Differentiation of MPNSTs from Benign Neurofibromas in Neurofibromatosis Patients Using ADC and SUV Biomarkers in children and young adults

Abstract

Purpose

This study evaluated the sensitivity and specificity of the apparent diffusion coefficient (ADC) and standardized uptake value (SUV) to differentiate malignant peripheral nerve sheath tumors (MPNSTs) from benign neurofibromas in children and young adults with NF1.

Methods

We included 31 patients (19 male, 12 female, age 1–24 years) with 33 nerve sheath tumors, including 11 MPNSTs and 22 benign neurofibromas underwent MRI with diffusion-weighted imaging (DWI, n = 30 lesions, 22 benign, 8 malignant) or 18F-FDG PET (n= 22 lesions, 11 benign, 11 malignant). Tumor apparent diffusion coefficients (ADCmean, ADCmin, and ADCmax) and standardized uptake values (SUVmax) were compared between benign and malignant lesions using Wilcoxon rank-sum tests. In addition, ROC curves were generated to determine optimal cutoffs for each parameter and their respective area under the curve (AUC), sensitivity and specificity. We used Spearman’s rank correlation to assess the correlation between ADCmin and SUVmax.

Results

Mean ADCmin and ADCmean were significantly lower in MPNST compared to benign neurofibromas (p<0.001). Mean SUVmax was significantly higher in MPNSTs (11.05±4.80) compared to benign lesions (2.21±0.68, p<0.001). An ADCmin cutoff of 1.1×10⁻³ mm²/s differentiated MPNST from benign neurofibromas with a sensitivity of 87%, specificity of 100%, and AUC of 0.97. A SUVmax threshold of 4.9 yielded a sensitivity of 91%, specificity of 100%, and AUC of 0.99. ADCmin correlated inversely with SUVmax (rho= −0.81, p < 0.0001). Longitudinal clustered regression showed progressive declines in ADCmean, ADCmax, and ADCmin in benign neurofibromas, with significant reductions in ADCmean and ADCmax at 2 and 3 years and in ADCmin at 3 years relative to baseline.

Conclusion

Either ADC metrics or SUVmax could differentiate MPNSTs from benign neurofibromas in children and young adults with high sensitivity and specificity.

Graphical Abstract

Introduction

Patients with neurofibromatosis type 1 (NF1) frequently develop numerous benign peripheral nerve sheath tumors (neurofibromas) (1). About 8–13% of these lesions will transform to malignant peripheral nerve sheath tumors (MPNSTs) (2). MPNSTs are rare, aggressive, and high-grade soft-tissue sarcomas, characterized by a high rate of local recurrence, metastasis, and poor survival (3). The 5-year survival rate for NF1-associated MPNST is about 40% (2, 4) and 40% of patients with MPNSTs develop distant metastasis (5). Late diagnosis of MPNST is associated with poor prognosis (6), therefore, early detection is critical.

MRI is the standard imaging modality in tumor surveillance in NF1 patients, offering high-resolution anatomical detail and soft tissue contrast (7). However, MRI features of benign and malignant peripheral nerve sheath tumors can overlap significantly, particularly in lesions that are deep-seated, large, or atypical in appearance (8). Histopathological examination continues to be regarded as the definitive 'gold standard' for diagnosis (9–11). However, the routine use of surgical biopsy is invasive and impractical for patients with multiple suspicious lesions. The development of quantitative imaging thresholds that could direct or, in certain circumstances, obviate the necessity for surgical biopsy would be highly beneficial for patient management.

18F-fluorodeoxyglucose positron emission tomography (18F-FDG PET) can provide quantitative measures of tumor glucose metabolism (6, 12, 13); in parallel, MRI diffusion-weighted imaging (DWI) has emerged as a valuable adjunct technique, allowing quantitative assessment of tumor cellularity through the apparent diffusion coefficient (ADC) (14). Ristow et al. (15), Combemale et al. (6), Ahlawat et al. (16), Chirindel et al. (17), and Reinert et al. (18) reported different thresholds of tumor SUV values for the differentiation of benign and malignant nerve sheath tumors in adult patients. In addition, Ahlawat et al. (16), Koike et al. (19), and Reinert et al. (18) reported different thresholds of apparent diffusion coefficients for the differentiation of benign and malignant nerve sheath tumors in adult patients.

Limited information is available about longitudinal changes of ADC and SUV of PNSTs (malignant and benign) in children and young adults. Most previous studies have been conducted in patient cohorts with wide age rages of 60 years or more. Therefore, the purpose of our study was to evaluate the sensitivity and specificity of ADC and SUV to differentiate MPNSTs and benign neurofibromas in children and young adults with NF1.

Material and Methods

Study Population

This retrospective study was approved by the institutional review board at Stanford University under protocol IRB-78002. Informed consent was waived due to the retrospective nature of our image analyses. We retrospectively reviewed pediatric and young adult patients (age < 25 years) with a confirmed diagnosis of NF1 and peripheral nerve sheath tumors (PNST) who underwent either MRI with diffusion-weighted imaging (DWI) or 18F-FDG PET at our institution. Given the limited number of patients, we separately identified eligible cohorts for MRI and PET analyses to maximize inclusion. For the MRI cohort, patients were included if they had available ADC maps. Patients with incomplete or poor-quality imaging (n = 2) were excluded, resulting in 29 patients with evaluable DWI-MRI, including 7 patients with MPNSTs (8 lesions) and 22 patients with benign neurofibromas (lesion count matched patient count) (Fig. 1A). MPNST diagnosis was confirmed by histopathology in all cases via biopsy or surgical resection. Benign neurofibromas were confirmed either histologically, when tissue samples were available, or through imaging and clinical follow-up demonstrating stability over a period of at least 12 months without progression or malignant features. For the PET cohort, patients were included if they had undergone 18F-FDG PET imaging. A total of 20 patients met inclusion criteria, comprising 9 with MPNSTs (11 lesions) and 11 with benign neurofibromas (Fig.1B). Among the MPNST lesions, 8 had both MRI with ADC maps and SUVmax values acquired on a combined PET/MRI system, while the remaining 3 had PET data available without corresponding ADC maps. All 22 benign lesions underwent MRI with ADC analysis; among these, 11 patients had paired SUVmax measurements from PET/MRI studies.

Fig. 1: Patient with (A) MRI and (B) PET scan selection flowchart.

In addition, 16 of the 22 benign neurofibromas underwent follow-up MRI within 1 year, 19 of the 22 benign neurofibromas underwent follow-up MRI within 2 years, and 15 of 22 benign neurofibromas underwent follow-up MRI within 3 years after the initial scan. Patients with benign neurofibroma included 16 males (72.7%) and 6 females (27.3%) with a mean age of 9.3 ± 5.6 years (range: 1–22 years). Patients with MPNSTs included 3 males (33.3%) and 6 females (66.7%) with a mean age of 15.1 ± 4.7 years (range: 8–24 years). Of the 29 patients, 27 were aged ≤19 years at the time of imaging; 1 patient with MPNST and 1 with a benign lesion were aged 20–24 years.

Magnetic Resonance Imaging

MRI studies were performed using a scanner (GE Healthcare, Milwaukee, WI) or 3T PET/MRI system (Signa; GE Healthcare), using appropriate surface coils and the following pulse sequences: Non-fat-saturated T1-SE (spin echo), T2-FSE (fast spin echo) with and without fat-saturation, diffusion-weighted single-shot spin-echo echo-planar imaging (EPI) with two b-values (b = 0–50 s/mm² and b = 800–1000 s/mm²) and a postcontrast fat-saturated T1-weighted gradient-echo sequence after intravenous injection of gadobutrol (Gadavist, 0.1 mmol Gd/kg). Pulse sequence parameters for all sequences are listed in Table 1. ADC maps were automatically generated from the DWI sequences using the GE Signa scanner software. This process involved calculating the slope of the logarithmic signal intensity relative to the b-value, allowing quantitative assessment of the tissue diffusivity.

Table 1:

MRI acquisition parameters

Sequence	Slice thickness (mm)	Spacing (mm)	FOV (mm)	Matrix size (pixel)	TR (ms)	TE (ms)	Flip angle (°)
T1 Spin echo	1 – 6	1 – 6	160 – 320	240–512 × 192–320	457 – 1083	8.7 – 16.8	90 – 142
T2 Fast spin echo	2 – 4	2 – 5	160 – 240	320–512 × 224–320	3212 – 8636	79.3 – 106	111 – 142
T2 Fat saturated Fast spin echo	1 – 5	0.5 – 6	140 – 220	256–512 × 192–260	2646 – 7468	81.6 – 107	90 – 160
Diffusion weighted	3 – 6	3 – 6	160 – 320	80–128 × 80–128	4000 – 11000	48.1 – 104	90
Post-contrast T1 gradient echo	1 – 2	0.5 – 2	140 – 380	240–416 × 220–416	4.3 – 8.6	0.9 – 3.3	8 – 15

A postdoctoral research fellow (ZSV) and a general radiologist with five years of experience (YUL) identified solid, contrast-enhancing, and non-cystic portions of each lesion on conventional MR imaging and manually drew regions of interest (ROIs) over the solid tumor part to measure ADCmin and ADCmean of PNST, using Sectra PACS software (version 25.2.22.7826, Sectra AB, Sweden). The readers were blinded to the clinical information of the patients and the histological tumor diagnoses.

18F-FDG Positron Emission Tomography

All 18F-FDG PET scans were performed on a system (Signa; GE Healthcare). Patients fasted for at least 4 hours prior to 18F-FDG radiotracer injection, and serum glucose levels were confirmed to be <120 mg/dL at the time of injection. A standardized dose of 3 MBq/kg 18F-FDG was administered intravenously, in accordance with guidelines from the Society of Nuclear Medicine and Molecular Imaging and the European Association of Nuclear Medicine (20). During the uptake phase, patients were instructed to rest quietly with their eyes closed in a dimly lit room without sedation.

For attenuation correction, non-breath-hold axial Dixon sequences were acquired (TR/TE: 4.6 – 6.1/1.71–1.74 ms, FOV: 500–600 mm, slice thickness: 2.78 mm, matrix: 320 × 224, using a body receiver coil), and segmented attenuation maps were generated. High-resolution breath-hold gradient echo sequences were obtained for PET-MRI co-registration.

A postdoctoral research fellow (ZSV) and a general radiologist with five years of experience (YUL), blinded to the clinical and histopathologic information measured tumor SUVmax using MIM software (version 7.2.8; MIM Software Inc.) as the maximum voxel-level standardized uptake value within an operator defined region of interest (ROI), that outlined the entire tumor on an individual image slice.

Statistical analysis

Wilcoxon rank-sum tests were used to compare imaging metrics (ADCmin, ADCmean, ADCmax, SUVmax) between MPNST and benign neurofibroma groups. Receiver operating characteristic (ROC) curves were generated to assess the diagnostic performance of each metric and to calculate the area under the curve (AUC). The optimal cutoff point for each parameter was identified using the Youden index, and corresponding sensitivity and specificity values were calculated. To evaluate the monotonic relationship between SUVmax and ADCmin values, Spearman’s rank correlation analysis was performed, and a rank-rank scatter plot was generated by plotting the rank order of each lesion’s ADCmin value against its corresponding SUVmax rank. Longitudinal data for ADC metrics were analyzed using clustered linear regression. Cutoff values were determined by maximizing Youden's index, the sum of sensitivity and specificity. A p-value < 0.05 was considered statistically significant. All statistical analyses were performed using the statistical package for social sciences software (IBM SPSS statistics for Mac, version 28.0, Armonk, NY: IBM Corp.). Graphics were created using R version 4.4.1. and the R package pROC version 1.18.5.

Results

All 11 the MPNST lesions had histologic confirmation based on biopsy or surgical pathology. Among the 22 benign neurofibromas, 7 lesions were histologically confirmed, and the remaining were clinically diagnosed and monitored with serial imaging for at least 12 months, showing no interval growth or imaging features suggestive of malignant transformation.

Magnetic Resonance Imaging

Benign neurofibromas presented as well-circumscribed, T1-hypointense, and T2-hyperintense lesions compared to adjacent skeletal muscle, with mild to moderate contrast enhancement on post-contrast T1-weighted images. Ten of 22 benign neurofibromas demonstrated a target sign on T2-weighted MRI (Fig. 2 B-D). Benign neurofibromas exhibited high signal intensity on ADC maps, consistent with low cellularity and unrestricted diffusion (Fig. 2 E & F) (Table 3).

Fig. 2. Representative imaging studies of a benign plexiform neurofibroma in a 7-year-old male with neurofibromatosis type 1 (NF1).

(A) Axial fused 18F-FDG PET/MR shows no metabolic activity of the lesion (arrows, both with SUVmax: 1.5). (B) Axial T1-weighted image demonstrates a well-circumscribed, mildly hypointense soft tissue mass (arrows). (C) Axial T2-weighted fat-suppressed (FS) fast spin echo (FSE) image reveals a typical neurofibroma with target sign (hypointense center and hyperintense periphery, red arrow). In addition, there is a lobulated, T2-hyperintense lesion which does not show the target sign (arrows). (D) Axial post-contrast T1-weighted FS-FSE image shows minimal contrast enhancement of the lesion (arrows). (E) Axial diffusion-weighted image (DWI) shows high diffusion restriction within the lesion (arrow). (F) Axial ADC map demonstrates high signal intensity consistent with unrestricted diffusion. The measured ADCmean was 2.0 ×10⁻³ mm²/s (yellow arrow), consistent with a benign lesion.

Table 3:

Summary of imaging characteristics for benign neurofibromas and MPNSTs in pediatric and young adult NF1 patients. Quantitative variables include mean lesion volume, lesion growth rate, ADCmin, and SUVmax.

Parameter	Benign Neurofibroma	MPNST	p-value
Mean ± SD Volume, cm3	49.39 ± 118.2	388.1 ± 609.8	0.002
Lesion Growth Rate	+18.24 cm3 per year (p value=0.028)	-	-
ADCmin, mm2/s	1.9 ± 0.3 ×10⁻3	0.9 ± 0.1 ×10⁻3	p < 0.001
SUVmax	2.2 ± 0.7	11.0 ± 4.8	p < 0.001

In contrast, MPNSTs presented as large, heterogeneous T1-hypointense, and T2-hyperintense lesions with avid, inhomogeneous contrast enhancement (Fig. 3 C-E). None of 11 MPNST demonstrated a target sign on T2-weighted MRI and nine of the 11 MPNST demonstrated central necrosis on contrast-enhanced MRI. MPNST exhibited low signal intensity on ADC maps, consistent with high cellularity and restricted diffusion (Fig. 3 F) (Table 3).

Fig. 3. Representative imaging studies of a malignant peripheral nerve sheath tumor (MPNST) in a 17-year-old male with NF1.

(A) Coronal 18F-FDG PET image shows intense focal 18F-FDG uptake in a large soft tissue lesion in the left upper extremity (arrow). (B) Sagittal fused 18F-FDG PET/MR image demonstrates high metabolic activity within the mass (SUVmax: 10). (C) Sagittal T1-weighted SE image shows a large soft tissue mass. Large portions of the mass are isointense compared to adjacent muscle (arrow), with a focal hyperintense area of central necrosis (arrowhead). (D) Sagittal post-contrast T1-weighted fat-suppressed FSE image reveals heterogeneous hyperintense enhancement of the mass. (E) Sagittal T2-weighted fat-suppressed FSE image demonstrates heterogenous hyperintense signal of the mass, with loss of the target sign. (F) Axial ADC map shows markedly low diffusion signal, reflecting restricted diffusion (arrow). The central hemorrhagic necrosis does not show restricted diffusion (arrowhead). (G) Sagittal DWI image displays high signal intensity within the soft tissue component of the mass, indicating restricted diffusion (arrow). (H) Sagittal ADC map confirms low ADC values of the soft tissue components of the mass, with a measured ADCmean of 0.8 ×10⁻³ mm²/s, consistent with hypercellularity.

Accordingly, mean ADCmin of MPNSTs was significantly lower (0.9 ± 0.1 ×10⁻³ mm²/s) compared to benign neurofibromas (1.9 ± 0.3 ×10⁻³ mm²/s, p < 0.001, Fig. 4A). Similarly, ADCmean of MPNSTs (1.3 ± 0.3 ×10⁻³ mm²/s) was significantly lower compared to benign lesions (2.1 ± 0.3 ×10⁻³ mm²/s, p < 0.001, Fig. 4A).

Fig. 4. Comparison of apparent diffusion coefficients (ADC) and standardized uptake values (SUV) of benign neurofibromas and MPNSTs.

(A) ADCmin and ADCmean were significantly lower in MPNSTs (n = 8) vs. benign lesions (n = 22). (B) Mean SUVmax was significantly higher in MPNSTs (n = 11) compared to benign neurofibromas (n = 11). (C) Rank-order scatterplot shows a strong inverse correlation of SUVmax and ADCmin (Spearman’s ρ = –0.81, p < 0.0001); MPNSTs consistently demonstrated high SUV and low ADC values.

The ROC analyses demonstrated clear separation between MPNST and benign neurofibroma lesions for ADCmin and ADC mean (Fig. 5). An ADCmin cutoff of 1.1×10⁻³ mm²/s provided the best overall performance, achieving a sensitivity of 87%, specificity of 100%, and an AUC of 0.97 (Table 2).

Fig. 5. Receiver operating characteristic (ROC) analysis of ADC and SUV imaging biomarkers of benign and malignant peripheral nerve sheath tumors.

ROC curves for SUVmax, ADCmin, and ADCmean demonstrate high diagnostic performance. SUVmax had the highest AUC (0.99), followed by ADCmin (0.97) and ADCmean (0.94). Optimal cutoffs: SUVmax, 4.9; ADCmin, 1.0 × 10⁻³ mm²/s; and ADCmean, 1.9 × 10⁻³ mm²/s.

Table 2:

Optimal cutoff values for ADC and SUVmax, with corresponding sensitivity, specificity, and area under the curve (AUC) for distinguishing benign neurofibromas from MPNSTs in pediatric and young adult NF1 patients.

Parameter	Optimal cutoff	Sensitivity (%)	Specificity (%)	AUC
ADCmin	1.1 × 10−3 mm2/s	87	100	0.97
ADCmean	1.9 × 10−3 mm2/s	100	77	0.94
SUVmax	4.9	91	100	0.99

18F-FDG Positron Emission Tomography

Benign neurofibromas exhibited mild or no FDG uptake on 18F-FDG PET (Fig. 2), while, MPNSTs presented strong FDG uptake, consistent with elevated metabolic activity (Fig. 3A, Fig. 4A and B)

Accordingly, the mean SUVmax was significantly higher in MPNSTs (11.0 ± 4.8) compared to benign neurofibromas (2.2 ± 0.7, p < 0.001, Fig. 4B). The ROC analyses demonstrated clear separation between MPNST and benign neurofibroma lesions for SUVmax (Fig. 5). An SUVmax threshold of 4.9 yielded a sensitivity of 91%, specificity of 100%, and an AUC of 0.99.

A significant inverse correlation was observed between SUVmax and ADCmin across all lesions (Spearman’s rho = −0.81, p< 0.0001), suggesting that lesions with higher metabolic activity tend to exhibit lower diffusivity, consistent with increased cellular density (Fig. 4C).

Longitudinal analysis using clustered linear regression revealed a progressive decline in all ADC metrics of benign neurofibromas over time. Compared to baseline, ADCmean decreased significantly at both the 2-year (−363.2, p = 0.001; 95% CI: −565.6 to −160.7) and 3-year (−517.5, p < 0.001; 95% CI: −737.3 to −297.6) follow-up points. Similarly, ADCmax showed significant reductions at 2 years (−556.9, p = 0.001; 95% CI: −867.1 to −246.7) and 3 years (−710.8, p < 0.001; 95% CI: −988.6 to −432.9). For ADCmin, the decline reached statistical significance only at 3 years (−341.9, p = 0.018; 95% CI: −619.0 to −64.8) (Fig. 6).

Fig. 6. Longitudinal changes in ADC of benign and malignant peripheral nerve sheath tumors over time.

Box plots illustrate changes in ADCmin and ADCmean at baseline, 1-year, 2-year, and 3-year follow-up MRI scans. All data represent mean data of n = 19 imaging studies and standard deviations. Significant reductions in ADCmean were observed at both 2- and 3-year follow-ups compared to baseline, while ADCmin showed a significant decrease only at the 3-year time point.

Discussion

Our study demonstrated that either ADC or SUVmax can effectively differentiate MPNST from benign neurofibromas in children and young adults with neurofibromatoses. Among the quantitative imaging metrics evaluated, ADCmin and SUVmax showed the highest diagnostic accuracy.

Our findings align with previous research demonstrating that MPNSTs exhibit significantly lower ADCmin values compared to benign neurofibromas. Ahlawat et al. reported a mean ADCmin of 0.6 ×10⁻³ mm²/s in MPNSTs versus 1.6 ×10⁻³ mm²/s in benign lesions (p < 0.0001), with an ADCmin threshold of 1.0 ×10⁻³ mm²/s yielding 100% sensitivity and 94% specificity for distinguishing malignant from benign peripheral nerve sheath tumors (16). While this study included patients with a wide age range (8–53 years), our study focused on children and young adults. In our younger patient cohort, an ADCmin threshold of 1.1×10⁻³ mm²/s differentiated MPNST from benign neurofibromas with a lower sensitivity of 87%, specificity of 100%. Our longitudinal studies demonstrated decreasing ADC values of benign lesions with increasing patient age and therefore, generally lower values in children. Using a similar threshold in children as previous authors used in adults, a relatively higher threshold in our study could be conservative enough to prevent benign lesions from being misidentified as malignant but not nuanced enough to capture the full range of ADC values indicative of malignancy in a younger population. This results in higher specificity at the cost of some sensitivity, primarily due to the distinct and uniform diffusion characteristics of benign lesions versus the more variable ADC values in malignant tumors in pediatric patients.

Similarly, our SUVmax threshold of 4.9 was higher than SUVmax thresholds reported in prior studies. Chirindel et al. reported an SUVmax threshold of 4.1 for the differentiation of MPNST and benign neurofibromas on 18F-FDG PET/CT, yielding 96% sensitivity and 88% specificity (17). Martin et al. conducted a meta-analysis of PET/MRI studies of 2939 lesions in patients with a wide age range of 2–88 and suggested an SUVmax threshold of ≥3.5 as a cutoff between benign and malignant peripheral nerve sheath tumors (21). Our higher threshold of 4.9 yielded 91% sensitivity and 100% specificity, because it minimized the inclusion of benign neurofibromas, which typically have lower SUVmax values, thus reducing false positives, while still accurately identifying most malignant cases.

Reinert et al. reported no significant difference in ADCmean between MPNSTs and benign plexiform neurofibromas (18), using ROIs that covered the entire tumor. This approach averaged the high ADC values from central necrotic and myxoid extracellular matrices with the low ADC values from solid tumor components, confounding the mean ADC data by including areas of both high and restricted diffusion. Additionally, their cohort had a broader age range and included composite lesions where malignant transformation was embedded within larger benign lesions, potentially diluting ADC differences.

Our data demonstrates a modest, gradual decline in ADC over time in benign PNST. This can reflect benign maturation, driven by increasing collagen and hyalinization that restrict diffusion. Benign neurofibromas can accumulate collagen, undergo hyalinization, and develop increased fibrous content, all of which restrict water diffusion and lower ADC. Pathology studies have described age-related changes in neurofibromas, including increased fibrosis and reduced proliferative activity, and conventional MRI often shows evolution of the “target sign,” with progressive central T2 hypointensity consistent with fibrous tissue. Within benign PNSTs, heterogeneity in ADC has been linked to the relative proportions of myxoid and fibrous components, where fibrous regions demonstrate lower ADC than myxoid regions (22, 23). In this context ADC typically remains in the range expected for benign neurofibromas. Previous studies generally placed ADCmin for benign neurofibromas around 1.1–1.6 ×10⁻³ mm²/s, for atypical neurofibroma/ANNUBP in an intermediate range of roughly 1.0–1.2 ×10⁻³ mm²/s, and for MPNST in a lower range often 0.7–0.9 ×10⁻³ mm²/s (exact cutoffs vary by sequence parameters and ROI methods) (24, 25). For longitudinal assessment, we suggest interpreting a slow decline that stabilizes above about 1.1–1.2 ×10⁻³ mm²/s as compatible with benign maturation, a drop into the ANNUBP range as a signal of emerging atypia that warrants closer follow-up or targeted biopsy, and a fall toward or below roughly 0.9–1.0 ×10⁻³ mm²/s as concerning for malignant transformation, especially if accompanied by accelerated growth, new pain, loss of the target sign, increasing heterogeneity, or infiltrative margins. In keeping with the lower variability of diffusion metrics, ADCmin or focal regions of lowest ADC may be particularly useful for biopsy targeting. While FDG SUV can provide supportive information, it overlaps substantially across PNST subtypes and in our cohort did not resolve uncertainty when ADC was equivocal. Therefore, longitudinal ADC trajectories and absolute thresholds, calibrated to local protocols, should anchor decision-making about an evolution to malignancy.

Our study has several limitations. First, the sample size was relatively small, as we specifically focused on children and young adults. However, data at this young age group are essential to better understand the unique imaging characteristics of these populations. Second, the progression from benign to malignant peripheral nerve sheath tumors (PNSTs) represents a biological continuum, making it unlikely that any threshold will be universally perfect. Several lesions in our cohort were unequivocally benign or malignant based on conventional MRI morphologic features, and in those cases ADC and SUVmax contributed little additional diagnostic information. The principal added value of ADC and SUVmax was in lesions with intermediate or overlapping features, where quantitative markers of cellularity (ADC) and glycolytic activity (SUVmax) helped shift diagnostic confidence from “indeterminate” to “likely benign” or “likely malignant.” These metrics reduced subjectivity, and improved triage in NF1 patients with multiple PNSTs by identifying which lesion warranted biopsy or urgent referral. They also highlighted intralesional heterogeneity, allowing biopsy to be directed to regions with low ADC/high SUVmax that are more likely to harbor malignant transformation, and supported longitudinal surveillance by flagging rising SUVmax or falling ADC as early indicators of progression. Accordingly, the added value of ADC and SUVmax is best understood as complementary to expert morphologic interpretation—most impactful in borderline cases, for biopsy guidance and risk stratification—rather than as replacements for clear-cut conventional MRI diagnoses. PNST/MPNST is rare, and our cohort represents a relatively large sample for a single-center study. However, the study design did not allow for advanced statistical analyses such as net reclassification improvement or decision-curve analysis. Future multicenter studies focusing on indeterminate PNSTs should use these methods to quantify the added value of ADC and SUVmax. Our relatively small cohort did not allow for formal advanced statistical analyses of the combined diagnostic performance of DWI and 18F-FDG PET. However, based on our practical clinical experience, we generated the following guidelines for our practice: lesions with ADCmin ≥1.2 ×10⁻³ mm²/s and low metabolic activity (SUVmax <3.5, or preferably a tumor to liver ratio <1.5) are very likely benign; lesions with ADCmin in the intermediate range of 1.0–1.2 ×10⁻³ mm²/s (consistent with ANNUBP) and elevated metabolism (SUVmax ≥4–5 or tumor to liver ratio ≥1.5) warrant closer surveillance or targeted biopsy; lesions with ADCmin ≤0.9–1.0 ×10⁻³ mm²/s are high risk for MPNST regardless of SUVmax, with SUVmax ≥6 further reinforcing suspicion. In our dataset, this combined approach reclassified a small subset of indeterminate lesions and aided biopsy targeting. These thresholds are provisional, require calibration to local acquisition and reconstruction parameters, and should be validated prospectively.

Nevertheless, clinical decision-making is inherently binary, requiring a clear recommendation for or against biopsy or surgery. Therefore, applying threshold values remains clinically justified. Third, we did not incorporate additional variables such as blood biomarkers or a broader integration of imaging, clinical, and histological data. Future studies could evaluate the value of incorporating such biomarkers into a comprehensive diagnostic algorithm, potentially combining ADC and SUV metrics to further enhance diagnostic accuracy.

The development of consensus guidelines for quantitative DWI and 18F-FDG PET would significantly improve consistency of NF1 surveillance protocols, such as whole-body MRI with targeted DWI, or selective PET imaging for lesions exceeding defined size or growth thresholds.

Conclusions

ADC measures and SUVmax are effective non-invasive imaging biomarkers for differentiating MPNST from benign neurofibromas in children and young adults with neurofibromatosis. Optimized ADCmin and SUVmax thresholds can reduce unnecessary biopsies, support more targeted biopsies or surgeries, and improve clinical outcomes for patients with NF1.

Abbreviations

NF

Neurofibromatosis

MPNST

malignant peripheral nerve sheath tumor

ADC

Apparent diffusion coefficient

SUV

standardized uptake values

SD

Standard deviation

TR

repetition time

TE

echo time

18F-FDG PET

Fluorodeoxyglucose positron emission tomography

FOV

Field of View

GLM

Generalized linear regression model

ROC

Receiver operating characteristic

AUC

Area under the curve