Quantitative white matter alterations of the anterior visual pathway predict injury following radiotherapy for nasopharyngeal carcinoma: a high-angular-resolution diffusion imaging study

Abstract

Objective

This study aimed to use high-angular-resolution diffusion imaging (HARDI) to evaluate microstructural alterations and temporal evolution patterns in the anterior visual pathway (AVP) white matter of patients with nasopharyngeal carcinoma (NPC) following radiotherapy (RT).

Methods

150 patients with nasopharyngeal carcinoma (NPC) were retrospectively enrolled and stratified into four groups according to the post-radiotherapy (post-RT) interval: G0 (pre-RT, n = 39), G1 (< 1 month post-RT, n = 30), G2 (< 6 months post-RT, n = 33), and G3 (> 6 months post-RT, n = 48). Based on high-angular-resolution diffusion imaging (HARDI), generalized q-sampling imaging (GQI) reconstruction was performed to derive GQI parameters for the anterior visual pathway, including generalized fractional anisotropy (GFA) and normalized quantitative anisotropy (NQA). Corresponding diffusion tensor imaging (DTI) parameters, including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), were also obtained. One-way analysis of variance (ANOVA) was used to compare all parameters across the four groups, followed by post hoc pairwise comparisons. Cohen’s d was calculated to assess the effect size of each parameter relative to pre-RT values.

Results

Compared with the G0 group, fractional anisotropy (FA) and normalized quantitative anisotropy (NQA) values were significantly decreased in G1 (P < 0.05). Conversely, mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) values were significantly increased bilaterally in G1 (P < 0.05). No significant differences were observed in other post-RT groups. No significant differences in generalized fractional anisotropy (GFA) values were observed between the pre-RT and any post-RT groups. Compared with the G0 group, effect-size analysis demonstrated that the largest effect sizes were present in G1. Effect sizes were generally smaller in G2. In G3, effect sizes were predominantly small to moderate.

Conclusions

DTI and GQI together exhibit the potential to serve as a biomarker of detecting microstructural alterations in the white matter of the anterior visual pathway before visible changes appear on conventional MR images, enabling a better understanding of radiation-induced injury.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12880-026-02565-z.

Introduction

Nasopharyngeal carcinoma (NPC) is a common head and neck malignancy, with a high incidence in North Africa, Southeast Asia, and southern China [1–3]. Radiotherapy (RT) is the primary treatment for NPC [4]. Owing to the specific anatomical location and complex structure of the nasopharynx, the radiotherapy target volume is frequently adjacent to critical structures including the optic nerves, parotid glands, and brain parenchyma. Consequently, these tissues along the RT pathway and near tumors are inevitably irradiated [5]. Previous evidence indicates that the visual pathway is more vulnerable to radiation damage compared with other white matter tracts, with injury occurring even at relatively low radiation doses. Radiation-induced optic neuropathy (RION) is a devastating complication following external beam radiotherapy (EBRT). Clinically, RION is characterized by rapid, progressive visual loss when symptoms occur and is categorized into acute anterior injury and late delayed injury. Late delayed injury occurs after a prolonged latent interval and often presents normal in the early phase. This preclinical course may therefore provide an opportunity for therapeutic intervention, as visual symptoms may not occur until months or even years after exposure of the anterior visual pathways to ionizing radiation. However, no effective treatment has been proven to stabilize or reverse vision loss; therefore, early detection of visual pathway alterations is of critical importance. With advances in radiotherapy techniques and segmentation, acute optic pathway injury has become relatively uncommon. Currently, RION is mostly of the delayed type, which has been reported to range from 3 months to 14 years after radiotherapy, with no obvious clinical symptoms in the early stage. It is predominantly characterized by sudden, painless, and irreversible visual loss, often accompanied by visual field defects and color vision deficiency [6–9].When visual decline occurs, optic atrophy is often already present, with no effective treatment available, resulting in a poor prognosis. Although conventional MRI and visual evoked potentials (VEP) reflect structural and functional changes, respectively, the complex anatomy of the visual pathway makes it challenging to quantitatively assess the microstructural integrity of the visual pathway. This limitation hinders a comprehensive understanding of the radiation-induced optic pathway damage following radiotherapy for NPC.

Recent studies have demonstrated that DTI can detect early microstructural damage to the visual pathway even when there are no visible changes on conventional MR images [10–12]. In other studies involving the optic pathway, it has been reported that FA values correlate significantly with optic nerve atrophy, retinal structural damage, and the severity of visual field defects [13, 14], suggesting that DTI can detect microstructural alterations in the visual pathway, including the optic chiasm, and is closely associated with both structural and functional integrity. In the field of radiotherapy, DTI has also been shown to identify early microstructural WM changes. Yazol et al. found that as early as three months after Gamma Knife radiosurgery (GKRS), patients exhibited decreased FA and increased apparent diffusion coefficient (ADC) and RD values in the bilateral optic nerves and anterior optic radiations, even with normal visual symptoms, demonstrating that DTI can detect radiation-induced early microstructural damage before clinical symptoms emerge [15]. Collectively, these findings suggest that DTI can detect microstructural WM differences in the visual pathway following radiotherapy for NPC, thereby facilitating early clinical intervention for visual pathway impairment.

Most studies on the visual pathway following radiotherapy for NPC have focused on the optic radiation. In contrast, research on the anterior visual pathway remains relatively insufficient, mainly due to its small size and the presence of crossing fibers, which cause challenges to reconstruct the entire visual pathway. Previous studies have consistently shown that the anterior visual pathway is particularly sensitive to high-dose ionizing radiation, due to its innate, precarious blood supply [16]. Therefore, investigating the microstructural changes of the anterior visual pathway in patients with NPC is of particular importance.The Q-ball imaging technique introduced by Tuch et al. has demonstrated that Q-ball imaging can capture multi-directional fiber orientation distributions, making it capable of tracking anatomical fiber crossings in the optic chiasm, whereas conventional DTI fails by its low angular resolution [17, 18]. Liang et al. quantitatively assessed the visual pathway using DSI parameters, namely QA and GFA based on GQI. They found that decrease of QA and GFA values were associated with visual field defects and optic chiasm compression, suggesting that GQI-derived parameters may serve as quantitative biomarkers for assessing microstructural damage in the visual pathway and enabling early evaluation [19].

In this study, we combined DTI and GQI to study changes in the WM microstructure of the anterior visual pathways in patients with NPC before and after radiotherapy, with the aim of exploring the pathophysiological changes and facilitating early clinical identification.

Materials and methods

Subjects

The study was approved by the Human Ethics Committee (Institutional Review Board, IRB) of Hainan Hospital Affiliated to Hainan Medical University and Hainan Provincial People’s Hospital. Inform consent was obtained from all subjects. Magnetic resonance imaging examinations were performed according to the time before and after completion of RT: G0 (pre-RT, n = 39), G1(< 1 month post-RT, n = 30), G2 (< 6 months post-RT, n = 33), and G3 (> 6 months post-RT, n = 48) [20, 21].

Inclusion criteria were as follows: (1) Pathologically confirmed newly diagnosed nasopharyngeal carcinoma without intracranial extension; (2) No history of other malignancies; (3) Absence of neurological or ophthalmic disorders, including glaucoma, optic neuritis, ischemic optic neuropathy and congenital optic nerve anomalies; (4) No history of hypertension or diabetes mellitus; (5) Normal visual pathway on conventional brain MRI (T1WI, T2WI, FLAIR).

Exclusion criteria were as follows: (1) insufficient clinical or imaging data; (2) receipt of radiotherapy other than intensity-modulated radiotherapy (IMRT); and (3) severe brain disorders or psychiatric illnesses, such as significant head trauma, stroke, or brain tumors.

Clinical variables included age, sex, T stage, N stage, and M stage.

MRI data acquisition

All MRI examinations were performed using a 3.0-T MR scanner (Skyra, Siemens Healthcare, Erlangen, Germany) equipped with a 32-channel head and neck coil. All participants underwent the same imaging protocol, which included a high-resolution three-dimensional T1-weighted (3D-T1) structural sequence and a high-angular-resolution diffusion imaging (HARDI) sequence. The acquisition parameters were as follows: (1)High-resolution 3D-T1 structural sequence: repetition time (TR) = 2530 ms, echo time (TE) = 2.98 ms, flip angle (FA) = 7°, field of view (FOV) = 256 mm × 256 mm, matrix = 240 × 240, slice thickness = 1.0 mm with no interslice gap, and 192 slices. (2) HARDI sequence (spin-echo echo-planar imaging): b-values = 2000 s/mm² and 0 s/mm², 64 diffusion-encoding directions, TR = 112,000 ms, TE = 52 ms, FA = 90°, FOV = 224 mm × 224 mm, and slice thickness = 2.5 mm with no interslice gap. Image quality control was performed by radiologists with more than five years of experience in diagnostic imaging.

Parameter analysis

Diffusion source data were first converted to NIFTI format and then processed and analyzed with DSI Studio (http://dsi-studio.labsolver.org). Following quality control, diffusion features were computed from the normalized and preprocessed diffusion-weighted images.

Fiber tracking of the optic tracts was conducted using DSI-Studio (http://dsistudio.labsolver.org/). Diffusion data were reconstructed with GQI at a diffusion sampling length ratio of 1.25 [22], employing a quantitative anisotropy-based generalized deterministic tracking algorithm [23]. Regions of interest (ROIs) were mutually placed along the course of the anterior visual pathway on generalized fractional anisotropy map. According to the anatomical atlas of the optic nerve (ON) and the fiber tracking strategy described by Yoshino et al., ROIs were placed on the intracanalicular segment of the optic nerve and the lateral geniculate body to reconstruct fiber tracts encompassing the optic nerve, optic chiasm, and optic tract (Figs. 1 and 2). According to previous studies, the tracking parameters were set as follows: quantitative anisotropy threshold = 0.20, angle threshold = 70°, step size = 1.2 mm, smoothing = 0.80, minimum fiber length = 30 mm, and maximum fiber length = 300 mm [19, 24]. Based on the spatial coordinates of the reconstructed fiber tracts, the normalized quantitative anisotropy (NQA), GFA, FA, MD, AD, and RD values were sampled for different fibers of the anterior visual pathway. The mean parameters were then calculated for fiber tracts.

Fig. 1.

Regions of interest (ROIs) for the anterior visual pathway (AVP), including the optic chiasm, were placed as follows: (A) the intracranial segment of the optic nerve; (B) in the lateral geniculate body

Fig. 2.

Fiber tract reconstruction of the anterior visual pathway (AVP), including the optic chiasm, is shown in (A) axial, (B) sagittal, and (C) coronal planes

Statistical analysis

Statistical analyses were performed using R software (version 4.4.2). Continuous variables were expressed as mean ± standard deviation (SD), while categorical variables (e.g., sex, disease stage) were presented as frequencies (percentages). Levene’s test was used to confirm that all data analyzed by ANOVA met the assumption of equal variances (all P > 0.05), which satisfies the assumptions for ANOVA. One-way analysis of variance (ANOVA) was used to compare diffusion metrics across the four time points. For parameters with statistically significant differences, post-hoc pairwise comparisons (G1, G2, and G3 vs. G0) were conducted using independent-samples t-tests, with multiple comparisons corrected using the Benjamini–Hochberg false discovery rate (FDR) procedure. Effect sizes were quantified using Cohen’s d, Effect sizes were interpreted as small (d = 0.2), medium (d = 0.5), and large (d = 0.8) according to Cohen’s (1988) conventions [25]. A two-tailed significance level of α = 0.05 was adopted.

Results

Demographic data

In this cross-sectional study, 150 patients with pathologically confirmed NPC were ultimately retrospectively enrolled from Hainan General Hospital (Affiliated Hainan Hospital of Hainan Medical University). Table 1 summarizes the clinical characteristics of 150 patients with NPC, including G0 (pre-RT, n = 39), G1 (< 1 month post-RT, n = 30), G2 (< 6 months post-RT, n = 33), and G3 (> 6 months post-RT, n = 48). There were no significant differences in age and gender ratio among groups (G0–G3) (P > 0.05).

Table 1.

Basic characteristics of patients in different groups

Demographic	Patients, No. (%)				P-value
	G0(Pre-RT, n = 39)	G1(< 1 m, n = 30),	G2(1–6 m, n = 33)	G3(> 6 m, n = 48)
Age, years	43.28 ± 10.96	48.43 ± 11.95	46.30 ± 11.62	47.10 ± 11.50	0.271a
Sex(male/female)	32/7	23/7	28/5	37/11	0.816b
Cancer Characteristic
T stage					0.587b
T1	0/39(0.0%)	0/30(0.0%)	1/33(3.0%)	0/48(0.0%)
T2	8/39(20.5%)	8/30(26.7%)	5/33(15.2%)	13/48(27.1%)
T3	22/3(56.4%)	12/30(40.0%)	19/33(57.6%)	26/48(54.2%)
T4	9/39(23.1%)	10/30(33.3%)	8/33(24.2%)	9/48(18.8%)
N stage					0.432b
N0	0/39(0.0%)	3/30(10.0%)	1/33(3.0%)	4/48(8.3%)
N1	11/39(28.2%)	8/30(26.7%)	8/33(24.2%)	18/48(37.5%)
N2	16/39(41.0%)	11/30(36.7%)	17/33(51.5%)	18/48(37.5)
N3	12/39(30.8%)	8/30(26.7%)	7/33(21.2%)	8/48(16.7%)
M stage					0.883b
M0	38/39(97.4%)	29/30(96.7%)	33/33(100.0%)	47/48(97.9%)
M1	1/39(2.6%)	1/30(3.3%)	0/33(0.0%)	1/48(2.1%)

Pre-RT before radiotherapy, m months;

^a p values were calculated with two-tailed t tests

^b p value was obtained using a chi-squared test

Group comparisons of parameters

Fiber tractography of the anterior visual pathway using a high-definition approach was successfully achieved in all participants. After fiber tracking, DTI and GQI parameters were extracted for group comparisons (Table 2). All parameters exhibited consistent temporal patterns of evolution (Fig. 3). Compared with the pre-RT group (G0), the most sigfinicant changes occurred within 1 month after RT (G1), then improved and became comparable to their pre-treatment values, with similar trends observed bilaterally. As shown in Table 2, FA values decreased significantly within 1 month after RT (G1) compared with pre-RT (G0) values (P < 0.05), and gradually returned after the first month. Mean MD, RD, and AD values increased significantly within 1 month after RT (G1) compared with pre-RT (G0) values (P < 0.05), and subsequently returned toward pre-RT levels after 1 month. No statistically significant differences were observed in mean GFA values across the groups before or after RT (P > 0.05). Mean NQA values decreased significantly within 1 month after RT (G1) compared with the pre-RT (G0) group (P < 0.05), and returned to pre-RT (G0) values after 1 month. Line graphs further indicated that, although fluctuations were observed after the first month, both DTI and GQI metrics showed a trend toward normalization, with no statistically significant differences compared with pre-RT values (P3:G2 vs. G0; and P4:G3 vs. G0, both P > 0.05; Fig. 4a and b).

Table 2.

Changes in each parameter of DTI, GQI in NPC before and post-RT in the anterior Visual Pathway

	nerve	G0(Pre-RT)		G1(< 1 m)	G2(1–6 m)	G3(> 6 m)	F	P1	P2	P3	P4
DTI parameters
FA	AVP_L		0.296 ± 0.029	0.274 ± 0.035	0.312 ± 0.090	0.295 ± 0.032	2.764	0.019*	0.027*	0.551	0.872
	AVP_R		0.299 ± 0.027	0.279 ± 0.036	0.322 ± 0.088	0.297 ± 0.033	4.077	0.022*	0.043*	0.307	0.872
MD	AVP_L		0.992 ± 0.067	1.054 ± 0.080	1.004 ± 0.194	1.016 ± 0.077	1.768	0.012*	0.012*	0.872	0.280
	AVP_R		0.988 ± 0.080	1.052 ± 0.084	0.979 ± 0.191	1.013 ± 0.086	2.563	0.013*	0.012*	0.872	0.307
AD	AVP_L		1.279 ± 0.071	1.332 ± 0.081	1.296 ± 0.243	1.310 ± 0.079	1.085	0.042*	0.027*	0.872	0.163
	AVP_R		1.277 ± 0.082	1.336 ± 0.083	1.276 ± 0.242	1.309 ± 0.080	1.576	0.035*	0.004*	0.988	0.074
RD	AVP_L		0.848 ± 0.070	0.915 ± 0.085	0.858 ± 0.172	0.869 ± 0.082	2.245	0.010*	0.012*	0.872	0.368
	AVP_R		0.843 ± 0.081	0.911 ± 0.088	0.830 ± 0.168	0.866 ± 0.092	3.178	0.010*	0.023*	0.988	0.190
GQI parameters
GFA	AVP_L		0.137 ± 0.009	0.132 ± 0.011	0.137 ± 0.010	0.138 ± 0.020	2.428	0.072	-	-	-
	AVP_R		0.137 ± 0.008	0.133 ± 0.009	0.139 ± 0.011	0.138 ± 0.010	2.292	0.083	-	-	-
NQA	AVP_L		0.177 ± 0.025	0.157 ± 0.025	0.179 ± 0.033	0.168 ± 0.029	3.793	0.005	0.011*	0.872	0.307
	AVP_R		0.178 ± 0.024	0.159 ± 0.024	0.184 ± 0.031	0.171 ± 0.029	5.286	0.002	0.012*	0.489	0.388

Abbreviations: AVP_L, Left Anterior Visual Pathway; AVP_R, Right Anterior Visual Pathway; Pre-RT before radiotherapy, m months; RT, radiotherapy; DTI, diffusion tensor imaging; GQI, generalized q‑sampling imaging; FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; GFA, generalized fractional anisotropy; NQA, normalized quantitative anisotropy; L, left; R, right.Data are mean ± SD. P1: overall ANOVA; P2:G1 vs. G0; P3:G2 vs. G0; P4:G3 vs. G0;*P < 0.05, FDR corrected

Fig. 3.

Pre-RT before radiotherapy, m months; Temporal trends of each parameter across the four time points. Pre-RT before radiotherapy, m months; G0 (pre‑RT), G1 (< 1 m post‑RT), G2 (1–6 m post‑RT), and G3 (> 6 m post‑RT)

Fig. 4.

Bar chart of each parameter across different time points. L(A) and R(B) represent the left and right sides, respectively. * indicate significant differences compared with the G0 group (*P < 0.05, **P < 0.01, ***P < 0.001).G0 (pre‑RT), G1 (< 1 m post‑RT), G2 (1–6 m post‑RT), and G3 (> 6 m post‑RT). Pre-RT before radiotherapy, m months

Group discrimination of parameters based on Cohen’s d

Cohen’s d (with 95% confidence intervals) was used to quantify intergroup differences for each parameter compared with the pre-RT group (G0) (Fig. 5). Overall, all metrics exhibited the largest effect sizes when comparing the within 1 monthpost-RT group (G1) with the pre-RT group (G0), suggesting that microstructural alterations are most pronounced during this period across all post-RT groups. In contrast, effect sizes were smaller when comparing the 1–6 months post-RT group (G2) with the pre-RT group (G0) (left: 0.03–0.24; right: 0.00–0.37), and were small to moderate when comparing the >6-month post-RT group (G3) with the pre-RT group (G0) (left: 0.05–0.42; right: 0.06–0.39). The effect size heatmap (Fig. 6) further demonstrated that in the G1 group, effect sizes for most parameters were consistently higher, with similar patterns between the left and right sides. Among these, larger effect sizes were observed for RD, MD, and NQA (d ≈ 0.79–0.87), indicating that these parameters tend to more sensitive to microstructural changes during the G1 period.

Fig. 5.

Forest plot of effect sizes.Effect sizes are presented for pairwise comparisons: Pre-RT before radiotherapy, m months; L and R represent the left and right sides, respectively.G0 (pre‑RT), G1 (< 1 m post‑RT), G2 (1–6 m post‑RT), and G3 (> 6 m post‑RT). G1 vs. G0, G2 vs. G0, and G3 vs. G0. The points represent Cohen’s d estimates, and the error bars indicate the 95% confidence intervals

Fig. 6.

Effect size heatmap. The color intensity corresponds to the magnitude of |d|.Pre-RT before radiotherapy, m months; G0 (pre‑RT), G1 (< 1 m post‑RT), G2 (1–6 m post‑RT), and G3 (> 6 m post‑RT). Results for the left and right sides are shown separately in two panels

Discussion

As demonstrated in previous studies, DTI and GQI can detect microstructural alterations in white matter, and the derived fiber tract parameters show strong concordance with histopathologic changes [26–28]. To our knowledge, this study is the first to explore dynamic changes in the anterior visual pathway at different post-RT stages in NPC patients using both DTI and GQI. The results exhibited dynamic change patterns across different post-RT stages: significant alterations occurred within the first month post-RT, followed by partial improvement. These findings may highlight critical windows for early intervention following radiotherapy in NPC patients.

Diffusion parameters: MD, RD, and AD

MD often reflects overall tissue water diffusivity and is widely used to assess changes in diffusion properties resulting from alterations in the intra- and extracellular microenvironment after radiotherapy [29]. Radiation injury to vascular endothelial cells can occur within 24 h after exposure, leading to blood–brain barrier disruption and subsequent vasogenic edema, which increases the proportion of free water and enhances water diffusion in all directions, with a resultant increase in MD values. Increased RD is typically associated with compromised myelin integrity, whereas AD may reflect alterations in the axonal microenvironment and intratract water distribution. In a longitudinal DTI study, Nagesh et al. reported that early white matter injury following radiotherapy is characterized by a prominent increase in RD, suggesting that demyelination constitutes the primary mechanism underlying early neural fiber damage [30]. Using a mouse model of congenital dysmyelination (shiverer mice), Song et al. histopathologically demonstrated that isolated demyelination presents as increased RD with relatively preserved AD [31]. In contrast, our study showed that MD, AD, and RD were all significantly increased within the first month after RT, a profile distinct from that of isolated demyelination. Using Monte Carlo simulations and tissue phantom experiments, Chiang et al. further validated that a marked elevation in AD is indicative of increased vasogenic edema [32]. Based on these observations, we speculate that the early-stage injury in our study may involve both demyelinating changes and vasogenic edema. Given that multiple pathological changes may coexist in early injury, an increasing number of recent studies have tended to suggest that the core of early tissue injury after radiotherapy may be more attributed to vascular endothelial damage. Levin et al. provided direct histopathologic evidence supporting this concept in their study of irradiated optic nerves, demonstrating a reduction of approximately 65% in the number of endothelial cells on cross-section in nerves exposed to high-dose radiation compared with those receiving low-dose radiation—a finding that confirms vascular endothelial cells as one early target of radiation injury [33]. In a systematic review, Gorbunov et al. proposed that following cranial irradiation, microvascular endothelial cells—owing to their high radiosensitivity—sustain damage that disrupts the blood–brain barrier (BBB) and triggers vasogenic inflammatory responses. This process is recognized as a common pathological substrate underlying radiation-induced brain injury across the early, progressive, and late-delayed phases [34]. As reported in a systematic review of preclinical studies by Nakkazi et al., radiation exposure increases BBB permeability, reduces endothelial cell density, and downregulates the expression of tight junction proteins, noting that “in most cases, vascular damage is an early event and does not resolve quickly” [35]. Furthermore, Li et al., using a rat spinal cord model of radiation-induced central nervous system injury, demonstrated that vascular endothelial damage may directly contribute to barrier disruption and vasogenic edema [36]. Collectively, the marked increase in MD, AD, and RD observed in the G1 group in our study are best interpreted as a composite pathological process initiated by vascular endothelial injury, followed sequentially by vasogenic edema, demyelinating changes, and axonal swelling. Consistent with this interpretation, effect size analysis revealed strong effects for both MD and RD bilaterally during the G1 period, indicating that these parameters may offer better diagnostic performance in this early phase.

After radiotherapy, MD significantly decreased during the G1 period, then recovered toward pre-RT values with no statistical difference compared with pretreatment values, suggesting gradual resolution of acute edema and demyelination-related changes and indicating some degree of spontaneous recovery of nerve fiber integrity. These findings are consistent with those of Duan et al., who reported similar longitudinal changes in WM microstructure using whole-brain DTI in patients with NPC after radiotherapy [37]. The decrease in RD to pre-RT levels one month after radiotherapy suggests that myelin-related abnormalities may be partially reversible or may have entered a reparative phase. AD did not increase sustainably further, which argues against typical progressive axonal disruption, implying that the observed abnormalities more likely reflect mild, reversible microstructural alterations rather than Wallerian degeneration [38].

Duan et al. reported that FA and MD exhibited a trend toward recovery between six and twelve months after radiotherapy, indicating that these changes recovered gradually. However, at twelve months, FA remained significantly lower than pretreatment levels in the right frontal and parietal lobes, and MD remained significantly higher in the right occipital lobe, suggesting incomplete recovery [37]. In our study, MD, RD, and AD tended to increase at six months after radiotherapy, which suggests that radiation-induced injury to the anterior visual pathway may not fully recover and may persist. Previous studies have shown that microstructural changes of the anterior visual pathway have already occurred in patients before the visual symptoms [39, 40]. Wang et al. reported that the mean latency from radiotherapy to the onset of visual loss is approximately 36 months [41]. Wu et al., reported that radiation-induced optic neuropathy (RION) has a long latent period, with patients experiencing a long deterioration course and irreversible visual loss, and most cases occurs within 3 years after radiotherapy [42]. Accordingly, given the limited follow-up time of our study, this finding indicates that the “recovery” observed after early radiation-induced injury is not a true tissue repair. The microstructural abnormalities observed in our study may represent preclinical pathological changes of RION, which could progress to overt clinical manifestations after a latency of months to years. Once RION develops, it leads to severe and irreversible visual loss, adversely affecting patient quality of life.

Anisotropy and fiber structure metrics: FA, GFA, and NQA

The FA and GFA are proven to be sensitive to coherent orientation, such as the integrity of the axonal membrane and myelin. Higher FA often reflects more coherent orientation and better white matter fiber tract integrity. In our study, both FA and NQA showed significant decreases in the G1 group compared with pre-RT values, suggesting early disruption of structural integrity in the white matter of the anterior visual pathway after radiotherapy, likely resulting from demyelination, axonal injury, and microenvironmental alterations. The increased MD, AD, and RD, together with decreased FA, collectively suggest that complex pathological changes exist during this period [12, 17, 22]. Ling et al. used GQI to assess the microstructural impairment of the visual pathway in patients with type 2 diabetes mellitus and found that NQA, QA, and GFA were significantly lower than those in healthy controls, indicating that long-term hyperglycemia leads to disruption of axonal and myelin microstructural integrity. These findings provide direct evidence linking decreased NQA values to axonal and myelin damage [43]. In our study, most diffusion parameters recovered to near pre-RT levels one month after radiotherapy. However, at six months, AD and GFA increased, whereas NQA decreased, concomitant with increases in MD and RD. Together, these parameters suggest that radiation-induced microstructural injury to the anterior visual pathway exhibited dynamic temporal changes and a complex pathological process.

Previous studies have shown that fiber tract architecture may become distorted in the setting of marked edema, and factors such as fiber distortion, crossing fibers, or white matter injury can contribute to decreased FA. Moreover, the effect size for FA in our study was less than 0.8. Therefore, decreased FA may have limitations when used to assess microstructural injury to the anterior visual pathway [44]. Compared with conventional DTI, GQI offers advantages in resolving crossing, branching, or otherwise complex intravoxel fiber orientations [17, 22]. Although radiotherapy irreversibly leads to microstructural damage within the irradiation field, GFA did not show significant differences between pre-RT and post-RT groups in our study. However, this does not imply that radiotherapy fails to cause microstructural injury to the anterior visual pathway. As an orientation distribution function (ODF)- or spin distribution function (SDF)-derived parameter, GFA reflects the overall coherent orientation of fiber tracts and may be influenced easily by partial volume effects, contamination from adjacent cerebrospinal fluid (CSF), and echo-planar imaging (EPI) distortions or registration errors — particularly in small structures such as the anterior visual pathway [17, 22, 23, 45–47]. The anterior visual pathway contains crossing nerve fibers, and its proximity to air-containing cavities such as the skull base and paranasal sinuses may also be prone to causing magnetic field inhomogeneities in the local region, leading to geometric distortion and signal dropout that reduce the reliability of GFA, thereby lowering its sensitivity for detecting radiotherapy-induced microstructural injury. Shen et al. reported that GFA shows non-ideal diagnostic performance for white matter injury when assessing relatively isotropic tissues or more complex structures [48]. In contrast, NQA minimizes interindividual variability and reduces partial volume effects, providing more reliable values of white matter (WM) microstructure and greater suitability for group comparisons, with better reliability than QA. In our study, the effect size for NQA in the G1 group was higher than GFA、FA, suggesting that NQA may be more sensitive and stable for detecting microstructural injury during this period in the anterior visual pathway.

Furthermore, higher-order anisotropy parameters are highly dependent on signal-to-noise ratio (SNR) and b-value. Although the number of MR diffusion directions in this study met the minimum requirement for HARDI, SNR and residual distortions may still limit the stability and sensitivity of parameters such as GFA for detecting microstructural alterations [49, 50]. In comparison, some DTI parameters showed slightly larger effect sizes than GQI parameters, likely because these parameters are more sensitive to changes in the setting of edema and myelin damage. This finding should not be interpreted as evidence that GQI is inferior to DTI [30, 51–53]. Although the anterior visual pathway contains anatomical fiber crossings, the proportion of crossing fibers is relatively small; thus, this effect may be diminished across the anterior visual pathway.

Limitations

This study has several limitations. First, the absence of histopathologic correlation and limited follow-up time make it difficult to fully verify the pathological changes in the anterior visual pathway after radiotherapy. Second, the cross-sectional study and the small number of patients limit the ability to fully characterize microstructural alterations after radiotherapy. Future multicenter, large-scale prospective longitudinal studies are needed to validate these findings and to inform early intervention strategies aimed at reducing the risk of irreversible visual impairment. Third, Some differences in the results may be introduced by manual tracking settings, including ROI location, ROI size. Finally, Therapeutic regimens and overall radiotherapy prescription doses were not reported in this study, which may affect results. Future studies with dosimetric data are needed to validate our conclusions.

Conclusion

In conclusion, we employed DTI and GQI to investigate changes in the anterior visual pathway white matter at different time points after radiotherapy for NPC. The RD values showed greater sensitivity within one month after radiotherapy. Our findings significantly advance the understanding of the pathophysiological mechanisms underlying RION and provide a technique to monitor injury in the anterior visual pathway.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Yaqian Xing: Writing‑original draft, Formal analysis. Tiansheng Li: Writing‑review and editing, Formal analysis. Yiying Zhang: Data curation. Linjie Zhou: Data curation. Gang Wu: Data curation, Methodology, Supervision, Writing‑review and editing. Weiyuan Huang: Writing‑original draft, Funding acquisition, Conceptualization, Supervision. All authors read and approved the final manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (82260339 and 82402240), Academic Enhancement Support Program of Hainan Medical University (XSTS2026178) and the Hainan Province Clinical Medical Center.

Data availability

The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Human Ethics Committee (Institutional Review Board, IRB) of Hainan Hospital Affiliated to Hainan Medical University and Hainan Provincial People’s Hospital. Inform consent was obtained from all subjects. The study was carried out in accordance with Helsinki Declaration.

Consent for publication

Not applicable. This manuscript does not contain any identifying images or clinical details of patient(s) that would compromise anonymity.

Competing interests

The authors declare no competing interests.