Contributions
Abstract: P1224
Type: Poster
Abstract Category: Therapy - disease modifying - 30 Tools for detecting therapeutic response
Background: Magnetic resonance imaging (MRI) is crucial for in vivo detection and characterization of white matter lesions (WML) in multiple sclerosis. The most widely established MRI outcome measure is T2-weighted lesion (T2L) volume. Unfortunately, T2L volume is non-specific for the level of tissue destruction and shows a weak relationship to clinical status. Consequently, some researchers have focused on T1-weighted hypointense lesion (T1L) (“black holes”) volume quantification to provide more specificity for axonal loss and a closer link to neurologic disability.
Objectives: This study aimed to adapt and assess the performance of an automatic T2L segmentation algorithm for segmenting T1L. First, to ensure the method performed well on the images of interest, we compared the automated T2L segmentation with gold standard manual segmentations. We then adapted the pipeline and compared the automated T1L segmentation with gold standard manual segmentations.
Methods: High-resolution 3D T1, T2, and FLAIR sequences were acquired from 40 MS subjects on a Siemens 3T Skyra unit at the Brigham and Women's Hospital (Boston, MA). Trained observers under the supervision of an experienced observer manually segmented T2L and T1L. We processed data by N4 inhomogeneity correction, rigid registration, skull stripping, and intensity normalization. We then employed MIMoSA, an automated segmentation algorithm, which has previously shown competitive performance for segmenting T2L. MIMoSA utilizes complementary imaging modalities to emphasize different tissue properties, which can help identify and characterize interrelated features of lesions, in a local logistic regression to model the probability that any voxel is part of a lesion.
Results: Using bootstrap cross-validation, we found that MIMoSA is a robust method that can be applied to segment both T2L and T1L. For T2L, we found a Sorenson's dice coefficient (DSC) of 0.61 and partial AUC (pAUC) up to 1% false positive rate of 0.69. For T1L, we found 0.48 DSC and 0.63 pAUC.
Conclusions: MIMoSA is a fully automated segmentation algorithm. Though originally designed to segment T2L, MIMoSA is able to segment T1 black holes with both sensitivity and specificity in patients with MS.
Study support: NIH R01NS085211, R21NS093349, and R01MH112847. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Disclosure:
Dr. Bakshi has received consulting fees from EMD Serono, Genentech, Sanofi-Genzyme, and Novartis and research support from Biogen, EMD-Serono, Novartis, and Sanofi-Genzyme.
Dr. Shinohara has received consulting fees from Genentech and Roche.
Abstract: P1224
Type: Poster
Abstract Category: Therapy - disease modifying - 30 Tools for detecting therapeutic response
Background: Magnetic resonance imaging (MRI) is crucial for in vivo detection and characterization of white matter lesions (WML) in multiple sclerosis. The most widely established MRI outcome measure is T2-weighted lesion (T2L) volume. Unfortunately, T2L volume is non-specific for the level of tissue destruction and shows a weak relationship to clinical status. Consequently, some researchers have focused on T1-weighted hypointense lesion (T1L) (“black holes”) volume quantification to provide more specificity for axonal loss and a closer link to neurologic disability.
Objectives: This study aimed to adapt and assess the performance of an automatic T2L segmentation algorithm for segmenting T1L. First, to ensure the method performed well on the images of interest, we compared the automated T2L segmentation with gold standard manual segmentations. We then adapted the pipeline and compared the automated T1L segmentation with gold standard manual segmentations.
Methods: High-resolution 3D T1, T2, and FLAIR sequences were acquired from 40 MS subjects on a Siemens 3T Skyra unit at the Brigham and Women's Hospital (Boston, MA). Trained observers under the supervision of an experienced observer manually segmented T2L and T1L. We processed data by N4 inhomogeneity correction, rigid registration, skull stripping, and intensity normalization. We then employed MIMoSA, an automated segmentation algorithm, which has previously shown competitive performance for segmenting T2L. MIMoSA utilizes complementary imaging modalities to emphasize different tissue properties, which can help identify and characterize interrelated features of lesions, in a local logistic regression to model the probability that any voxel is part of a lesion.
Results: Using bootstrap cross-validation, we found that MIMoSA is a robust method that can be applied to segment both T2L and T1L. For T2L, we found a Sorenson's dice coefficient (DSC) of 0.61 and partial AUC (pAUC) up to 1% false positive rate of 0.69. For T1L, we found 0.48 DSC and 0.63 pAUC.
Conclusions: MIMoSA is a fully automated segmentation algorithm. Though originally designed to segment T2L, MIMoSA is able to segment T1 black holes with both sensitivity and specificity in patients with MS.
Study support: NIH R01NS085211, R21NS093349, and R01MH112847. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Disclosure:
Dr. Bakshi has received consulting fees from EMD Serono, Genentech, Sanofi-Genzyme, and Novartis and research support from Biogen, EMD-Serono, Novartis, and Sanofi-Genzyme.
Dr. Shinohara has received consulting fees from Genentech and Roche.