Publications
Text queries can be conducted by Author, Title, or Keyword.
Cross Validation of Tissue Classification and
Surface Modeling Algorithms for Determining Growth
Rates of Malignant Gliomas: Prognostic Value of Growth
Rates and MR Spectroscopy
Source: 2000 Int’l Conference on Mathematics and Engineering Techniques in Medicine 2000;.
Author: Haney S, Thompson PM, Cloughesy TF, Alger JR, Frew A, Toga AW
Abstract:
A tissue classification method and a surface modeling algorithm were compared in their ability to analyze changes in brain tumors, based on volumetric MRI data. Measures were derived from serially acquired T2 and gadolinium-enhanced T1-weighted SPGR (spoiled Grass) MRIs. Volumes for contrast enhancing tissue, necrosis, and edema were determined and cross-validated against manually defined volumes. Volumes generated by both algorithms were highly correlated with volumes generated by manual segmentation (r2=0.99 for the tissue segmentation method; r2=0.96 for the surface modeling algorithm). Growth rates were calculated from contrast enhancing tissue volumes. Growth rates derived from the tissue classification approach were highly correlated with growth rates derived from manually segmented images (r2=0.94). Growth rates were significantly correlated with survival (p<0.03) as was the choline to creatine ratio (CHO/CRE; p<0.02). [1H]-MR spectroscopy measures, linked to the rates of cellular proliferation, were also examined to assess their relationship with growth rates.
Source: 2000 Int’l Conference on Mathematics and Engineering Techniques in Medicine 2000;.
Author: Haney S, Thompson PM, Cloughesy TF, Alger JR, Frew A, Toga AW
Abstract:
A tissue classification method and a surface modeling algorithm were compared in their ability to analyze changes in brain tumors, based on volumetric MRI data. Measures were derived from serially acquired T2 and gadolinium-enhanced T1-weighted SPGR (spoiled Grass) MRIs. Volumes for contrast enhancing tissue, necrosis, and edema were determined and cross-validated against manually defined volumes. Volumes generated by both algorithms were highly correlated with volumes generated by manual segmentation (r2=0.99 for the tissue segmentation method; r2=0.96 for the surface modeling algorithm). Growth rates were calculated from contrast enhancing tissue volumes. Growth rates derived from the tissue classification approach were highly correlated with growth rates derived from manually segmented images (r2=0.94). Growth rates were significantly correlated with survival (p<0.03) as was the choline to creatine ratio (CHO/CRE; p<0.02). [1H]-MR spectroscopy measures, linked to the rates of cellular proliferation, were also examined to assess their relationship with growth rates.