MRI Protocols in Experimental Stroke Xenios Milidonis Centre for Clinical Brain Sciences The University of Edinburgh MultiPART Meeting, Barcelona, 2 October 204
Objective In MultiPART magnetic resonance imaging (MRI) data from different scanners will be combined. Effective pooling will require standardisation of acquisition protocols and postprocessing methods. We aim to examine the variability in the use of MRI in animal models of stroke for measuring the infarct size, and develop standardised strategies to facilitate multicentre studies.
Do you use the following to assess infarct volume? TTC MRI Yes 62% No 38% Yes 67% No 33% Histology No 3% Yes 97%
Type of Histology? Luxol fast blue Nitroblue tetrazolium (NBT) NeuN Immunohistochemistry Silver Thionin H&E Cresyl Violet 0 0 20 30
Systematic review of MRI in experimental stroke 54 articles identified in CAMARADES metaanalysis database 78 scanning protocols for the assessment of lesion sizes: T2weighted imaging Diffusionweighted imaging Tweighted imaging Perfusionweighted imaging Plasma volume imaging Unknown 45 23 4 3 2 46 comparisons between MRI and histology: T2weigthted imaging Diffusionweighted imaging Tweighted imaging Perfusionweighted imaging TTC Hematoxylin & eosin Cresyl violet 33 4 4 Evaluate the heterogeneity and reporting of MRI parameters Identify which MRI method best correlates with histology using metaregression
Criterion Study quality Peerreviewed publication Control of temperature Compliance with animal welfare regulations Use of anaesthetic without marked intrinsic neuroprotective activity Random allocation to groups Blinded assessment of outcome Statement of potential conflicts of interest Blinded induction of ischemia 9942000 2002005 2006200 Overall Use of comorbid animals Sample size calculation 0 0 20 30 40 50 60 70 80 90 00 Prevalence (%)
Basic imaging characteristics () Magnet Type Monitoring Anaesthetic of field radiofrequency during strength for scanning MRI (Tesla) coil 94% 30% 63% 52% of studies 7 Temperature Two Ketamine types 4.7 Isoflurane Volume 3 Temperature & Halothane.5 other 2.35 Surface Pentobarbital 2 Human wrist Other 9.4 FFM 2 34 8 8 0 9 4 9 25 38 44 36 36 29 29 29 29 29 0 0 0 0 0 20 20 30 20 20 3040 50 30 40 (%)
Basic imaging characteristics (2) Oedema Method Presentation Side of of correction ischaemia lesion of lesion delineation for brain infarct size 24% of 78% 65% 99% (relevant) of studies studies Volume Thresholding Swanson's Right Compared method to ipsi * Change from baseline Manual tracing Compared to contra Normalised to Compared Left to brain hemisheric Automated ratio 4 Area at one slice 20 9 8 298 5 8 58 76 782 0 0 200 020 40 3020 40 60 40 60 80 90 60 (%) * J. Cereb. Blood Flow Metab. 990;0:290293
MRI scanning parameters Parameter Echo time (ms) Repetition time (ms) T2weighted imaging Diffusionweighted imaging range median range median 3020 3020 7508000 0003000 Number of slices 635 525 Slice thickness (mm) Field of view (mm mm) Matrix size (pixels pixels) Number of averages 0.353 0.5 22 2220 20 25.6 2.820 20 64 6452 52 28 64256 256 28 86 60 66 3000 2550 4 3 4 2 3300 30 0004000 000000 43 5 2 0.5 22 2280 80 22 6 64 64256 256 28 64 24 N/G 40 30 3000 000 8 5.5 0.5 3 N/G Rat Mouse
T2weighted imaging correlates with histology Method Time poststroke (h) Compar. n Effect size [95% CI] (%) Histology 20 25.49 [7.52, 33.46] 48 336 33 T2WI 40 22.22 [7.90, 26.55] Histology 05 3.85 [24.02, 39.68] 2 74 DWI 47 42.43 [30.88, 53.98] Metaregression Coeff. R 2 p.08 0.699 <0.00 0.3 0.433 0.36
Assessment of MRI using a structural phantom Internal dimensions Assessment of scaling in 3 directions Cylindrical volume Assessment of acquisition parameters n... 2 n i= Area i thickness
Scaling error (%) Differences between pulse sequences 7T Agilent preclinical scanner, 20mm bore gradient coil, 72mm (ID) volume transmit & 2 channel mouse phased array surface receive coil Fast spin echo (FSE) vs. Gradient echo (GE) (9.2 9.2mm FOV, 256 256 matrix, mm slice thickness, 2 averages) 6 5 4 3 X (rightleft) Y (anteriorposterior) Z (superiorinferior) 2 FSE GE Type of pulse sequence
Phantom volume (mm 3 ) Impact of scanning parameters on volume measurements Fast spin echo (TR/TE eff =500/20ms, 9.2 9.2mm FOV, 256 256 matrix, mm slice thickness, ETL=4, 2 averages)...then changed one parameter at a time 260 69% 240 220 true volume 200 80 4% 60 40 TR (ms) TE eff (ms) FOV (mm) Matrix size Slice th. (mm) Averages ETL RB (khz)
Scaling error (%) Scaling error (%) Within and betweenscanner variability Edinburgh Preclinical Imaging (EPI) 7T Agilent scanner Glasgow Experimental MRI Centre (GEMRIC) 7T Bruker scanner 7 6 7 6 ** * * *p<0.0 **p<0.00 5 5 4 3 X Y Z 4 3 X Y Z 2 2 0 April 4 May 4 July 4 0 EPI GEMRIC Time of scanning (EPI) Preclinical centre (July 4)
Future plans Assessment of MRI acquisition protocols using the phantom Other pulse sequences Other coils Other scanners Assessment of postprocessing methods for calculating the infarct size from animal data Evaluate performance based on ground truth (manual, histology?) Estimate betweenmethod variation Survey under construction