Nicholas Tustison, Presenter: Nothing to Disclose

Talissa A. Altes MD, Abstract Co-Author: Nothing to Disclose

Eduard E. De Lange MD, Abstract Co-Author: Nothing to Disclose

Brian B. Avants, Abstract Co-Author: Nothing to Disclose

John P. Mugler PhD, Abstract Co-Author: Research grant, Siemens AG Research Consultant, Siemens AG

James C. Gee PhD, Abstract Co-Author: Nothing to Disclose

To develop an automated computerized segmentation protocol to automatically quantify the ventilated lung volume on 3He MR images of the lung in patients with asthma and normal volunteers, and to correlate the results with those from human readers who scored the images.

Our computational processing (CP) for each subjects consisted of the following 3 steps: 1) bias correction to remove inhomogeneities of the main magnetic field, 2) whole lung segmentation using a template-based approach, and 3) subdivision of the whole lung segmentation into regions of similar ventilation. Based on this subdivision, regions of poor ventilation are automatically identified and can be directly compared to the readers’ assessment. The results of the CP were compared to the scores by two independent human readers who were blinded to all clinical information and who scored the number of ventilation defects on previously obtained 3He ventilation MR scans in 43 subjects (8 normal and 35 asthmatic). Inter-rater agreement between CP and the two readers was characterized using Bland-Altman test plotting between pairings, and intra-class correlation coefficients (ICC) were calculated.

There was excellent correlation between CP and reader #1 (ICC = 0.86), CP and reader #2 (ICC = 0.85), and between the two readers (ICC = 0.97).

We have developed and validated an automated computerized method for quantifying the ventilated lung volume on 3He MRI, and achieved excellent concordance with human reader scoring. The findings strongly indicate that our proposed automated model may be a reliably, automatic method for evaluating ventilation defects in subjects with asthma.

We have developed an accurate segmentation technique for assessing ventilation defects in 3He ventilation MRI that could potentially substitute for human scorings, which are tedious and subjective.

Tustison, N, Altes, T, De Lange, E, Avants, B, Mugler, J, Gee, J, Development of an Automated Computerized Segmentation Technique to Quantify Ventilation Defects on 3He MR Images of the Lung in Patients with Asthma and Normal Volunteers.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9015168.html