J Kullberg1, H Ahlström, L Johansson, H Frimmel. 1. Department of Oncology, Radiology and Clinical Immunology, Uppsala University Hospital, Uppsala, Sweden. joel.kullberg@radiol.uu.se
Abstract
OBJECTIVES: (1) To develop a fully automated algorithm for segmentation of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT), excluding intermuscular adipose tissue (IMAT) and bone marrow (BM), from axial abdominal magnetic resonance imaging (MRI) data. (2) To evaluate the algorithm accuracy and total method reproducibility using a semi-automatically segmented reference and data from repeated measurements. BACKGROUND: MRI is a widely used in adipose tissue (AT) assessment. Manual analysis of MRI data is time consuming and biased by the operator. Automated analysis spares resources and increase reproducibility. Fully automated algorithms have been presented. However, reproducibility analysis has not been performed nor has methods for exclusion of IMAT and BM been presented. METHODS: In total, 49 data sets from 31 subjects were acquired using a clinical 1.5 T MRI scanner. Thirteen data sets were used in the derivation of the automated algorithm and 36 were used in the validation. Common image analysis tools such as thresholding, morphological operations and geometrical models were used to segment VAT and SAT. Accuracy was assessed using a semi-automatically created reference. Reproducibility was assessed from repeated measurements. RESULTS: Resulting AT volumes from the automated analysis and the reference were not found to differ significantly (2.0+/-14% and 0.84+/-2.7%, given as mean+/-s.d., for VAT and SAT, respectively). The automated analysis of the repeated measurements data significantly increased the reproducibility of the VAT measurements. One athletic subject with very small amounts of AT was considered to be an outlier. CONCLUSIONS: An automated method for segmentation of VAT and SAT and exclusion of IMAT and BM from abdominal MRI data has been reported. The accuracy and reproducibility of the method has also been demonstrated using a semi-automatically segmented reference and analysis of repeated acquisitions. The accuracy of the method is limited in lean subjects.
OBJECTIVES: (1) To develop a fully automated algorithm for segmentation of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT), excluding intermuscular adipose tissue (IMAT) and bone marrow (BM), from axial abdominal magnetic resonance imaging (MRI) data. (2) To evaluate the algorithm accuracy and total method reproducibility using a semi-automatically segmented reference and data from repeated measurements. BACKGROUND: MRI is a widely used in adipose tissue (AT) assessment. Manual analysis of MRI data is time consuming and biased by the operator. Automated analysis spares resources and increase reproducibility. Fully automated algorithms have been presented. However, reproducibility analysis has not been performed nor has methods for exclusion of IMAT and BM been presented. METHODS: In total, 49 data sets from 31 subjects were acquired using a clinical 1.5 T MRI scanner. Thirteen data sets were used in the derivation of the automated algorithm and 36 were used in the validation. Common image analysis tools such as thresholding, morphological operations and geometrical models were used to segment VAT and SAT. Accuracy was assessed using a semi-automatically created reference. Reproducibility was assessed from repeated measurements. RESULTS: Resulting AT volumes from the automated analysis and the reference were not found to differ significantly (2.0+/-14% and 0.84+/-2.7%, given as mean+/-s.d., for VAT and SAT, respectively). The automated analysis of the repeated measurements data significantly increased the reproducibility of the VAT measurements. One athletic subject with very small amounts of AT was considered to be an outlier. CONCLUSIONS: An automated method for segmentation of VAT and SAT and exclusion of IMAT and BM from abdominal MRI data has been reported. The accuracy and reproducibility of the method has also been demonstrated using a semi-automatically segmented reference and analysis of repeated acquisitions. The accuracy of the method is limited in lean subjects.
Authors: Defeng Wang; Lin Shi; Winnie C W Chu; Miao Hu; Brian Tomlinson; Wen-Hua Huang; Tianfu Wang; Pheng Ann Heng; David K W Yeung; Anil T Ahuja Journal: Med Biol Eng Comput Date: 2015-08-06 Impact factor: 2.602
Authors: Jon D Klingensmith; Addison L Elliott; Amy H Givan; Zechariah D Faszold; Cory L Mahan; Adam M Doedtman; Maria Fernandez-Del-Valle Journal: J Med Imaging (Bellingham) Date: 2019-02-07
Authors: Aziz H Poonawalla; Brett P Sjoberg; Jennifer L Rehm; Diego Hernando; Catherine D Hines; Pablo Irarrazaval; Scott B Reeder Journal: J Magn Reson Imaging Date: 2012-10-10 Impact factor: 4.813