PNist: interactive volumetric measurements of plexiform neurofibromas in MRI scans

Weizman, Lior; Helfer, D.; Bashat, Dafna Ben; Pratt, Li-tal; Joskowicz, Leo; Constantini, Shlomi; Shofty, Ben; Ben-Sira, Liat

doi:10.1007/s11548-013-0961-0

“…When comparing the performance of these two segmentation methods in our study, no significant difference was observed between manual (ITK) and semiautomatic (3D Slicer) segmentation when they were applied to mMRI or DWIs, which is consistent with other studies 22 24 .…”

Section: Discussionsupporting

confidence: 91%

“…Morphological imaging (e.g., CT, MR T1WI or T2WI) and functional imaging (e.g., MR DWI and positive emission tomography, PET) have been used for the delineation of GTV measurements 10 13 18 23 24 . Among these imaging modalities, MRI provides the best soft tissue contrast and spatial resolution compared to other modalities 4 .…”

Section: Discussionmentioning

confidence: 99%

MR diffusion-weighted imaging-based subcutaneous tumour volumetry in a xenografted nude mouse model using 3D Slicer: an accurate and repeatable method

Ma

¹

,

Chen

²

,

Huang

³

et al. 2015

Sci Rep

View full text Add to dashboard Cite

Accurate and repeatable measurement of the gross tumour volume(GTV) of subcutaneous xenografts is crucial in the evaluation of anti-tumour therapy. Formula and image-based manual segmentation methods are commonly used for GTV measurement but are hindered by low accuracy and reproducibility. 3D Slicer is open-source software that provides semiautomatic segmentation for GTV measurements. In our study, subcutaneous GTVs from nude mouse xenografts were measured by semiautomatic segmentation with 3D Slicer based on morphological magnetic resonance imaging(mMRI) or diffusion-weighted imaging(DWI)(b = 0,20,800 s/mm2) . These GTVs were then compared with those obtained via the formula and image-based manual segmentation methods with ITK software using the true tumour volume as the standard reference. The effects of tumour size and shape on GTVs measurements were also investigated. Our results showed that, when compared with the true tumour volume, segmentation for DWI(P = 0.060–0.671) resulted in better accuracy than that mMRI(P < 0.001) and the formula method(P < 0.001). Furthermore, semiautomatic segmentation for DWI(intraclass correlation coefficient, ICC = 0.9999) resulted in higher reliability than manual segmentation(ICC = 0.9996–0.9998). Tumour size and shape had no effects on GTV measurement across all methods. Therefore, DWI-based semiautomatic segmentation, which is accurate and reproducible and also provides biological information, is the optimal GTV measurement method in the assessment of anti-tumour treatments.

show abstract

“…The time necessary for a trained imaging scientist to process the images and delete normal anatomy took on average 6 min compared to an average of 18 min for manual volumetric analysis. Furthermore, Weizman et al determined that for complex cases, up to 35 min per case would be required for manual segmentation 13 . If one expands this to the typical case load encountered by radiologists at our institution, this amounts to 0.007-0.014 FTE, or $3500-7000/case in annualized cost savings.…”

Section: Discussionmentioning

confidence: 99%

“…This approach, which overcomes the slice limitations of previous investigators, still requires that a trained expert identify each tumor in the image volume. Finally, Weizman et al 13 also uses a STIR histogram matching algorithm which requires a one-time training requiring manual volumetric analysis of a tumor and identifying of a small frame including most of the tumor and a large frame which includes tumor and background. In the segmentation phase, the user is required to scribble over the tumor, which can then propagate in 3D space; however, like all the described methods, manual error correction of STIR hyper-intense normal structures adjacent to tumor is required.…”

Section: Discussionmentioning

confidence: 99%

Image segmentation of plexiform neurofibromas from a deep neural network using multiple b-value diffusion data

Ho

¹

,

Kindler

²

,

Persohn

³

et al. 2020

Sci Rep

View full text Add to dashboard Cite

We assessed the accuracy of semi-automated tumor volume maps of plexiform neurofibroma (PN) generated by a deep neural network, compared to manual segmentation using diffusion weighted imaging (DWI) data. NF1 Patients were recruited from a phase II clinical trial for the treatment of PN. Multiple b-value DWI was imaged over the largest PN. All DWI datasets were registered and intensity normalized prior to segmentation with a multi-spectral neural network classifier (MSNN). Manual volumes of PN were performed on 3D-T2 images registered to diffusion images and compared to MSNN volumes with the Sørensen-Dice coefficient. Intravoxel incoherent motion (IVIM) parameters were calculated from resulting volumes. 35 MRI scans were included from 14 subjects. Sørensen-Dice coefficient between the semi-automated and manual segmentation was 0.77 ± 0.016. Perfusion fraction (f) was significantly higher for tumor versus normal tissue (0.47 ± 0.42 vs. 0.30 ± 0.22, p = 0.02), similarly, true diffusion (D) was significantly higher for PN tumor versus normal (0.0018 ± 0.0003 vs. 0.0012 ± 0.0002, p < 0.0001). By contrast, the pseudodiffusion coefficient (D*) was significantly lower for PN tumor versus normal (0.024 ± 0.01 vs. 0.031 ± 0.005, p < 0.0001). Volumes generated by a neural network from multiple diffusion data on PNs demonstrated good correlation with manual volumes. IVIM analysis of multiple b-value diffusion data demonstrates significant differences between PN and normal tissue.

show abstract

“…This approach, which overcomes the slice limitations of previous investigators, still requires that a trained expert identify each tumor in the image volume. Finally, Weizman et al 13 also uses a STIR histogram matching algorithm which requires a one-time training requiring manual volumetric analysis of a tumor and identifying of a small frame including most of the tumor and a large frame which includes tumor www.nature.com/scientificreports/ and background. In the segmentation phase, the user is required to scribble over the tumor, which can then propagate in 3D space; however, like all the described methods, manual error correction of STIR hyper-intense normal structures adjacent to tumor is required.…”

Section: Discussionmentioning

confidence: 99%

Image Segmentation of Plexiform Neurofibromas from a Deep Neural Network Using Multiple b-value Diffusion Data.

Ho

¹

,

Kindler

²

,

Persohn

³

et al. 2020

Preprint

0

View full text Add to dashboard Cite

We assessed the accuracy of semi-automated tumor volume maps of plexiform neurofibroma (PN) generated by a deep neural network, compared to manual segmentation using diffusion weighted imaging (DWI) data. NF1 Patients were recruited from a phase II clinical trial for the treatment of PN. Multiple b-value DWI was imaged over the largest PN. All DWI datasets were registered and intensity normalized prior to segmentation with a multi-spectral neural network classifier (MSNN). Manual volumes of PN were performed on 3D-T2 images registered to diffusion images and compared to MSNN volumes with the Sørensen-Dice coefficient. Intravoxel incoherent motion (IVIM) parameters were calculated from resulting volumes. 35 MRI scans were included from 14 subjects. Sørensen-Dice coefficient between the semi-automated and manual segmentation was 0.77 ± 0.016. Perfusion fraction (f) was significantly higher for tumor versus normal tissue (0.47 ± 0.42 vs. 0.30 ± 0.22, p = 0.02), similarly, true diffusion (D) was significantly higher for PN tumor versus normal (0.0018 ± 0.0003 vs. 0.0012 ± 0.0002, p < 0.0001). By contrast, the pseudodiffusion coefficient (D*) was significantly lower for PN tumor versus normal (0.024 ± 0.01 vs. 0.031 ± 0.005, p < 0.0001). Volumes generated by a neural network from multiple diffusion data on PNs demonstrated good correlation with manual volumes. IVIM analysis of multiple b-value diffusion data demonstrates significant differences between PN and normal tissue. Abbreviations PN Plexiform neurofibroma NF1Neurofibromatosis type 1 MSNN Multispectral neural network IVIM Intravoxel incoherent motion Neurofibromatosis type 1 (NF1) is a common genetic neurocutaneous disease, affecting 1 in 3000 newborn infants 1 . There are over 1000 identified mutations of the NF1 gene which produces a tumor suppressor protein called neurofibromin 2 . A prototypical NF1 tumor is the plexiform neurofibroma (PN), a neurofibroma variant with tumor cells that spread along multiple nerve fascicles, resulting in an extensive mass of thickened nerve bundles in a proteinaceous matrix. These PNs affect 25-50% of NF1 patients and can occur anywhere there are nerve fibers leading to significant morbidity and mortality depending on the size and location adjacent to vital structures 3 . Approximately 10% of PNs can transform to become malignant peripheral nerve sheath tumors over the lifetime of the patient 3 . Currently, surgical resection is the only available therapy, but is infrequently curative due to extensive invasion, making complete resection difficult 3 . No effective medical therapy currently exists; however, clinical trials are underway to evaluate chemotherapy which shrink tumor or limit growth. Therefore, accurate measurements of tumor burden are necessary to assess for tumor response. Two dimensional measurements are limited in assessing treatment change because of the common irregularity and

show abstract

PNist: interactive volumetric measurements of plexiform neurofibromas in MRI scans

Abstract: PNist is a new method for the semiautomatic segmentation of PN lesions. Its simplicity and reliability make it unique among other state-of-the-art methods. It has the potential to become a clinical tool that allows the reliable evaluation of PN burden and progression.

Cited by 8 publications

References 16 publications

MR diffusion-weighted imaging-based subcutaneous tumour volumetry in a xenografted nude mouse model using 3D Slicer: an accurate and repeatable method

MR diffusion-weighted imaging-based subcutaneous tumour volumetry in a xenografted nude mouse model using 3D Slicer: an accurate and repeatable method

Image segmentation of plexiform neurofibromas from a deep neural network using multiple b-value diffusion data

Image Segmentation of Plexiform Neurofibromas from a Deep Neural Network Using Multiple b-value Diffusion Data.

Contact Info

Product

Resources

About