PURPOSE: To develop a three-dimensional (3D) segmentation and computerized volumetry technique for use in the assessment of neurofibromatosis and to assess the ability of this technique to aid in the calculation of tumor burden in patients with neurofibromatosis types 1 and 2 (NF1 and NF2, respectively) and schwannomatosis detected with whole-body magnetic resonance (MR) imaging. MATERIALS AND METHODS: Institutional review board approval and written informed consent were obtained for this prospective HIPAA-compliant study. Fifty-two subjects (27 women, 25 men; mean age, 42 years +/- 15 [standard deviation]; age range, 24-86 years) underwent whole-body MR imaging performed with coronal short inversion time inversion-recovery (STIR) sequences. Whole-body tumor burden was estimated with a 3D segmentation method (the dynamic-threshold [DT] level set method) in 29 subjects (16 with NF1, six with NF2, and seven with schwannomatosis) in whom at least one nerve sheath tumor was reliably identified on MR images. Fifty tumors (25 plexiform and 25 discrete tumors) were randomly selected and subjected to manual and computerized volumetry to assess reliability. Ten plexiform tumors 5 cm or larger in diameter were retrospectively selected and segmented with three initialization methods for computerized volumetry and manually contoured by three radiologists to assess repeatability. Bland-Altman analysis was performed, and intraclass correlation coefficients (ICCs) were calculated. RESULTS: A total of 398 nerve sheath tumors (185 plexiform and 213 discrete tumors) were identified in 29 subjects. Volumetric measurements obtained with the computerized method and manual contouring were highly correlated (r(ICC) = 0.99). Bland-Altman analysis showed that computerized volumetry had a mean difference of -2.6% compared with manual volumetry. The repeatability coefficient of the computerized scheme was +/-5% compared with +/-30% for manual contouring. CONCLUSION: This 3D segmentation and computerized volumetry technique is reliable relative to manual segmentation and has the advantage of being less labor intensive and more repeatable. This technique can be paired with whole-body MR imaging to determine tumor burden in patients with neurofibromatosis. Supplemental material: http://radiology.rsnajnls.org/cgi/content/full/250/3/665/DC1
PURPOSE: To develop a three-dimensional (3D) segmentation and computerized volumetry technique for use in the assessment of neurofibromatosis and to assess the ability of this technique to aid in the calculation of tumor burden in patients with neurofibromatosis types 1 and 2 (NF1 and NF2, respectively) and schwannomatosis detected with whole-body magnetic resonance (MR) imaging. MATERIALS AND METHODS: Institutional review board approval and written informed consent were obtained for this prospective HIPAA-compliant study. Fifty-two subjects (27 women, 25 men; mean age, 42 years +/- 15 [standard deviation]; age range, 24-86 years) underwent whole-body MR imaging performed with coronal short inversion time inversion-recovery (STIR) sequences. Whole-body tumor burden was estimated with a 3D segmentation method (the dynamic-threshold [DT] level set method) in 29 subjects (16 with NF1, six with NF2, and seven with schwannomatosis) in whom at least one nerve sheath tumor was reliably identified on MR images. Fifty tumors (25 plexiform and 25 discrete tumors) were randomly selected and subjected to manual and computerized volumetry to assess reliability. Ten plexiform tumors 5 cm or larger in diameter were retrospectively selected and segmented with three initialization methods for computerized volumetry and manually contoured by three radiologists to assess repeatability. Bland-Altman analysis was performed, and intraclass correlation coefficients (ICCs) were calculated. RESULTS: A total of 398 nerve sheath tumors (185 plexiform and 213 discrete tumors) were identified in 29 subjects. Volumetric measurements obtained with the computerized method and manual contouring were highly correlated (r(ICC) = 0.99). Bland-Altman analysis showed that computerized volumetry had a mean difference of -2.6% compared with manual volumetry. The repeatability coefficient of the computerized scheme was +/-5% compared with +/-30% for manual contouring. CONCLUSION: This 3D segmentation and computerized volumetry technique is reliable relative to manual segmentation and has the advantage of being less labor intensive and more repeatable. This technique can be paired with whole-body MR imaging to determine tumor burden in patients with neurofibromatosis. Supplemental material: http://radiology.rsnajnls.org/cgi/content/full/250/3/665/DC1