the Wilcoxon rank sum test for two-group analyses and the Kruskal–Wallis test for multi-group comparisons, followed by Dunn’s post hoc test. Paired MRC scores before and after surgery were analyzed with the Wilcoxon signed rank test. Survival curves were analyzed by log-rank tests with Bonferroni correction. Univariate and multivariate logistic regression analyses were employed to evaluate associations between tract types, demographic factors (eg, sex, age, and preoperative tumor volume), and molecular markers of gliomas (eg, IDH, P53, and ATRX mutations).

Results

Characteristics of Patients

The baseline characteristics of the 360 patients included in this study are shown in Table 1. Among the cohort, 35 patients (9.7%) exhibited displacement-type tracts, 283 patients (78.6%) had infiltration-type tracts, and 42 patients (11.7%) were classified as having disruption-type tracts. For all enrolled patients, the inter-rater reliability between the 2 neurosurgeons yielded a κ value of 0.94. The median age of patients with disruption-type tracts was 52 years, significantly older than those with displacement-type (median 44 years, P = .011) and infiltration-type tracts (median 42 years, P = .006). No significant age difference was observed between displacement and infiltration types (P = .607). The preoperative tumor volume in the disruption group (median 71.4 cm³) was significantly larger than in the displacement group (median 42.7 cm³, P = .021) and the infiltration group (median 41.4 cm³, P = .001). The distribution of tumor-tract types did not differ significantly between sexes (P = .838). In our cohort, 131 patients (36.4%) had tumor volumes exceeding 60 cm³, 13% of which involved the insular and basal ganglia regions, contributing to residual tumor in some cases.

Table 1.

Baseline Characteristics of the Study Cohort

Characteristics	Total	Displacement	Infiltration	Disruption	P value
	(n = 360)	(n = 35)	(n = 283)	(n = 42)
Sex					.838b
Female	158 (43.9%)	16	122	20
Male	202 (56.1%)	19	161	22
Age (years)					.005 a
Median [IQR]	44 [34–53]	44 [33.5–51]	42 [34–52]	52 [44–57]
Side of tumor					.073b
Left	184 (51.1%)	21	140	23
Right	163 (45.3%)	12	136	15
Both side	13 (3.6%)	2	7	4
Tumor location					.619b
Frontal	190 (52.8%)	16	152	22
Parietal	59 (16.4%)	5	45	9
Temporal	64 (17.8%)	10	49	5
Insula	47 (13.0%)	4	37	6
Tumor volume (cm3)					.002 a
Median [IQR]	43.5 [23.5–77.2]	42.7 [19.2–73.0]	41.4 [22.9–75.8]	71.4 [40.8–117.8]
Preoperative symptomsa					/
Seizures	94 (38.1%)	5	81	8
Aphasia	25 (10.1%)	4	17	4
Limb weakness	43 (17.4%)	4	31	8
Limb numbness	26 (10.5%)	1	23	2
Headache	36 (14.6%)	6	26	4
Other symptoms	40 (16.2%)	3	30	7
No symptoms	33 (13.4%)	11	28	2
Tumor grade					<.001 b
Grade 2	173 (48.1%)	21	150	2
Grade 3	62 (17.2%)	4	53	5
Grade 4	125 (34.7%)	10	80	35
Tumor histology					<.001 b
Oligodendroglioma	104 (28.9%)	5	95	4
Astrocytoma	135 (37.5%)	22	110	3
Glioblastoma	121 (33.6%)	8	78	35
Molecular markers
IDH1c					<.001 b
Wild type	118 (34.7%)	8	76	34
Mutant type	222 (65.3%)	26	191	5
ATRXc					.028 b
Wild type	75 (27.3%)	10	62	3
Mutant type	200 (72.7%)	12	164	24
P53c					.302b
Wild type	66 (23.2%)	4	57	5
Mutant type	219 (76.8%)	21	170	28
EORc					.028 b
Total	102 (34.2%)	16	76	10
Subtotal	85 (28.5%)	5	75	5
Partial	111 (37.3%)	6	91	14

Abbreviations: EOR, extent of resection; IQR, interquartile range.

Note: ^aP-values were derived from the Kruskal–Wallis test. ^bP-values were calculated using the chi-square test or Fisher's exact test, as appropriate. ^cSome variables have missing data for certain patients. Bold indicates statistical significance.

Tumor-Tract Types and Tumor Pathology

Tumor-tract interaction types varied across glioma grades and histological subtypes. In general, infiltration-type tracts were the most common across all tumor types. Displacement and infiltration types were more frequently observed in lower-grade gliomas (WHO grades 2 and grade 3) than in disruption-type tracts (P < .001). Conversely, disruption-type tracts were predominantly associated with WHO grade 4 gliomas (Table 1).

Tumor histological subtypes also displayed distinct distributions of tract types. Among displacement-type tracts, astrocytomas accounted for the largest proportion (62.9%, 22/35). In the infiltration-type group, the proportions of tumors across histological subtypes were relatively similar. Disruption-type tracts were dominated by glioblastomas, comprising 83.3% (35/42) of this group. Oligodendrogliomas were predominantly associated with infiltration-type tracts (91.3%, 95/104), significantly higher than the proportions in astrocytomas (81.5%, 110/135, P = .030) and glioblastomas (64.5%, 78/121, P < .001). Astrocytomas showed a higher proportion of infiltration-type tracts than glioblastomas (P = .002). Additionally, glioblastomas exhibited a significantly higher proportion of disruption-type tracts (28.9%, 35/121) compared to astrocytomas (2.2%, 3/135) and oligodendrogliomas (3.8%, 4/104) (P < .001).

A strong association was observed between tract types and key molecular phenotypes of gliomas. Disruption-type tracts were predominantly IDH wild type (87.2%, 34/39), significantly higher than infiltration-type (28.5%, 76/267) and displacement-type tracts (23.5%, 8/34) (P < .001). IDH-mutant gliomas were more frequently associated with displacement and infiltration-type tracts (P < .001). Among displacement-type tracts, ATRX mutations were found in 54.5% of cases, while the proportions were higher in infiltration-type (72.6%, P = .126) and disruption-type tracts (88.9%, P = .017). No significant differences were observed between tract types and P53 phenotypes (P = .302). Univariate and multivariate regression analyses indicated that disruption-type tracts were independently associated with IDH wild-type gliomas (OR = 39.220, 95% CI = 9.292–165.541, P < .001; Supplementary Table 1). Among the cases included in this study with available dMRI imaging and fiber tractography, the IDH-mutant to IDH wild-type ratio was 1.88 (222/118). This ratio is largely comparable to the mutant-to-wild type ratio of 1.33 (2266/1708) observed in primary glioma patients who underwent surgery at the neuro-oncological surgery wards of our institution during the same period.

Double Immunohistochemical Staining of Tumor Pathology

Double immunohistochemical staining of tumor pathology revealed 3 distinct patterns of marker expression, corresponding to the 3 tumor-tract interaction types identified on dMRI (Figure 2): (a) tumor cells (−), tracts (+); (b) tumor cells (+), tracts (+); and (c) tumor cells (+), tracts (−). These patterns were consistent with the displacement-type (Type I), infiltration-type (Type II), and disruption-type (Type III) classifications, respectively. Analysis of whole pathological slides from 3 tumors (1 displacement type, 1 infiltration type, and 1 disruption type) showed complete concordance between immunohistochemical findings and dMRI-based classifications (Supplementary Figure 1).

Figure 2.

Schematic representation of white matter fiber tract classification in gliomas. Type Ⅰ, displacement type: fiber tracts are displaced by the tumor but remain structurally intact. Immunohistochemistry: IDH (−), MBP (+). Type Ⅱ, infiltration type: distinct fiber tracts are observed traversing the tumor-infiltrated region. Immunohistochemistry: IDH (+), MBP (+). Type Ⅲ, disruption type: fiber tracts exhibit significant disruption and loss of continuity. Immunohistochemistry: IDH (+), MBP (−).

Tumor-Tract Types and Extent of Resection

Significant differences in the EOR were observed among the 3 tract types (P = .028). Gross total resection (EOR ≥ 0.9) was achieved in 59.3% of displacement-type cases, compared to 31.4% of infiltration-type and 34.5% of disruption-type cases. Displacement-type tracts were significantly more likely to undergo GTR compared to infiltration-type (P = .015).

Perioperative Symptoms and Tract Morphology

The most common preoperative symptoms among patients were seizures, motor deficits, and headaches. Significant differences in preoperative symptoms were observed among the three tumor-tract types (P < .001). Patients with infiltration-type and disruption-type tracts were more likely to present with clinical symptoms compared to those with displacement-type tracts (P < .001 for both). However, no significant differences were observed among the 3 tract types regarding the prevalence of specific symptoms, such as seizures, aphasia, or motor deficits.

For patients with tumors involving the corticospinal tract (CST), preoperative and postoperative diffusion data were analyzed using fiber tractography. The changes in CSTs across the three tract types are illustrated (Figure 3A). For displacement type, the affected CSTs exhibited partial morphological recovery postoperatively. For infiltration type, postoperative tracts showed reduced volume, with partial disruption of tract integrity in some patients. For disruption type, the affected CSTs were significantly reduced preoperatively, with disrupted integrity, and demonstrated minimal changes postoperatively compared to the preoperative state.

Figure 3.

Functional and survival outcomes associated with tumor-tract types. (A) Preoperative and postoperative status of the corticospinal tract in the 3 tumor-tract types. (B and C) Comparison of preoperative (B) and postoperative (C) muscle strength among the 3 tumor-tract types. (D) Paired analysis of pre- and postoperative motor strength changes within each tumor-tract type. (E) Kaplan–Meier curves for overall survival stratified by tumor-tract type. Note: Statistical significance is indicated as follows: *P < .05; ***P < .001.

Tumor-Tract Types and Clinical Outcome

In this study, a total of 147 patients underwent intraoperative awake surgery, including 116 patients with tumors involving the motor area. Preoperative and postoperative motor function were further analyzed in patients with CST involvement. Preoperatively, patients with disruption-type tracts exhibited significantly lower muscle strength scores compared to those with infiltration-type (P = .021) and displacement-type tracts (P = .020), indicating more severe motor deficits (Figure 3B). No significant differences were observed between infiltration-type and displacement-type tracts (P = .909). Postoperatively, muscle strength scores in infiltration-type and disruption-type tracts were significantly lower than in displacement-type tracts (Figure 3C, P = .032 and P < .001, respectively), while no significant differences were observed between disruption-type and infiltration-type tracts (P = .059).

Paired comparisons of preoperative and postoperative muscle strength scores revealed that patients with displacement-type tracts showed no significant changes (P = .152). In contrast, patients with infiltration-type (P < .001) and disruption-type tracts (P = .013) exhibited significant declines in muscle strength postoperatively (Figure 3D). Among the 198 patients with available survival data, groups were classified based on the type of fiber tract interaction (Figure 3E). Patients in the disruption group exhibited significantly shorter survival times compared to those in the infiltration group (P < .001). However, no statistically significant differences were observed between the disruption and displacement groups (P = .235), or between the infiltration and displacement groups (P = .851).

Discussion

This study introduces a novel classification of glioma-tract interactive relationships based on dMRI and advanced fiber tractography algorithms, encompassing displacement, infiltration, and disruption types. The accuracy of this imaging-based classification was validated at the histopathological level, and the study further explored the associations between these classifications and clinical features, tumor pathology, molecular markers, extent of resection, postoperative function, and overall survival. Notably, the tumor-CST interaction type was found to have a significant impact on postoperative motor function.

The classic classification of tumor-tract relationships was first proposed in 2002,²⁰ categorizing brain tumor impacts on white matter tracts into displacement, edema, infiltration, and disruption. Subsequently, researchers conducted a quantitative analysis based on DTI-related parameters.^21–23 However, the accuracy and scientific validity of these classifications had not been confirmed. In this study, we provided the first pathological-level characterization of tumor and white matter tract distributions using double immunohistochemical staining. We observed distinct tumor-tract interaction patterns corresponding to the displacement, infiltration, and disruption types, supporting the validity of this classification and its potential clinical applications.

Our findings demonstrated significant differences in tract type distributions among glioma grades and histological subtypes. Displacement and infiltration types were more frequently associated with lower-grade gliomas, whereas disruption types were predominant in WHO grade 4 gliomas. This aligns with the biological characteristics of gliomas: lower-grade gliomas tend to grow slowly with localized infiltration, whereas higher-grade gliomas grow rapidly, destroying normal brain tissue and posing greater risks to life. Similar observations have been reported in previous studies, where high-grade tumors were often linked to infiltration or disruption types, while low-grade tumors were associated with displacement types.²⁴

Furthermore, among low-grade gliomas, infiltration was the most common tract type, followed by displacement and disruption. Interestingly, displacement types were more prevalent in astrocytomas compared to oligodendrogliomas, suggesting a potential molecular subtype of astrocytomas that predominantly displace surrounding tracts, with clearer tumor boundaries and minimal disruption to adjacent tissue and function. For displacement-type gliomas, surgical strategies should carefully consider access routes, prioritizing resection through the interspaces of fiber tracts displaced by the tumor. This strategy facilitates complete tumor resection while minimizing the risk of damage to critical tracts and preserving functionality.

The tumor-tract types were also correlated with IDH status. Previous studies have shown that IDH-mutant gliomas are associated with lower malignancy, higher rates of GTR, and increased sensitivity to chemotherapy and radiotherapy.²⁵ Our findings are consistent with these observations, offering a novel perspective by linking the IDH mutation with tumor-tract interaction types. Displacement-type tumors were more likely to achieve extensive resection and had a higher proportion of IDH-mutant gliomas compared to infiltration and disruption types. Conversely, IDH wild-type tumors, which exhibit higher proliferation and destructive capacity, were more prevalent in disruption types. These results emphasize the association between imaging-based tumor-tract relationships and the molecular biology of gliomas. Furthermore, IDH status was reflected in dMRI metrics that indicate tumor invasiveness.²⁶ This suggests that surgical and adjuvant treatment strategies should account for the differences in tumor biology and tract relationships when managing IDH-mutant and wild-type gliomas.^25,27 For immunohistochemical staining, the sensitivity and specificity for detecting IDH1 were 98.2% and 84.2% respectively²⁸; a sensitivity of 83% and a specificity of 100% for MBP detection.²⁹

The primary goal of glioma surgery is to achieve maximal safe resection while preserving critical functions.^5,30 Previous studies have established that greater tumor resection during initial surgery correlates with longer overall survival.^31–33 Our findings showed significantly lower rates of GTR for infiltration-type gliomas compared to displacement types. This may be due to 2 factors. First, glioma infiltration into white matter tracts results in a mixed composition of tumor and normal tissue, making it more challenging for neurosurgeons to identify precise tumor margins. This increases the likelihood of either incomplete resection or unintended damage to adjacent normal brain tissue. Second, when tumors infiltrate but do not destroy critical tracts, surgeons may adopt a more conservative resection strategy to avoid functional impairments. Supporting this observation, intraoperative electrical stimulation during awake surgeries has shown that even when tumors appear to invade critical tracts on imaging, these tracts may retain functional integrity, with positive stimulation points identified within tumor-infiltrated regions. This indicates that the tumor may invade tracts without fully disrupting their structure and function.

When tumors invade the CST, different tumor-tract interaction types have significant implications for postoperative motor function. Our findings indicate that disruption-type CST is more likely to result in preoperative motor strength deficits compared to the other 2 types. This can be attributed to 2 factors. First, the tumor directly damages the CST, leading to evident functional impairments. Second, disruption-type interactions are frequently associated with glioblastomas, which grow rapidly, causing repeated destruction of surrounding tissues before compensatory mechanisms can develop. Previous studies have also demonstrated that direct tumor invasion of white matter tracts weakens neural conduction, contributing to motor strength deficits.³⁴ Patients with disruption-type CST often present with preoperative motor impairments, making the value of intraoperative functional preservation variable and patient-specific; functional preservation should not always be prioritized over achieving adequate resection.

Compared to displacement-type CST, patients with infiltration-type CST exhibited more pronounced postoperative motor strength declines. This may stem from cases where infiltration-type CST appears functionally intact preoperatively, leading neurosurgeons to assume that GTR would not impair function. However, resection along imaging-defined tumor margins may inadvertently disrupt the integrity of critical tracts, resulting in new postoperative deficits. Our group has previously demonstrated the existence of functional regions within tumors using resting-state functional MRI and individualized brain network analyses, enabling visualization of these regions.³⁵ The current study provides a subcortical-level explanation and validation of these findings, suggesting that intratumoral functional regions may be attributable to tract infiltration without structural disruption. Prior research also highlights the importance of preserving critical white matter tracts during surgery to minimize postoperative functional impairments, although invasive tumors increase the challenge of such preservation.^36,37

Paired analyses of pre- and postoperative motor function further revealed that disruption-type and infiltration-type CSTs are associated with poorer functional outcomes compared to displacement-type CSTs. For patients with infiltration-type CST involvement, complete resection may not be both safe and feasible. We strongly recommend comprehensive preoperative planning, incorporating advanced techniques such as direct electrical stimulation, electrophysiological monitoring, and neuronavigation during surgery. These methods help localize and preserve critical motor functions, reducing postoperative deficits. For unresectable tumor components, more aggressive and targeted radiotherapy may be necessary. Conversely, in gliomas with fully disrupted CSTs, resection does not exacerbate postoperative deficits, suggesting that a more aggressive resection strategy can be safely implemented.

This study has several limitations. First, it is a qualitative study that classifies white matter tracts into 3 types based on predefined criteria, primarily focusing on associations between tract types and clinical outcomes at the population level. Future research should incorporate quantitative analyses of white matter tracts using metrics such as fractional anisotropy (FA), mean diffusivity (MD), and fiber density index (FDi)³⁸ to more precisely characterize tumor-tract interactions. Second, while our study confirms that gliomas can infiltrate tracts, it remains unclear how to determine whether infiltrated tracts retain functionality using imaging techniques. Future work should focus on functional visualization technologies for assessing the functional integrity of affected tracts. Third, this study is based on single-center data, which may limit the generalizability of our findings. Although the algorithms and methods used are open-source and widely applicable, future studies should include multicenter datasets to validate our conclusions. Prospective data collection is also necessary to confirm our findings and ensure their applicability across diverse clinical settings and imaging modalities.

Conclusion

This study introduces a novel 3-tier classification of tumor-tract relationships—displacement, infiltration, and disruption types—validated by pathological evidence. Significant associations were identified between these tract types and glioma pathology, molecular characteristics, extent of resection, and functional outcomes. The findings deepen our understanding of glioma growth, invasion, and their interactions with white matter tracts, while providing critical evidence and new perspectives for developing personalized surgical strategies tailored to tumor-tract interaction types.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic.oup.com/neuro-oncology).

Funding

This research was supported by the Beijing Natural Science Foundation (JQ23040), Beijing Natural Science Foundation (L241027), Science and Technology Project of Hebei Province (22377718D), and the National Natural Science Foundation of China (82203170).

Conflict of interest statement. None declared.

Authorship statement

J.H. and S.F. collected and organized all the patients' dMRI data, performed the preprocessing, and carried out the tract segmentation; J.H. and H.B. contributed to data analysis and writing the manuscript. X.L., Z.Y., Z.W., and R.Z. participated in the immunohistochemical staining work. R.W. and T.P. collected clinical information of the patients. C.L. and Z.C. provided valuable suggestions for revising the manuscript. T.J. and Y.W. contributed to conceptual design and project integrity. All authors read and approved the final manuscript.

Data availability

The data that support the findings of this study are available upon request from the corresponding authors. The data are not publicly available due to privacy or ethical restrictions.