Semi-auto. lung seg. of DCE-MRI data sets of 2-year old children after CDH repair
Link: http://www.ncbi.nlm.nih.gov/pubmed/26277730
Magn Reson Imaging. 2015 Aug 13. pii: S0730-725X(15)00196-4. doi: 10.1016/j.mri.2015.08.003. [Epub ahead of print]
Semi-automatic lung segmentation of DCE-MRI data sets of 2-year old children after congenital diaphragmatic hernia repair: Initial results.
Z?llner FG1, Daab M2, Weidner M3, Sommer V3, Zahn K4, Schaible T5, Weisser G3, Schoenberg SO3, Neff KW3, Schad LR2.
Author information
Abstract
In congenital diaphragmatic hernia (CDH), lung hypoplasia and secondary pulmonary hypertension are the major causes of death and severe disability. Based on new therapeutic strategies survival rates could be improved to up to 80%. However, after surgical repair of CDH, long-term follow-up of these pediatric patients is necessary. In this, dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) provides insights into the pulmonary microcirculation and might become a tool within the routine follow-up program of CDH patients. However, whole lung segmentation from DCE-MRI scans is tedious and automated procedures are warranted. Therefore, in this study, an approach to semi-automated lung segmentation is presented. Segmentation of the lung is obtained by calculating the cross correlation and the area under curve between all voxels in the data set and a reference region-of-interest (ROI), here the arterial input function (AIF). By applying an upper and lower threshold to the obtained maps and intersecting these, a final segmentation is reached. This approach was tested on twelve DCE-MRI data sets of 2-year old children after CDH repair. Segmentation accuracy was evaluated by comparing obtained automatic segmentations to manual delineations using the Dice overlap measure. Optimal thresholds for the cross correlation were 0.5/0.95 and 0.1/0.5 for the area under curve, respectively. The ipsilateral (left) lung showed reduced segmentation accuracy compared to the contralateral (right) lung. Average processing time was about 1.4s per data set. Average Dice score was 0.7?0.1 for the whole lung. In conclusion, initial results are promising. By our approach, whole lung segmentation is possible and a rapid evaluation of whole lung perfusion becomes possible. This might allow for a more detailed analysis of lung hypoplasia of children after CDH.
Copyright ? 2015. Published by Elsevier Inc.
KEYWORDS:
Congenital diaphragmatic hernia (CDH) repair; DCE-MRI; Lung perfusion; Segmentation; Semi-automatic
PMID: 26277730 [PubMed - as supplied by publisher