Program Information
Development of An MRI Radiomics Phantom
J Yang*, A Steinmann , D Mackin , R Stafford , D Followill , J Li , L Court , The University of Texas MD Anderson Cancer Center, Houston, TX
Presentations
TU-H-FS4-9 (Tuesday, August 1, 2017) 4:30 PM - 6:00 PM Room: Four Seasons 4
Purpose: To identify materials that could be used in a radiomics phantom for MRI studies. This phantom will be used to evaluate uncertainties of MRI radiomic features such as scanning protocol variability and inter-scanner variability studies.
Methods: Ninety-seven patients were identified who had had pathologic confirmation of necrosis or progression of their brain lesions. The lesion were manually contoured on both T1-weighted and T2-weighted MR images at 2 time points per patient, and 90 texture features were extracted using IBEX software, including 19 intensity features, 5 neighborhood gray-tone matrix features, and 66 gray-level co-occurrence matrix features. A total of 194 patient population samples were generated for each feature, each MR series. T1-weighted and T2-weighted MR images were then taken of 57 different phantom materials. For each material the same 90 texture features were extracted. The suitability of each material for use in a radiomics phantom was evaluated by comparing each feature value with the patient population data (mean ± 2 standard deviations).
Results: For T1-weighted images, an average of 91.5% phantom texture features were within the patient population range (range: 78.9%-98.9%). Thirty-six materials had more than 90% of texture features within this range. For T2-weighted images, an average of 78.1% phantom texture features were within the patient population range (range: 55.6%-94.4%). Only one material had more than 90% of texture features within this range.
Conclusion: Potential materials were identified for developing MR radiomics phantoms. For different MR series such as T1-weighted or T2-weighted images, different materials should be used for the radiomics phantom.
Contact Email: