CLINICAL ARTICLE - PEDIATRICS

Tumor burden evaluation in NF1 patients with plexiformneurofibromas in daily clinical practice

L. Pratt & D. Helfer & L. Weizman & B. Shofty &

S. Constantini & L. Joskowicz & D. Ben Bashat & L. Ben-Sira

Received: 12 November 2014 /Accepted: 29 January 2015 /Published online: 14 March 2015# Springer-Verlag Wien 2015

AbstractBackground Existing volumetric measurements of plexiformneurofibromas (PNs) are time consuming and error prone, asthey require delineation of PN boundaries, a procedure that isnot practical in the typical clinical setting. The aim of this studyis to assess the Plexiform Neurofibroma Instant SegmentationTool (PNist), a novel semi-automated segmentation programthat we developed for PN delineation in a clinical context. PNistwas designed to greatly simplify volumetric assessment of PNsthrough use of an intuitive user interface while providing ob-jectively consistent results with minimal interobserver andintraobserver variabilities in reasonable time.

Materials and methods PNs were measured in 30 magneticresonance imaging (MRI) scans from 12 patients with neuro-fibromatosis 1. Volumetric measurements were performedusing PNist and compared to a standard semi-automated vol-umetric method (Analyze 9.0).Results High correlation was detected between PNist and thesemi-automated method (R2=0.996), with a mean volumeoverlap error of 9.54% and low intraobserver and interobserv-er variabilities. The segmentation time required for PNist was60 % of the time required for Analyze 9.0 (360 versus 900 s,respectively). PNist was also reliable when assessing changesin tumor size over time, compared to the existing commercialmethod.Conclusions Our study suggests that the new PNist methodis accurate, intuitive, and less time consuming for PNsegmentation compared to existing commercial volumet-ric methods. The workflow is simple and user-friendly,making it an important clinical tool to be used by radi-ologists, neurologists and neurosurgeons on a daily ba-sis, helping them deal with the complex task of evalu-ating PN burden and progression.

Keywords Plexiform neurofibroma . Neurofibromatosis 1 .

Tumor burden . Volumetry

Introduction

Plexiform neurofibromas (PNs) are one of the primary fea-tures of neurofibromatosis 1 (NF1) [1, 2]. The reported inci-dence ranges between 30 and 40 % of NF1 patients. This isprobably an underestimation, as internal PNs are often unde-tected without appropriate imaging [1–3]. These tumors havea significant size range, but typically are large and extensive,

Presentation at a conference:L. Pratt, D. Helfer, L. Weizman, B. Shofty, S. Constantini, L. Joskowicz,D. Ben Bashat, L. Ben-Sira. "PNist”: a novel semi-automated volumetricmethod for easy segmentation of plexiform neurofibromas—a practicaltool for the clinicians. Poster presentation at EANS 2013, 11–14 Novem-ber 2013, Tel-Aviv, Israel

L. Pratt (*) : L. Ben-SiraImaging Division, Tel Aviv Sourasky Medical Center, 6 WeizmannStreet, Tel Aviv 64239, Israele-mail: litalpratt@gmail.com

D. Helfer : L. Weizman : L. JoskowiczSchool of Engineering and Computer Science, The HebrewUniversity of Jerusalem, Jerusalem, Israel

B. Shofty : S. ConstantiniDepartment of Pediatric Neurosurgery, and the Gilbert InternationalNeurofibromatosis Center, Tel Aviv Sourasky Medical Center, TelAviv, Israel

S. Constantini :D. Ben Bashat : L. Ben-SiraSackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel

D. Ben BashatFunctional Brain Center, Tel Aviv Sourasky Medical Center, TelAviv, Israel

Acta Neurochir (2015) 157:855–861DOI 10.1007/s00701-015-2366-z

with irregular complex shapes. Their growth rate is erratic andunpredictable [4]. They can involve different parts of the bodyand may infiltrate, displace, or compress the surroundingstructures [1, 2]. Thus, in addition to esthetic disfiguration,they may lead to substantial morbidity [3–5]. In 10 % of thecases these lesions transform into malignant peripheral nervesheath tumors (MPNSTs) [2, 3]. This serious complication isassociated with higher tumor burden and rapid tumor progres-sion. Therefore, tumor burden assessment is essential for de-tecting aggressive lesions at an early stage and monitoringresponse to therapy [6, 7].

Magnetic resonance imaging (MRI) is the imaging modal-ity of choice to detect and characterize PNs, which appearbright in relation to adjacent normal tissues when using fat-suppression techniques such as STIR (short-T1 inversionrecovery) [8–10]. Currently, evaluation of tumor progressionand treatment decisions are based on tumor volume assess-ment on MRI scans, through a process which is both timeconsuming and error prone. Delineation of tumor boundariesis often difficult due to tumor inhomogeneity, blurred tumormargins, and signal intensity overlapping with surroundingtissues [5]. In addition, there are technical difficulties relatedto the quality and consistency of the MRI images, such asinhomogeneity of the magnetic field. Thus, traditional unidi-mensional and bidimensional measurements are not suitablefor efficient, reliable PN volume assessment.

Several semi-automatic segmentation volumetric methodshave previously been described in the literature, aiming main-ly to address the difficulties in PN volume measurement [4, 5,11]. These methods require intensive and time consuming userinteraction during the entire segmentation process for multiplenoncontinuous or complex lesions, and have been describedmainly as research tools.

In previous publications, we presented semi-automaticmethods for PN segmentation [12, 13]. PNist (PN Instant Seg-mentation Tool) [13] is an advanced method that enables easyvolumetric quantification of PNs on MRI scans with minimaluser intervention. It is designed to be user-friendly and com-patible with the busy daily routine of a clinician. PNist relieson simple, easy-to-learn, user interaction to allow accurate PNdelineation regardless of size and complexity.

This paper describes the validation study of PNist by com-paring it to standard-of-care manual measurements, and itsability to evaluate tumor progression, estimating tumor burdenvolume changes on consecutive scans.

Materials and methods

Tel-AvivMedical Center Institutional ReviewBoard (HelsinkiCommittee) approved this study. The data collection andvolumetric measurements have been performed between theyears 2011 to 2012.

Study population

MRI scans of 12 NF1 patients with PNs were retrospectivelyobtained. All of these patients are routinely followed by theGilbert Israeli Neurofibromatosis Center (GINFC). Patients’ages ranged from 4 to 24 years.

A total of 30 MRI scans were included in the study. Tenpatients had two or three MRI studies in different time inter-vals that included STIR sequences; two patients had only oneMRI study each (Table 1).

MRI methodology

MRI scans were acquired by a 1.5-T MR System (Signa Ex-cite HDx; GE, Milwaukee, WI, USA), at Tel Aviv SouraskyMedical Center. Images were acquired in coronal or axialplanes, and included STIR sequences. The number of slicesin each sequence varied between 14 and 48. Voxel sizes variedbetween 0.4 × 0.4 × 3.3 mm3 and 1.9 × 1.9 × 9.0 mm3. Thescans showed lesions in various locations: scalp, neck, shoul-der, spine, abdomen, pelvis, and calves. All scans were per-formed between the years 2006 to 2012.

Tumor classification

Tumors were classified by radiologist evaluation into threecategories, based on tumor complexity and morphologicalpatterns:

1. Simple tumor: Lesion with sharp boundaries, without ex-tension or infiltration to adjacent tissues.

2. Intermediate tumor: Lesion with sharp boundaries, withextension or infiltration to adjacent tissues.

Table 1 Study cohort: demographics and tumor location

Patient Gender No. of MRI studies Tumor location Type

1 M 2 Abdomen Intermediate

2 F 3 Scalp Simple

3 F 2 Abdomen Intermediate

4 F 3 Scalp Simple

5 F 2 Calves Simple

5 F 2 Pelvis Intermediate

5 F 1 Total body Intermediate

6 F 3 Scalp Intermediate

7 M 2 Shoulder Intermediate

8 M 3 Neck Complex

9 F 3 Neck Complex

10 F 2 Neck Complex

11 F 1 Neck Complex

12 M 1 Abdomen Complex

856 Acta Neurochir (2015) 157:855–861

3. Complex tumor: Lesion with blurred margins, in which itwas difficult to distinguish between tumor and surround-ing tissues, with extension or infiltration to adjacenttissues.

PNist overview

We used PNist for segmentation of the tumors (for a de-tailed technical description of the method, see [13]). PNistrelies on predefined tumor models obtained from a train-ing phase. The input to the training phase is a set of STIRMR scans and manual delineation of PNs in those scans.Scans and delineations are used to create models of ex-pected intensity distribution in the vicinity of tumors,which are then stored in a histogram database. Informa-tion from the database, together with a user scribble on anew scan is then fed into the interactive phase of themethod to segment PNs in the new STIR MRI scans.Figure 1 illustrates the flow of the algorithm.

The segmented area from one slice can be copied automat-ically to the adjacent slice and used as input there, effectivelyworking in three-dimensions. The interactive phase is repeat-ed as many times as required, with an option of manual cor-rection, until the user is satisfied that all tumor voxels werelabeled.

A simple to use graphical user interface (GUI) is used todelineate the tumors (Fig. 2). In addition to basic draw anderase operations, it has a Smart Draw operation: instantlysegmenting bright objects in the vicinity of the first simplemouse stroke. For illustration clip please refer to http://youtu.be/XNcIl7_l29E.

Study methodology

To evaluate the new PNist segmentation method, an expertradiologist (LP) segmented the PNs on STIR sequences ineach MRI study from the data set. A total of 90 independentsegmentation sessions were completed (three segmentationsfor each of the 30 MRI scans). All 30 scans were then seg-mented using a commercial general-purpose volumetric seg-mentation tool (Analyze 9.0). The measurements were per-formed on different days for each scan. The interaction timerequired for 21 PNist segmentation sessions was recorded andcompared to the interaction time required for Analyze 9.0segmentation. To evaluate inter-observer variability, a secondexpert radiologist (L.B.S.) segmented ten scans from the dataset, chosen arbitrarily.

Results

Correlation between PNist and analyze 9.0

The volumes measured for all PNs, using both methods,ranged from 5.1 to 611 cm3. There was high correlation(R2=0.996) between the volumetric measurements donewith PNist compared to Analyze 9.0 (Fig. 3). The cor-relation was lower (R2=0.887) for smaller tumors (lessthan 50 cm3).

Volume overlap error

Overlap is a measure of similarity between segmentations,regarding their location in the space. The volume overlap dif-ference (volume overlap error [VOE]) between the sets

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

0 10 20 30 40 500

0.05

0.1

(a) (b) (c) (d)

Fig. 1 Stages of the segmentation algorithm. a Learning phase: Imageintensity histograms of previously segmented PN tumor environments arenormalized and clustered into 15 groups, creating our database ofsamples. This initialization process is performed only once. b Userinput: Given a new coronal STIR MRI of the abdomen, user scribblesover the tumor to sample the local intensity distribution. The histogram ofthe pixels under the thick red line is computed and matched against thesamples database. c Thresholding: Using the most similar histogram in

the samples database, the algorithm automatically finds a threshold greylevel that distinguishes the tumor from its background. The thresholdingmask acknowledges the fact that tumors are not the only structuresbrighter than the selected threshold. d Component filtering: Brightpixels are classified as tumors only if they belong to a connectedcomponent that intersects with the original user input (in this case, thered line in b)

Acta Neurochir (2015) 157:855–861 857

measured with both algorithms was calculated using the fol-lowing formula:

1−2⋅ A∩Pj jA���þ ���P��� ���

where |·| denotes set size, A is the Analyze 9.0 segmentedvolume, and P is the PNist segmented volume.

The mean VOE was computed for the 30 scans that weresegmented three times, for every pair of segmentations (threesegmentations madewith PNist were paired with the matchingsingle segmentation made with Analyze 9.0, yielding threedifferent pairs). The simple tumor group had a lower averageoverlap error value (6 %) than the intermediate and complextumor groups (11.9 and 10.2 % respectively). The overallaverage volume overlap error was 9.54 %.

Segmentation time

Twenty-one segmentations were performedwhile documentingthework-flow, includingmeasuring the duration of the segmen-tation process. Overall, the segmentation time measured with

PNist was shorter when compared to Analyze 9.0. The averageuser interaction time for the 21 measurements made with PNistwas 360 s, compared to 900 s with Analyze 9.0. This consti-tutes an overall reduction of 60 % with PNist compared toAnalyze 9.0. Table 2 shows that the user interaction time wasshorter for all subgroups (simple, intermediate, and complex).

Intraobserver variability

The intraobserver variability of the volume being measured (co-efficient of variation, CV=standard deviation/mean), calculatedfor the 90 segmentations generated from the 30 MRI scans,ranged from 0.17 to 6.28 % (mean 3 %, STD 1.8 %). Fig. 3illustrates the proximity of the maximal, minimal, and averagevolumetric values measured with PNist for each MRI study.

Interobserver variability

There was high correlation between measurements of the vol-ume made by the two radiologists (R2=0.997). The interob-server variability ranged from 0.8 to 11.4 % (mean 4.25 %,STD 6.4 %).

Fig. 2 PNist segmentation process graphical user interface. a Scan loading: Coronal STIR MRI of the shoulder. b Manual tracing of the tumor (userinput). c Segmentation created immediately (red coloring), including the hyperintense components of the tumor

Fig. 3 Correlation between PNistand Analyze 9.0. We analyzed thecorrelation between volumetricvalues measured with Analyze9.0 and the maximal, minimal,and average volumetric valuesmeasured with PNist (PNmax,PNmin, and PN Average,respectively) for each study

858 Acta Neurochir (2015) 157:855–861

Comparison of tumor volume over follow-up time

To evaluate the accuracy of PNist when assessing tumor vol-ume changes on follow-up studies (the “Delta” between se-quential measurements), we defined a Delta Comparison (DC)quantifier:

Delta Comparsion ¼ PNist△Analyze△

¼P1

.P0

A1

.A0

Where P0 and A0 denote a previous volumemeasurement withPNist and Analyze 9.0 respectively, and P1 and A1 denote theconsecutive volumemeasurement with PNist and Analyze 9.0respectively. Ideally, DC should be as close to 1 as possible,indicating that both tools measure volume changes identically.

Ten study patients had two or three MRI scans of the samePN. We computed the DC for every pair of consecutiveexams, a total of 15 pairs. Figure 4 illustrates the high accu-racy of the segmentation tools, with a mean DC=0.983 and astandard deviation of 8.7 %.

Systematic and/or random errors that might exist in the ab-solutemeasurements might decrease or increase the accuracy ofthe measured tumor volume change. To evaluate this scenario,we computed the following absolute and relative values:

Abs Error ¼ maxP1

A1−1

��������; P0

A0−1

��������

� �� 100%

and

Relative Error ¼ PNist△Analyze△

−1� �

� 100% ¼P1

.P0

A1

.A0

������������−1

24

35� 100%

Abs Error represents the maximal mismatch between the seg-mentation methods in the absolute measurement values of thesame tumor on consecutive studies. Relative Error representsthe delta mismatch between the tools.

Figure 5 illustrates the relationship between absolute andrelative errors. The points below the diagonal slope representmeasurements in which the delta mismatch is smaller com-pared to the individual absolute volume values obtained withthe two algorithms.

Note the two measurements in the bottom right corner ofthe graph: both have an absolute volume measurement mis-match between the tools that exceeded 25 %. However, thedelta mismatch for these measurements was smaller than 3 %.This indicates a systematic mismatch between the tools, whichcanceled out in the relative delta measurements. For somemeasurements the volume change mismatch is larger thanthe individual values mismatch (points above the diagonalline), however, they are still much lower than the worst caseline, which denotes the worst case summation of the errors.

Discussion

In this study we propose a novel tool that allows for easy, fast,tumor burden assessment of PNs. Using this tool, the treatingphysician and radiologist are able to rapidly assess tumor bur-den and progression in their office setting, with minimal inter-ruption to their daily workflow.

The simplicity, ease of use and accuracy of PNist makes itunique among other volumetric methods in this field. Previ-ously described tools usually rely on delineation of a region ofinterest, which requires more time and a steady hand, orinvolves seed initiation, which may be tiresome when manysmall non-continuous masses are involved. The input to PNistis a free-hand brush over the object or objects of interest,performed in less than a second and requires less mental ex-ertion and accuracy than other types of input.

Few articles have been published regarding volumetricmethods designed specifically for accurate PN measurements.

Table 2 Average segmentation time (in seconds) according to lesiontype

Tumor type PNist time (seconds±SD) Analyze time (seconds±SD)

Simple 125 (±53) 240 (±66)

Intermediate 273 (±179) 394 (±162)

Complex 840 (±364) 1,350 (±642)

Fig. 4 Delta comparison between PNist and Analyze 9.0 Fig. 5 Relative versus absolute error measurements

Acta Neurochir (2015) 157:855–861 859

Solomon et al. [4] described a segmentation algorithm basedon histogram analysis, in which the analyst finds a thresholdvalue that distinguishes between intensity levels of normaltissue and lesions. Solomon’s two-dimensional volumetricprogram requires the user to define a region of interest in eachMRI slice containing the tumor. The program then automati-cally segments all tumors within the region of interest. Thismethod was proven reproducible with coefficient of variationbetween 0.6 and 5.6 %, and correlated with manual tumortracing (R=0.999). However, the algorithm eliminates con-nected components smaller than a predetermined size,resulting in an inability to detect small tumor lesions. Thismay result in overlooking active disease foci or missing sig-nificant data when assessing treatment response.

Cai et al. [11] measured tumor burden in patients withneurofibromatosis 1 and 2 and schwanomatosis throughwhole-body MRI, using coronal STIR scans. They estimatedtumor burden with a three-dimensional segmentation algo-rithm called the “dynamic–threshold level set” in which theuser has to identify the center of each PN. Their method thenexpands automatically from the user’s predefined central pointto segment the entire lesion in three dimensions, according tostatistical measures. Cai et al.’s method was reliable whencompared to manual segmentation (ricc=0.99), less labor in-tensive, and more repeatable. Nevertheless, these methods arebased on prolonged user-interaction and do not fit well intomost radiologists’ daily clinical work schedule.

Our results show that the PNist algorithm reliably definesand measures PN lesions when compared to an existing com-mercial method (Analyze 9.0), with high repeatability andreproducibility. Comparable results for interobserver variabil-ity [4, 5] and intraobserver variability [4] have been describedin previous publications. PNist was found to be reliable forabsolute volume values, as well as for three-dimensionalshape of created segmentations as measured by the VOE.The mean VOE between PNist and the ground truth was9.54 % in the present study. In our earlier work [12], the meanVOE between a semi-automated PN segmentation algorithmand the ground truth was 27 %. No other published articlesregarding proposed segmentation algorithms provided infor-mation regarding overlap quality in PN segmentation.

Another advantage of PNist is the short computationtime—a clear improvement in the required segmentation timecompared to Analyze 9.0. There are no previous data reportedregarding the required segmentation time with the aforemen-tioned alternative volumetric methods, but we estimate thatour method is less time consuming, especially for measuringcomplex tumor subgroups or lesions that contain many smallunconnected masses.

There are a few limitations to PNist. Although we found ahigh correlation between measurements obtained with PNistin comparison to those obtained with Analyze, the correlationwas slightly lower when comparing only small volume tumors

of less than 50 cm3. A similar trend was reported by Solomonet al. [4]. This finding is probably related to the fact that smallvolume tumors are prone to large relative volume differencesin consecutive segmentations. For example, an absolute vol-ume difference of 1 cm3 between two measurements, which ishardly detectable by the user, yields a relative volume error of25%when the tumor’s overall volume is 4 cm3. Nevertheless,previous works reported that PN size and location are the mostimportant factors in predicting clinical outcome and chancesfor malignant transformation. Thus, it is most important tomonitor large-sized tumors and internal PNs that are difficultto assess with physical examination.

Another limitation common to all volumetric techniquesrelying on diffusion of the created segmentation, has previ-ously described by Cai et al. [11], and concerns diffusion ofthe segmentation to high-signal-intensity structures in the vi-cinity of the segmented lesion. This happens whennontumoral structures (such as cerebrospinal fluid in the spi-nal canal or urinary bladder), are brighter than the algorithmthreshold and the structure’s connected component intersectswith the algorithm input. All of these bright structures willthen be mistaken for tumor. However, this can be very easilyand efficiently corrected in PNist by minimal user interven-tion. We also have developed a predelineation option for highsignal structures, thereby excluding them from segmentationeven before starting the measurement process.

One of the more attractive PNist capabilities is its ability tofollow volumetric PN changes on consecutive MRI scans. Ra-diologists must detect tumor volume changes to predict clinicalcomplications and evaluate the possibility of malignant trans-formation. An efficient tool is required to determine if a lesionhas been stable or is changing in size over time. Therefore, inaddition to evaluating PNist accuracy in measuring absolutevolumes for PNs, we also assessed the method’s ability to trackrelative volume changes over time.We demonstrated that PNistreliably assesses relative volume changes compared to Analyze9.0. Furthermore, even when there was a large mismatch be-tween the absolute volumetric values measured with both toolsfor the same lesion at a specific point of time, the relative sizechange of the lesion on consecutive scans was similar (withdifferences of less than 13 %) between the two tools.

There are not enough data in the literature regarding thenatural history of PNs, probably because of the complexity ofthese tumors, their frequent internal location, and the cumber-some volumetric methods currently in use, which are not suit-able for daily routine assessment. Most previous studiesassessed growth characteristics of PNs on regional MRI scansof known or suspected lesions [3, 14]. Whole-body tumorburden and asymptomatic tumors cannot be evaluated ade-quately with regional techniques, since PNs can appear indifferent locations in the body. In addition, it has been foundthat different PNs have different growth rates in the samepatient. Thus, a single lesion cannot represent the growth rate

860 Acta Neurochir (2015) 157:855–861

of the other lesions, and separate follow-up is needed for eachlesion [3].

Several studies have shown that whole-body MRI(WBMRI) is an efficient tool for evaluating PN tumor burdenin NF1 patients [11, 15, 16]. This technique enables detectionand characterization of PN size and morphology throughoutthe body, and can be used for both baseline assessment andfuture follow-up. Plotkin et al. [17] showed that WBMRI de-tects internal tumors in a far higher percentage of patientscompared with regional imaging methods, highlighting theimportance of using this technique for NF1 patients.They found that 60 % of NF1 patients had internalnerve sheath tumors identified by WBMRI, compared with16–39 % found using regional imaging techniques. Further-more, they characterized the distribution of lesions to distinctanatomic subgroups.

To date, only one published work addresses natural PNgrowth dynamic using volumetric estimation of tumor burdenon WBMRI scans of NF1 patients. In this study, NF1 patientswith serial WBMRI scans were monitored over 1.1–4.9 years.The results showed that PN growth rate is correlated withtumor burden and inversely correlated with age, suggestingthat younger patients, and patients with high tumor burden,warrant close clinical and radiological follow-up. In addition,patients with no internal PNs on the first MRI scan are unlike-ly to develop lesions later on. They also found that somelesions decreased in size during follow-up, but it seemed like-ly that most of these observations represented measurementerror. They concluded that long-term natural history studieswith WBMRI monitoring and clinical assessment are needed.Preliminary results show that PNist can reliably measure thetumor burden in whole-body coronal STIR MRI scans(WBMRI) in a relative short computation time, with minimalstaff intervention required.

Conclusion

We present the first clinical evaluation study of PNist, a novelsemi-automated segmentation tool for volumetric measure-ments of PNs. PNist is a clinical office-based tool designed tobe used by radiologists, neurologists and neurosurgeons on adaily basis, allowing them to deal with the complex task ofevaluating PN burden and progression. PNist was found to beaccurate for absolute volumetric measurement and for definingthe shape of the created segmentation. It has been proven reli-able in assessing PN size changes over-time. The workflow issimple and user-friendly, with a short, interactive computationtime that can easily be incorporated into a busy work schedule.

Conflicts of interest None of the authors has any conflict of interestand/or commercial stake in the evaluated software.

References

1. Mautner VF, Hartmann M, Kluwe L, Friedrich RE, Funsterer C(2006) MRI growth patterns of plexiform neurofibromas in patientwith neurofibromatosis type 1. Neuroradiology 48:160–165

2. Williams VC, Lucas J, Babcock MA, Gutmann DH, Korf B, MariaBL (2009) Neurofibromatosis type 1 revisited. Pediatrics 123:124–133

3. Tucker T, Friedman JM, Friedrich RW, Funsterer C, Mautner VF(2009) Longitudinal study of neurofibromatosis 1 associated plexi-form neurofibromas. J Med Genet 46(2):81–85

4. Solomon J, Warren K, Dombi E, Patronas N, Widemann B (2004)Automated detection and volume measurement of plexiform neuro-fibromas in neurofibromatosis 1 using magnetic resonance imaging.Comput Med Imaging Graph 28(5):257–265

5. Poussaint TY, Jaramillo D, Chang Y, Korf B (2003) Interobserverreproducibility of volumetric MR imaging measurements of plexi-form neurofibromas. AJR Am J Roentgenol 180(2):419–423

6. Nguyen R, Kluwe L, Fuensterer C, Kentsch M, Friedrich RE,Mautner VF (2011) Plexiform neurofibromas in children with neu-rofibromatosis type 1: frequency and associated clinical deficits. JPediatr 159(4):652–655

7. Nguyen R, Dombi E, Widemann BC, Solomon J, Fuensterer C,Kluwe L, Friedman JM, Mautner VF (2012) Growth dynamics ofplexiform neurofibromas: a retrospective cohort study of 201 patientswith neurofibromatosis 1. Orphanet J Rare Dis 7:75

8. Ros PR, Eshaghi N (1991) Plexiform neurofibroma of the pelvis: CTand MRI findings. Magn Reson Imaging 9:463–465

9. Stull MA, Moser RP, Kransdorf MJ, Bogumill GP, Nelson MC(1991) Magnetic resonance appearance of peripheral nerve sheathtumors. Skelet Radiol 20(1):9–14

10. Tien RD, Hesselink JR, Chu PK, Szumowski J (1991) Improveddetection and delineation of head and neck lesions with fat suppres-sion spin-echo MR imaging. AJNR Am J Neuroradiol 12:19–24

11. Cai W, Kassarjian A, Bredella MA, Harris GJ, Yoshida H, MautnerVF, Wenzel R, Plotkin SR (2009) Tumor burden in patients withneurofibromatosis types 1 and 2 and schwannomatosis: determina-tion on whole-body MR images. Radiology 250:665–673

12. Weizman L, Hoch L, Ben-Bashat D, Joskowicz L, Pratt LT,Constantini S, Ben Sira L (2012) Interactive segmentation of plexi-form neurofibroma tissue: method and preliminary performanceevaluation. Med Biol Eng Comput 50(8):877–884

13. Weizman L, Helfer D, Ben Bashat D, Pratt LT, Joskowicz L,Constantini S, Shofty B, Ben Sira L (2014) PNist: interactive volu-metric measurements of plexiform neurofibromas inMRI scans. Int JComput Assist Radiol Surg 9(4):683–693

14. Dombi E, Solomon J, Gillespie AJ, Fox E, Balis FM, Patronas ER,Korf BR, Babovic-Vuksanovic D, Packer RG, Belasco J, Goldman S,Jakacki R, Kieran M, Steinberg M, Weidemann BC (2007) NF1plexiform neurofibroma growth rate by volumetric MRI: relationshipto age and body weight. Neurology 68:643–647

15. Mautner VF, Asuagbor FA, Dombi E, Funsterer C, Kluwe L, WenzelR, Widemann BC, Friedman JM (2008) Assessment of benign tumorburden by whole-body MRI in patients with neurofibromatosis 1.Neuro Oncol 10:593–598

16. Van Meerbeeck SFL, Verstraete KL, Janssens S, Mortier G (2009)Whole body MR imaging in neurofibromatosis type 1. Eur J Radiol69:236–242

17. Plotkin SR, Bredella MA, CaiW, Kassarjian A, Harris GJ, Esparza S,Merker VL, Munn LL, Muzikansky A, Askenazi M, Nguyen R,Wenzel R, Mautner VF (2012) Quantitative assessment of whole-body tumor burden in adult patients with neurofibromatosis. PLoSONE 7(4):e35711

Acta Neurochir (2015) 157:855–861 861