ISBE-AstraZeneca Strategic Alliance Project 26 Evaluation of Crohn’s disease using T1-weighted dynamic contrast- enhanced MRI (DCE-MRI) Karl Embleton
Project objectives develop an M. R. based imaging technique to provide dynamic contrast enhanced images for investigation of inflammatory bowel disease derive disease volume measurements from contrast enhanced images of inflammatory bowel disease assess the application of endothelial permeability measurements to inflammatory bowel disease assess reproducibility of these measurements and the ability to identify treatment effects
Current recruitment Eight patients scanned at present Two of these scanned three times (or about to be), 2 x pre-treatment 1 x post-treatment Karl Embleton now has a contract with Hope Hospital Ethics in process of amendment to allow recruitment from active patients not undergoing conventional radiology (approximately 2x the number) Application in progress to allow scanning at Wellcome Trust Clinical Research Facility Hope hospital project nurse recruited for informing patients at time of conventional examination
Morphological images from consecutive days visit 1visit 2
Surface renderings of ROI drawn for inflamed region visit 1visit 2
Permeability parameters for inflamed ROI No voxels in ROI KtransVeVpKfpVpfp visit ± ± ± ± ± visit ± ± ± ± ±0.0183
Problems drawing ROI Drawing region of interest very time consuming Very difficult to draw accurately Voxel size my be greater than bowel wall thickness Substantial differences in position and shape of inflamed areas on consecutive days Can only include in ROI obviously inflamed and thickened tissue, which may be older, and fibroid in nature and not the most currently active region of disease By limiting the analysis to ROI we may actually miss the most interesting regions
Parameter based segmentation Alternative is to set thresholds on permeability parameters and measure volume of tissue above these thresholds. First step is to produce region of interest for all the abdominal cavity, excluding obviously non bowel organs e.g. uterus
Parameter based segmentation surface rendering of all voxels ktrans > 0.2 & Ve > 0.1 same voxels after application of a Gaussian filter kernel size 3x3
Surface renderings of voxels above various thresholds Ktrans > 0 Ktrans > 0.2 Ktrans > 0.4 Ktrans > 0.6 Ktrans > 0.8 Ktrans > 1.0 Ve > 0 Ve > 0.1 Ve > 0.2 Ve > 0.3 Ve > 0.4 Ve > 0.5
Voxel counts Ktrans > 0 Ktrans > 0.2 Ktrans > 0.4 Ktrans > 0.6 Ktrans > 0.8 Ktrans > 1.0 Ve > 0 Visit 1 Visit Ve > Ve > Ve > Ve > Ve >
Surface renderings on consecutive days visit 1 all voxels visit 2 all voxels Gaussian smoothed
Coronal sequences taken on consecutive days