Manual volumetric segmentation of brain hyperintensities on T2-weighted imaging (T2WI) and fluid-attenuated inversion recovery (FLAIR) sequences in children with neurofibromatosis type 1: inter-rater reliability and clinical correlates

Introduction

Neurofibromatosis type 1 (NF1) is an autosomal dominant neurocutaneous disorder resulting from pathogenic variants in the NF1 gene, which maps to chromosome 17q11.2. NF1 encodes neurofibromin, a protein that plays a pivotal role in regulating cell proliferation and differentiation. Pathogenic variants in NF1 disrupt neurofibromin function, leading to the manifestation of diverse clinical phenotypes that primarily involve the skin, nervous system, skeleton, and vascular system (1,2). Individuals with NF1 have an increased susceptibility to central nervous system (CNS) neoplasms, with optic pathway gliomas (OPGs) being the most common, followed by non-optic pathway tumors including pilocytic astrocytomas and brainstem gliomas (3,4). Approximately 70% of NF1 patients may also present with other non-neoplastic abnormalities. In pediatric NF1 patients, abnormal brain hyperintensities are frequently observed on T2-weighted imaging (T2WI) and fluid-attenuated inversion recovery (FLAIR) sequences—lesions previously termed “unidentified bright objects (UBOs)” or “focal areas of signal intensity (FASI)” in the literature—these hyperintensities are non-neoplastic and are instead linked to NF1-induced microstructural abnormalities in the CNS (5,6). Previous studies with small sample sizes have investigated the anatomical distribution, number, maximum diameter, and volume of brain hyperintensities, as well as the associations between these magnetic resonance imaging (MRI)-derived metrics and clinical characteristics, yielding inconsistent findings (7-9). This variability is likely attributable to differences in segmentation methodologies and the varying sensitivity of MRI sequences—specifically T2WI and FLAIR. In addition, prior discussions on the correlation between hyperintensity volume and clinical manifestations have focused predominantly on cognitive or language function outcomes (7,8). Most notably, no previous research has explored the relationship between hyperintensity volume and key NF1 clinical phenotypes—OPGs and plexiform neurofibromas (PNFs). To address these gaps, the present study aimed to quantify the burden of brain hyperintensities in a large pediatric NF1 cohort using T2WI and FLAIR sequences, evaluate the inter-reader reliability of manual volumetric segmentation, and investigate the relationship between hyperintensity volume and relevant clinical correlates. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2667/rc).

Methods

Study population

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Institutional Review Boards of The First Affiliated Hospital of USTC (No. 2024-RE-304) and Shanghai Ninth People’s Hospital (No. SH9H-2024-T153-2), and the requirement for individual consent for this retrospective analysis was waived. The study cohort, covering the period from May 2019 to May 2025, consisted of pediatric patients (aged <18 years) diagnosed with NF1 who met the following enrollment criteria: (I) fulfilment of the National Institutes of Health (NIH) revised diagnostic criteria for NF1 (10); (II) complete whole-body magnetic resonance imaging (WB-MRI) data; and (III) comprehensive clinical records. The exclusion criteria were as follows: failure to meet the diagnostic criteria for NF1; incomplete MRI examinations; presence of interfering artifacts (including motion artifacts, susceptibility artifacts, signal loss, and image blurring) that could compromise the accuracy of hyperintensity segmentation and volumetric measurement. Patients with other CNS comorbidities (e.g., demyelinating and inflammatory diseases, toxic and metabolic encephalopathy, and traumatic brain injury) were also excluded to eliminate the confounding effects of these diseases on the detection, quantification, and morphological characterization of hyperintensities. A total of 108 patients were enrolled, comprising 51 males and 57 females, with a mean age of 8.6±4.2 years and an age range of 1.5 to
17.75 years. Clinical data were collected, including neurofibroma subtypes, as well as family history of NF1, café-au-lait macules, axillary or inguinal freckling, and other NF1-related manifestations.

MRI acquisition protocol

All patients underwent MRI using a 3T MRI scanner (MAGNETOM Trio Tim, Vida, or Prisma, Siemens Healthineers, Erlangen, Germany). Routine MRI sequences included T1-weighted imaging (T1WI) and T2WI; additionally, short-tau inversion recovery (STIR) sequences were used for body imaging, Dixon sequences for orbital imaging, FLAIR, and diffusion-weighted imaging (DWI) sequences for head imaging. Detailed imaging parameters are summarized in Table S1. Contrast-enhanced MRI was performed in pediatric patients with suspected low-grade intracranial gliomas (LGGs) or OPGs.

Imaging processing and analysis

T2WI and FLAIR hyperintensities were defined as regions with signal intensity greater than that of normal gray matter, isointensity on T1WI, lack of mass effect, no postcontrast enhancement, and no diffusion restriction on DWI. However, non-optic gliomas often show mild-to-moderate enhancement, mass effect, or restricted diffusion.

We evaluated the anatomical distribution of hyperintensities on brain MRI images, including the basal ganglia region, thalamus, hippocampus, brainstem (midbrain, pons, medulla oblongata), pontine arm, and dentate nucleus of the cerebellum, as well as the boundaries (clear or unclear). All MRI data were evaluated separately by two neuroimaging-oriented physicians, each with over 8 years of experience, who processed and analyzed the data without access to the clinical information, and consensus was reached by negotiation in case of disagreement.

Two readers independently performed manual volumetric segmentation on randomly assigned T2WI and FLAIR sequences using ITK-SNAP software (https://www.itksnap.org/pmwiki/pmwiki.php). After an interval of 2 weeks, they segmented the alternate sequence, with subsequent volumetric calculations (mL). For each patient, the total hyperintensity volume at each anatomical location was documented. T2WI and FLAIR images were acquired consecutively in the same scanning session, ensuring their inherent spatial alignment in the scanner’s native coordinate system; hyperintensity segmentation was thus performed by directly referencing the original images.

Statistical analysis

Quantitative data are presented as mean ± standard deviation, and categorical data as counts (percentages) [n (%)]. Inter-reader reliability was assessed using intraclass correlation coefficient (ICC), using two-way mixed effect model, absolute consistency, and single measurement value calculation. Specifically, ICC values were computed separately for hyperintensity volume data derived from manual volumetric segmentation of T2WI and FLAIR sequences by two readers, aiming to evaluate the consistency and reliability of segmentation results across different sequences. The interpretation criteria were defined as follows: ICC ≥0.90 indicated excellent consistency, 0.75–0.89 indicated good consistency, and <0.75 indicated poor consistency. Additionally, Bland-Altman plots were utilized to analyze the mean difference and 95% limits of agreement (LoA) between the two readers' measurements.

Mann-Whitney U test was used to conduct univariate intergroup comparisons of hyperintensity volume among patients stratified by gender, age, and OPGs/PNFs status, for the preliminary screening of clinically relevant variables potentially associated with cerebral hyperintensity volume. Subsequently, variables with potential relevance (P<0.1 in univariate analysis) were included in a mixed-effects linear regression model, with site and scanner incorporated as random effects to account for inter-center and inter-scanner variability. Age and sex were included as fixed effects for adjustment. After controlling for confounding factors, the independent associations between each factor and cerebral hyperintensity volume were identified and quantified. A two-tailed P value <0.05 was considered statistically significant.

All statistical analyses were performed using the software SPSS 27.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 22.001 (MedCalc Software, Ostend, Belgium). Statistical significance was set at P<0.05.

Results

Clinical and WB-MRI features of participants

Genetic testing identified 27.8% of 108 pediatric patients as familial cases. Clinical physical examination detected axillary/inguinal freckling (51.9%) and cutaneous neurofibromas (CNFs; 56.5%). WB-MRI revealed PNFs (74.1%)—distributed in the head-neck region, abdominopelvic cavity, and paravertebral plexus—as well as 26 cases of OPGs (24.1%), and 17 cases of characteristic bone pathology (15.7%) including 10 cases of sphenoid wing dysplasia, 1 case of pseudoarthrosis of long bones, and 6 cases of scoliosis. Moyamoya disease was observed in
5 cases (4.6%) (Table 1).

Brian hyperintensities: imaging features, preferential locations, and age characteristics

Hyperintensities located in the basal ganglia region typically had clear borders and were round or ovoid; most of them showed apparent high signal intensity on T2WI and FLAIR sequences. Lesions in other cerebral regions had unclear borders and exhibited cloud-like, slightly high signal intensity on T2WI and FLAIR sequences and isointensity on T1WI sequences. The preferred locations of intracerebral hyperintensities were cerebellar dentate nucleus (89.4%), brainstem (76.5%), globus palidus (72.9%), brachium pontis (54.1%), thalamus (50.6%), hippocampus (45.9%), and deep white matter (4.7%) in the order of prevalence (Table 2); those in the deep white matter were unilateral, whereas the rest were more commonly bilateral (Figure 1A-1H). The globus pallidus was the location with the highest average age at detection for hyperintensities (8.95±3.64 years), whereas deep white matter is the location with the lowest average age at detection (5.38±0.06 years) (Table 2). Notably, 90.6% (77/85) of individual patients exhibited hyperintensities involving multiple distinct brain regions.

Figure 1 Hyperintensities identified in the basal ganglia (A,B), thalamus (C,D), brachium pontis (E,F), and cerebellar dentate nucleus (G,H) on T2WI and FLAIR sequences, respectively. FLAIR, fluid-attenuated inversion recovery; T2WI, T2-weighted imaging.

Volume segmentation reliability and consistency

Hyperintensities volume measured by two readers on T2WI and FLAIR images were 7.15±5.95 vs. 7.36±6.19 mL and 7.94±6.27 vs. 8.02±6.57 mL, with an ICC value of 0.993 vs. 0.994, respectively, the corresponding ICCs were 0.993 and 0.994, with all P values <0.001, indicating excellent inter-reader agreement and high reliability of the measurements for both sequences (Table 3). Bland-Altman plot analysis was performed to evaluate inter-reader agreement for intracerebral hyperintensity volume measurements on T2WI and FLAIR sequences. For T2WI, the mean difference between the two readers was –0.21, with 95% LoA of –1.16, 0.81; 95.3% of the data points fell within this range. For FLAIR, the mean difference was –0.15, with 95% LoA of –1.21, 0.14, and 96.5% of the points were within these bounds. These findings confirm excellent inter-reader agreement for hyperintensity volume segmentation on both T2WI and FLAIR sequences (Figure 2).

Figure 2 Consistency analysis of hyperintensity volume measurements by two readers in T2WI and FLAIR sequences: ICC (A,C) and Bland-Altman (B,D) plots, with segmented volumes expressed as logarithmic values [log(mL)]. ICC, intraclass correlation coefficient; FLAIR, fluid-attenuated inversion recovery; SD, standard deviation; T2WI, T2-weighted imaging.

Relationship between hyperintensity volume and clinical correlates

A multiple linear regression model was constructed in the present study, with hyperintensity volume as the dependent variable and age, gender, OPGs status, and pNF status as independent variables. The model was statistically significant overall (F=2.817, P=0.029, adjusted R²=0.132). After adjusting for confounding factors, age [β=–0.049, 95% confidence interval (CI): –0.096 to –0.002, P=0.040] and OPGs status (β=–0.556, 95% CI: –1.024 to –0.088, P=0.020) were identified as independent negative predictors of cerebral hyperintensity volume. In contrast, no significant independent associations were observed between cerebral hyperintensity volume and gender (β=–0.119, 95% CI: –0.504 to 0.266, P=0.540) or pNF status (β=–0.052, 95% CI: –0.516 to 0.412, P=0.824). The model showed no multicollinearity [variance inflation factor (VIF) <2 for all variables], and the residuals followed a normal distribution, indicating reliable model fitting (Table 4).

Discussion

NF1 is an inherited disorder characterized by neurofibromas as the typical phenotype and multisystem involvement. It is associated with high teratogenicity and disability rates, poses considerable therapeutic challenges, and remains a key focus and difficult issue in current medical diagnosis and management. In 2020, the first multicenter treatment collaborative group for NF1 was established in China, and a consensus of experts on the clinical diagnosis and treatment of NF1 was formulated (11). This study cohort was predominantly recruited from the largest specialized treatment center and represents one of the few cohorts evaluated using 3.0T WB-MRI for clinical phenotyping.

In the present study, sporadic cases accounted for 64.8% (70/108), a proportion higher than the previously reported 50% (12). This discrepancy may be associated with potential socioeconomic factors and elevated psychological expectations among parents. WB-MRI identified PNFs in 74.1% of cases, yielding a significantly higher detection rate than prior local MRI for clinical phenotyping (13-15). This superior detectability is particularly pronounced for lesions located in deep trunk regions and perivisceral areas—sites that are inaccessible via clinical physical examination. This finding supports WB-MRI as the preferred imaging modality for clinical phenotyping of NF1, particularly for comprehensively identifying deep-seated or clinically occult PNFs that are critical for treatment planning and prognostic assessment. Of course, the higher prevalence of PNFs in our cohort is also likely attributed to referral bias, given that our center specializes in the management of complex NF1 clinical phenotypes (e.g., symptomatic PNFs requiring clinical intervention). This selection bias may restrict the generalizability of our findings to unselected, population-based NF1 cohorts. A total of 26 cases of OPGs (24.1%) were found in this study, which is slightly higher than the prevalence reported in the literature, which is between 15% and 20% (16). This higher detection rate may be attributable to the routine use of orbital MRI in our clinical practice, which improves the identification of OPGs, and may also be an additional consequence of referral bias.

NF1 can involve the skeletal system (17). In this study, 10 cases of sphenoid wing dysplasia were identified, accounting for 9.3% of the total. Additionally, the most prominent clinical manifestation of this disease is facial and orbital asymmetry. As the disease progresses, it may lead to exophthalmos, incomplete eyelid closure, and even visual impairment (18). NF1 is occasionally seen as a vascular abnormality of the CNS, which mainly involves the anterior circulation, with moyamoya disease being the most common manifestation, followed by vascular malformations and aneurysms (5), and only 5 cases of moyamoya disease were identified, indicating an extremely low detection rate, which may be attributed to the lack of routine vascular imaging examinations, such as cranial magnetic resonance angiography (MRA) in our study’s initial imaging protocol.

The clinical phenotypes in the diagnostic criteria of NF1 are well known. However, we observed that children with NF1 demonstrated T2WI and FLAIR hyperintensities in multiple brain regions, with varying lesion counts and a high detection rate consistent with previous literature. These hyperintensities have been termed “unexplained hyperintensities” in prior studies (7,8). Pathologically, areas characterized by changes in myelin vacuoles correspond to hyperintensities, with no inflammatory reaction in the surrounding tissues and no significant demyelination (19).
We observed that basal ganglia hyperintensities were typically well-demarcated with regular morphology and a signal intensity approximating cerebrospinal fluid. These lesions lacked mass effect, showed no postcontrast enhancement, and exhibited no diffusion restriction on DWI sequences. This constellation of findings may be attributed to more prominent intramedullary edema in the basal ganglia compared with other brain regions. However, if the following manifestations were present—mass effect, contrast enhancement, or perilesional edema—close clinical follow-up was warranted to rule out glioma. Notably, in the present study, basal ganglia lesions had a later age at detection than hyperintensities in other brain regions, with the deep white matter exhibiting the youngest mean age at detection.

We also observed that hyperintensities were predominantly located in the cerebellar dentate nucleus, brainstem, and globus pallidus, followed by the pontine crus, thalamus, and hippocampus. These lesions were more frequently bilateral, whereas they were least common in the periventricular white matter and mostly unilateral—findings that are consistent with previous studies (20,21). This distribution pattern suggested that the formation of cerebral hyperintensities in children with NF1 was not a random process but was closely associated with the regional specificity of NF1-related pathological changes. From the perspective of disease pathogenesis, as an inherited neurocutaneous syndrome, the core pathological alterations of NF1 affected the development of cerebral structures during the embryonic stage and the regulatory mechanisms of neuroglial proliferation. Specifically, midline and para-midline structures such as the cerebellar dentate nucleus, brainstem, and globus pallidus were more sensitive to NF1-related genetic abnormalities during embryonic development and had a dense distribution of neuroglial cells. Consequently, pathological abnormalities were more likely to occur in these regions, ultimately presenting as bilaterally predominant hyperintensities. Furthermore, the rare occurrence of hyperintensities in the deep white matter—predominantly unilaterally—could serve as one of the differential diagnostic criteria to distinguish NF1-associated hyperintensities from those induced by other CNS diseases (e.g., leukodystrophy, vascular lesions).

In the present study, we manually segmented hyperintensity volumes using two conventional imaging sequences (T2WI and FLAIR); our initial findings indicated that the mean segmented volume was higher on FLAIR sequences than on T2WI sequences. This result is explainable by the higher sensitivity of FLAIR sequences for detecting hyperintensities in the white matter (22,23), rendering the segmented volumes relatively more reliable.

In addition, ICC analysis demonstrated a high degree of inter-reader consistency in the reliability of manual segmentation for the same sequence. Meanwhile, Bland-Altman analysis revealed that the measurement differences were within the 95% LoA, which is clinically acceptable. Notably, several recent studies utilizing deep learning algorithms for white matter hyperintensity segmentation have confirmed a strong correlation with manual segmentation, further validating the robustness of quantitative assessment methods in this field (24,25). However, some studies have shown that compared with manual segmentation, the algorithms used (LGA SPM 8, LGA SPM 12, and LPA SPM 12) all underestimate the burden of hyperintensities (26). Although manual segmentation is indeed time-consuming, it exhibits greater reliability in lesion segmentation—an observation that is further corroborated by the findings of the present study.

A multiple linear regression model constructed with age, gender, OPGs status, and PNFs status as independent variables explained only 13.2% of the variance in cerebral hyperintensity volume (adjusted R²=0.132). Although the model was statistically significant (P=0.023) and identified age and OPGs status as independent negative predictors of hyperintensity volume—meaning that increasing age and the presence of OPGs were both associated with a reduction in hyperintensity volume—the low explanatory power indicates that the changes in hyperintensity volume are regulated by multiple factors. The variables included in this study were limited, leaving 86.8% of the variance unexplained. This phenomenon may be related to the following key unincorporated factors: specific NF1 mutation types, other unmeasured clinical phenotypes (e.g., cognitive impairment), other unintegrated imaging biomarkers, as well as environmental factors or genetic modifiers. Future studies will further incorporate these parameters to optimize the model.

The negative association of age suggests that hyperintensity volume tends to decrease with increasing age, which is consistent with previous studies (5,16,27), these studies indicated that cerebral hyperintensities in pediatric patients with NF1 typically emerge around one year of age, increase in number and size, peak at approximately seven years of age, and then spontaneously diminish by 17 years. Pathologically, regions with myelin vacuolar changes correspond to cerebral hyperintensities detected on imaging. Histopathological analysis of these hyperintense areas has identified two key features: first, the absence of obvious inflammatory cell infiltration or accumulation of inflammatory mediators in the adjacent brain tissue, indicating that the formation of such hyperintensities is not driven by an inflammatory response. Second, only localized vacuolar degeneration (not widespread destruction) is present, with no significant demyelination and preserved structural integrity. The progressive reduction of cerebral hyperintensities over time is mainly driven by three core factors: reversibility of pathological changes, CNS maturation, and regressive features of disease natural courses. The most common mechanisms include repair of myelin vacuolar degeneration and improved myelination in childhood (5,27). This finding informs clinical practice: for patients with cerebral hyperintensities, particularly children and adolescents, lesion nature should be determined by integrating imaging and clinical context. Serial dynamic imaging follow-up to monitor lesion evolution is preferred over unnecessary intervention, aligning with the reversibility of most such lesions.

Of particular note, the presence of OPGs demonstrated a significant negative predictive value (β=–0.556, P=0.020), indicating that individuals with OPGs exhibit smaller volumes of intracranial hyperintensities. Although this observation might theoretically be linked to a potential neuroprotective effect, such an interpretation remains highly speculative in the absence of direct clinical or mechanistic evidence (e.g., longitudinal neuropsychological or tissue-level data) in the current study. Previous studies have suggested that OPG-related tissue remodeling, driven by NF1 mutation-mediated regulation of oligodendroglial lineage homeostasis and downstream signaling pathways, may influence myelin development in NF1 (28). Although this provides a plausible biological framework, it only offers indirect support for the observed inverse association. Further research with longitudinal and functional data is therefore needed to clarify whether such a relationship truly reflects a neuroprotective mechanism or represents coincidental pathophysiological variation in NF1.

We suggest several potential directions for future research in this domain. First, prospective longitudinal studies with larger cohorts are warranted to verify the inverse association between OPGs and intracranial hyperintensity volumes and clarify the authenticity of its potential neuroprotective mechanism. Second, population-based studies should be conducted to mitigate referral bias and generalize the findings to the broader NF1 population. Third, mechanistic investigations are required to explore relevant biological pathways (e.g., OPG-related myelin modulation). Fourth, future studies should collect neuropsychological data to clarify the functional implications of hyperintensity volumes and optimize imaging analysis protocols to improve result reproducibility.

Limitations

Our study has some limitations. Firstly, it is a cross-sectional study and lacks longitudinal imaging analysis of NF1 patients with regular MRI follow-ups of each NF1 child to further assess the volume changes of the hyperintensities. Secondly, cranial MRI sequences with reduced slice thickness have the potential to minimize segmentation discrepancies, which in turn facilitates the subsequent use of deep learning algorithms to verify their correlation with expert manual segmentation. In addition, a key limitation of this study is that we did not collect complete electroencephalography, seizure history, and formal neuropsychological assessment data in the current cross-sectional cohort, so we could not explore the correlation between hyperintensity volume and these clinical indicators. Future longitudinal studies integrating multi-modal clinical data are needed to verify these potential associations.

Conclusions

WB-MRI has significantly improved the detection of clinical phenotypes in children with NF1, providing a more comprehensive disease assessment. The present study confirms that T2WI and FLAIR hyperintensities are a prevalent imaging feature in the pediatric NF1 population. Furthermore, we identified age and the presence of OPGs as independent predictors of hyperintensity volume, offering potential imaging biomarkers for monitoring disease progression.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2667/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2667/dss

Funding: This work was supported by The National Natural Science Foundation of China (grant No. 82271991) and Joint Fund for Medical Artificial Intelligence of USTC (grant No. MAI2023C006).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2667/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Institutional Review Boards of The First Affiliated Hospital of USTC (No. 2024-RE-304) and Shanghai Ninth People’s Hospital (No. SH9H-2024-T153-2), and the requirement for individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Wang W, Wei CJ, Cui XW, Li YH, Gu YH, Gu B, Li QF, Wang ZC. Impacts of NF1 Gene Mutations and Genetic Modifiers in Neurofibromatosis Type 1. Front Neurol 2021;12:704639. [Crossref] [PubMed]
2. Zhu B, Zheng T, Wang W, Gu Y, Wei C, Li Q, Wang Z. Genotype-phenotype correlations of neurofibromatosis type 1: a cross-sectional study from a large Chinese cohort. J Neurol 2024;271:1893-900. [Crossref] [PubMed]
3. Costa AA, Gutmann DH. Brain tumors in Neurofibromatosis type 1. Neurooncol Adv 2019;1:vdz040. [PubMed]
4. Angelova-Toshkina D, Decker JA, Traunwieser T, Holzapfel J, Bette S, Huber S, Schimmel M, Vollert K, Bison B, Kröncke T, Bramswig NC, Wieczorek D, Gnekow AK, Frühwald MC, Kuhlen M. Comprehensive neurological evaluation of a cohort of patients with neurofibromatosis type 1 from a single institution. Eur J Paediatr Neurol 2023;43:52-61. [Crossref] [PubMed]
5. Russo C, Russo C, Cascone D, Mazio F, Santoro C, Covelli EM, Cinalli G. Non-Oncological Neuroradiological Manifestations in NF1 and Their Clinical Implications. Cancers (Basel) 2021;13:1831. [Crossref] [PubMed]
6. Di Pietro S, Reali L, Tona E, Belfiore G, Praticò AD, Ruggieri M, David E, Foti PV, Santonocito OG, Basile A, Palmucci S. Magnetic Resonance Imaging of Central Nervous System Manifestations of Type 1 Neurofibromatosis: Pictorial Review and Retrospective Study of Their Frequency in a Cohort of Patients. J Clin Med 2024;13:3311. [Crossref] [PubMed]
7. Di Stasi M, Cocozza S, Buccino S, Paolella C, Di Napoli L, D'Amico A, Melis D, Ugga L, Villano G, Ruocco M, Scala I, Brunetti A, Elefante A. The role of unidentified bright objects in the neurocognitive profile of neurofibromatosis type 1 children: a volumetric MRI analysis. Acta Neurol Belg 2024;124:223-30. [Crossref] [PubMed]
8. Harriott EM, Nguyen TQ, Landman BA, Barquero LA, Cutting LE. Using a semi-automated approach to quantify Unidentified Bright Objects in Neurofibromatosis type 1 and linkages to cognitive and academic outcomes. Magn Reson Imaging 2023;98:17-25. [Crossref] [PubMed]
9. Baudou E, Nemmi F, Biotteau M, Maziero S, Assaiante C, Cignetti F, Vaugoyeau M, Audic F, Peran P, Chaix Y. Are morphological and structural MRI characteristics related to specific cognitive impairments in neurofibromatosis type 1 (NF1) children? Eur J Paediatr Neurol 2020;28:89-100. [Crossref] [PubMed]
10. Legius E, Messiaen L, Wolkenstein P, Pancza P, Avery RA, Berman Y, Blakeley J, Babovic-Vuksanovic D, Cunha KS, Ferner R, Fisher MJ, Friedman JM, Gutmann DH, Kehrer-Sawatzki H, Korf BR, Mautner VF, Peltonen S, Rauen KA, Riccardi V, Schorry E, Stemmer-Rachamimov A, Stevenson DA, Tadini G, Ullrich NJ, Viskochil D, Wimmer K, Yohay KInternational Consensus Group on Neurofibromatosis Diagnostic Criteria (I-NF-DC). Huson SM, Evans DG, Plotkin SR. Revised diagnostic criteria for neurofibromatosis type 1 and Legius syndrome: an international consensus recommendation. Genet Med 2021;23:1506-1513. [Crossref] [PubMed]
11. National Multi-Center Treatment Collaboration Group For Neurofibromatosis Type, National Multi-Center Research Platform For Plastic And Reconstructive Surgery. Expert consensus on diagnosis and management of neurofibromatosis type 1 (2021 edition). Zhongguo Xiu Fu Chong Jian Wai Ke Za Zhi 2021;35:1384-95. [PubMed]
12. Güneş N, Yeşil G, Geyik F, Kasap B, Celkan T, Kebudi R, Tüysüz B. Neurofibromatosis type 1: Expanded variant spectrum with multiplex ligation-dependent probe amplification and genotype-phenotype correlation in 138 Turkish patients. Ann Hum Genet 2021;85:155-65. [Crossref] [PubMed]
13. Kang E, Kim YM, Choi Y, Lee Y, Kim J, Choi IH, Yoo HW, Yoon HM, Lee BH. Whole-body MRI evaluation in neurofibromatosis type 1 patients younger than 3 years old and the genetic contribution to disease progression. Orphanet J Rare Dis 2022;17:24. [Crossref] [PubMed]
14. Ly KI, Merker VL, Cai W, Bredella MA, Muzikansky A, Thalheimer RD, Da JL, Orr CC, Herr HP, Morris ME, Chang CY, Harris GJ, Plotkin SR, Jordan JT. Ten-Year Follow-up of Internal Neurofibroma Growth Behavior in Adult Patients With Neurofibromatosis Type 1 Using Whole-Body MRI. Neurology 2023;100:e661-70. [Crossref] [PubMed]
15. Attia S, Guirguis M, Le LQ, Chhabra A. Association of plexiform and diffuse neurofibromas with malignant peripheral nerve sheath tumor in NF I patients: a whole-body MRI assessment. Skeletal Radiol 2024;53:769-77. [Crossref] [PubMed]
16. Wang MX, Dillman JR, Guccione J, Habiba A, Maher M, Kamel S, Panse PM, Jensen CT, Elsayes KM. Neurofibromatosis from Head to Toe: What the Radiologist Needs to Know. Radiographics 2022;42:1123-44. [Crossref] [PubMed]
17. Liu Y, Zhang Z, Liang M, Liu Y, Zhang N, Xu K. Comprehensive imaging analysis of a patient with neurofibromatosis type 1 combined with hypophosphatemic osteomalacia: a case description. Quant Imaging Med Surg 2023;13:4777-84. [Crossref] [PubMed]
18. Rijken BFM, van Veelen-Vincent MLC, Mathijssen IMJ. Sphenoid dysplasia in patients with neurofibromatosis type 1: Clinical features and imaging findings including cerebrospinal fluid alterations. Eur J Paediatr Neurol 2023;42:28-33. [Crossref] [PubMed]
19. DeBella K, Poskitt K, Szudek J, Friedman JM. Use of "unidentified bright objects" on MRI for diagnosis of neurofibromatosis 1 in children. Neurology 2000;54:1646-51. [Crossref] [PubMed]
20. Ferraz-Filho JR, José da Rocha A, Muniz MP, Souza AS, Goloni-Bertollo EM, Pavarino-Bertelli EC. Unidentified bright objects in neurofibromatosis type 1: conventional MRI in the follow-up and correlation of microstructural lesions on diffusion tensor images. Eur J Paediatr Neurol 2012;16:42-7. [Crossref] [PubMed]
21. Sánchez Marco SB, López Pisón J, Calvo Escribano C, González Viejo I, Miramar Gallart MD, Samper Villagrasa P. Neurological manifestations of neurofibromatosis type 1: our experience. Neurologia (Engl Ed) 2022;37:325-33. [Crossref]
22. Wu X, Ya J, Zhou D, Ding Y, Ji X, Meng R. Pathogeneses and Imaging Features of Cerebral White Matter Lesions of Vascular Origins. Aging Dis 2021;12:2031-51. [Crossref] [PubMed]
23. Ribaldi F, Altomare D, Jovicich J, Ferrari C, Picco A, Pizzini FB, Soricelli A, Mega A, Ferretti A, Drevelegas A, Bosch B, Müller BW, Marra C, Cavaliere C, Bartrés-Faz D, Nobili F, Alessandrini F, Barkhof F, Gros-Dagnac H, Ranjeva JP, Wiltfang J, Kuijer J, Sein J, Hoffmann KT, Roccatagliata L, Parnetti L, Tsolaki M, Constantinidis M, Aiello M, Salvatore M, Montalti M, Caulo M, Didic M, Bargallo N, Blin O, Rossini PM, Schonknecht P, Floridi P, Payoux P, Visser PJ, Bordet R, Lopes R, Tarducci R, Bombois S, Hensch T, Fiedler U, Richardson JC, Frisoni GB, Marizzoni M. Accuracy and reproducibility of automated white matter hyperintensities segmentation with lesion segmentation tool: A European multi-site 3T study. Magn Reson Imaging 2021;76:108-115. [Crossref] [PubMed]
24. Diaz-Hurtado M, Martínez-Heras E, Solana E, Casas-Roma J, Llufriu S, Kanber B, Prados F. Recent advances in the longitudinal segmentation of multiple sclerosis lesions on magnetic resonance imaging: a review. Neuroradiology 2022;64:2103-17. [Crossref] [PubMed]
25. Egger C, Opfer R, Wang C, Kepp T, Sormani MP, Spies L, Barnett M, Schippling S. MRI FLAIR lesion segmentation in multiple sclerosis: Does automated segmentation hold up with manual annotation? Neuroimage Clin 2016;13:264-70. [Crossref] [PubMed]
26. Schmidt P, Pongratz V, Küster P, Meier D, Wuerfel J, Lukas C, Bellenberg B, Zipp F, Groppa S, Sämann PG, Weber F, Gaser C, Franke T, Bussas M, Kirschke J, Zimmer C, Hemmer B, Mühlau M. Automated segmentation of changes in FLAIR-hyperintense white matter lesions in multiple sclerosis on serial magnetic resonance imaging. Neuroimage Clin 2019;23:101849. [Crossref] [PubMed]
27. Calvez S, Levy R, Calvez R, Roux CJ, Grévent D, Purcell Y, Beccaria K, Blauwblomme T, Grill J, Dufour C, Bourdeaut F, Doz F, Robert MP, Boddaert N, Dangouloff-Ros V. Focal Areas of High Signal Intensity in Children with Neurofibromatosis Type 1: Expected Evolution on MRI. AJNR Am J Neuroradiol 2020;41:1733-39. [Crossref] [PubMed]
28. Aghoghovwia BE, Shen CE, Bano S, Shyamala N, Echeandia Marrero A, Irshad K, Sharafaldin S, Brossier NM, Pan Y. Current states in understanding oligodendroglia-mediated neurological issues in neurofibromatosis type 1 (NF1). Acta Neuropathol Commun 2025;13:202. [Crossref] [PubMed]