Multi-modal Glioblastoma Segmentation: Man versus Machine

Porz, Nicole; Bauer, Stefan; Pica, Alessia; Schucht, Philippe; Beck, John; Verma, Ruchika; Slotboom, Johannes; Wiest, Roland

doi:10.1055/s-0034-1383771

Cited by 33 publications

(49 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our analysis, a wavelet-based quality measure [46] is used for blurriness quantification. The quality measure is defined as Q = 3 . This case however, holds the highest level of blurriness, approximately 4-6 times higher than the average blurriness of DataSet1.…”

Section: Discussionmentioning

confidence: 99%

“…In clinical practice, expert radiologists obtain complete information about the brain tumor by multi-channel MR imaging and perform the delineation task manually. However, a non-automated segmentation process is subject to variations, not only between different radiologists but also by the same expert [3,4]. Moreover, the neoplastic tumor can vary significantly in geometric properties, location and degree of enhancement among patients, making the segmentation and characterization of brain abnormalities a very demanding, difficult and highly time consuming task.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A low cost approach for brain tumor segmentation based on intensity modeling and 3D Random Walker

Kanas

Zacharaki

Davatzikos

et al. 2015

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A low cost approach for brain tumor segmentation based on intensity modeling and 3D Random Walker

Kanas

Zacharaki

Davatzikos

et al. 2015

Biomedical Signal Processing and Control

View full text Add to dashboard Cite

“…50,68 Variability in segmentation and ROI selection is another issue, which may be addressed by ever-improving automated methods of extracting imaging feature sets. 73 The use of stereotactic biopsies, rather than random sampling of the tumor, will help generate microarray analysis that is able to link gene expression data with specific imaging traits within a single tumor. 38 Physiologic imaging with perfusion, diffusion MRI, and PET scans provides a wealth of information that is only just beginning to be incorporated into these analyses.…”

Section: Future Directionsmentioning

confidence: 99%

Genomics of Brain Tumor Imaging

Pope

2015

Neuroimaging Clinics of North America

View full text Add to dashboard Cite

“…The proposed method was used to identify the main types of metabolic patterns seen in high‐grade gliomas (HGGs). The spatial distribution of the metabolic abnormalities was compared with the tumor segmentation performed by BraTumIA (Brain Tumor Image Analysis), which is an automatic brain tumor segmentation method trained to reproduce the manual MRI segmentation performed by experienced neuroradiologists. Besides this, we also include an analysis of two follow‐up cases where the MRSI maps are compared with the corresponding structural MRI acquired several months after, showing the ability of MRSI to predict future contrast enhancement in HGG.…”

Section: Introductionmentioning

confidence: 99%

Analysis of metabolic abnormalities in high‐grade glioma using MRSI and convex NMF

et al. 2019

View full text Add to dashboard Cite

Clinical use of MRSI is limited by the level of experience required to properly translate MRSI examinations into relevant clinical information. To solve this, several methods have been proposed to automatically recognize a predefined set of reference metabolic patterns. Given the variety of metabolic patterns seen in glioma patients, the decision on the optimal number of patterns that need to be used to describe the data is not trivial. In this paper, we propose a novel framework to (1) separate healthy from abnormal metabolic patterns and (2) retrieve an optimal number of reference patterns describing the most important types of abnormality. Using 41 MRSI examinations (1.5 T, PRESS, T E 135 ms) from 22 glioma patients, four different patterns describing different types of abnormality were detected: edema, healthy without Glx, active tumor and necrosis. The identified patterns were then evaluated on 17 MRSI examinations from nine different glioma patients. The results were compared against BraTumIA, an automatic segmentation method trained to identify different tumor compartments on structural MRI data. Finally, the ability to predict future contrast enhancement using the proposed approach was also evaluated. KEYWORDS applications, cancer, head and neck cancer methods and engineering, MRS and MRSI methods, post-acquisition processing, spectroscopic imaging, visualization methods and engineering 1 | INTRODUCTION MRS provides relevant metabolic information for the assessment of brain tumors, allowing us to distinguish different tumor types and grades, 1-3 distinguish radiation effects (pseudoprogression) from true progression 4,5 and identify regions with high tumor cellularity that are not visible in structural MRI. 6-8 Many publications 9 focus on the translation of one or two MRS features, such as choline (Cho)/NAA (N-acetyl aspartate) andCho/Cr (creatine), into clinically meaningful information for the tasks described above. Regardless of what can already be achieved with the analysis of individual metabolite ratios, the use of metabolic patterns for the identification of tissue types and diseases has the potential to allow more precise characterization of brain tumors.Several approaches 10-21 have been suggested for the analysis of metabolic patterns present in brain tumor MRS data. Before such methods can be used to interpret new data, a library of metabolic patterns has to be defined. The decision on the order of the model used to interpret .3 + Lip0.9; Lip0.9, 0.9 ppm signals originating from the -CH3 moiety of lipid molecules (triglycerides); Lip1.3, 1.3 ppm signals originating from the -CH2 moiety of lipid molecules (triglycerides); LOH, loss of heterozygosity; LOOCV, leave-one-out cross-validation; NAA, N-acetyl aspartate; NNLS, non-negative least squares

show abstract

Multi-modal Glioblastoma Segmentation: Man versus Machine

Cited by 33 publications

References 32 publications

A low cost approach for brain tumor segmentation based on intensity modeling and 3D Random Walker

A low cost approach for brain tumor segmentation based on intensity modeling and 3D Random Walker

Genomics of Brain Tumor Imaging

Analysis of metabolic abnormalities in high‐grade glioma using MRSI and convex NMF

Contact Info

Product

Resources

About