Advanced MR Imaging in Epilepsy
Dr. Vincent Lai MBChB, FRCR(UK), FHKCR, FHKAM(Radiology) Consultant Radiologist, Hong Kong BapBst Hospital Honorary Clinical Assistant Professor, University of Hong Kong
Overview
General imaging findings & concept Various funcBonal imaging techniques Our preliminary work
Introduc:on
• Very heterogeneous imaging spectrum
• IdenBficaBon of epileptogenic lesion is crucial in achieving seizure free outcome aPer surgery
E:ologies of Epilepsy MalformaBon of corBcal development Focal corBcal dysplasia
Heterotopia Polymicrogyria
Mesial temporal sclerosis
Tumor DysembryoplasBc neuroepithelial tumor Ganglioglioma Astrocytoma Oligodendroglioma Cavernoma
Vascular cerebral insult Chronic corBcal/subcorBcal infarct Arteriovenous malformaBon Amyloid angiopathy
Nonvascular cerebral insult Post-‐traumaBc PostoperaBve PostencephaliBs Postanoxia
Others Abnormal venous drainage Arachnoid cyst/ neuroepithelial cyst Focal calcificaBon/ corBcal atrophy
Goal of Epilepsy Imaging
Detec:on of epileptogenic lesion
Localiza:on of epileptogenic lesion
TriangularizaBon amongst Seizure emiology, EEG & Imaging
Current MR Imaging Considera:on High resoluBon structural imaging
3D MPRAGE/SPGR T1W, Oblique Coronal/3D T2W & FLAIR T2W
SuscepBbility weighted imaging FuncBonal imaging
Radionuclide, T2 relaxometry, MRS, Diffusion, ASL, MR volumetry
Are we doing well?
ProblemaBc issue in MR-‐negaBve paBents
Does this really exist?
Where are we upto ?
• 2/3 of MR negaBve paBents have idenBfiable lesion (oOen subtle MCD) on 3.0 T – Temporal 50% – Frontal 40% – Majority is FCD
Knake S et al. 2005 Neurology
• 65% of drug resistant epilepsy has idenBfiable lesion – Frequently MTS
Vezina LG 2011 Epilepsia
M/ 22 yrs old
LeO hippocampal FCD
Taylor’s FCD with balloon cells, radial band
Colombo N et al. 2003 AJNR
F/4.5 yrs old
Right insular FCD
Malforma:on of Cor:cal Development
Agyria Polymicrogyria
Grant PE et al. 1997 AJNR
TPO Syndrome/ Posterior Quadran:c Cor:cal Dysplasia
Low Grade Astrocytoma
MTS
Parahippocampus
Forms mesial & inferior gyrus of temporal lobe
Includes: Entorhinal & perirhinal corBces Parahippocampal cortex Contributes to: Seizure iniBaBon epileptogenesis
Parahippocampal epilepsy
A subset of MTLE A cause of MR-‐negaBve MTLE
T2W hyperintense signal in parahippocmapal WM
Blurring of GW juncBon Normal corBcal thickness
Pillay N er al. 2009 Epilepsia
Challenging Issue
Microdysgenesis of neocortex or subtle MCD
Majority of MR-‐nega:ve PET-‐posi:ve cases
GeneBc Early
environmental factors
Disrup:on of normal cor:cal development
Mild MCD (I & II) in 12-40%: Carne RP et al. 2004 Brain; Huba R et al. 2012 Epliepsy & Behavior
So, how can we do it?
Radionuclide Imaging
PET vs SPECT
Noninvasive Presurgical mapping
Uses: • MR-‐negaBve • Several lesions • Discordant findings between EEG & structural imaging
Advanced MR Imaging Techniques
T2 Relaoxmetry MR Volumetry
MR Spectroscopy Diffusion Tensor Imaging Arterial Spin labeling
T2 Relaxometry
T2 relaxaBon Bme ↑ in the epilepBc focus Woermann et al. 1998; Namer et al. 1998; Van Paesschen et al. 1995; Jackson et al. 1993)
SuggesBon of superior detecBon rate as compared with volumetry
Bernasconi A et al. 2000 Neuroimage
But not confirmed in later and recent study with more advanced MR volumetry: – Improved detec_on rate in 19% of pa_ents only
Coan AC et al. 2013; AJNR
T2 Relaxometry
False +ve in upto 50% of visually detected T2 signal changes in hippocampus
Sumar I et al. 2011; Epilepsy research
Voxel based quan:ta:ve analysis is more reliable
MR Volumetry i. Segmenta_ons by VBM ii. Orienta_on-‐corrected, spaBally normalized, Bssue classified iii. Par__on the whole-‐brain to GM, WM and iv. FIRST: fieng a mesh to the surface of the amygdala & hippocampus
Manual vs automated segmenta:on
In a study of 46 paBent Manual method vs various automated methods
LocalInfo > HAMMER > FreeSurfer
Akhondi-Asl A et al. 2011 Neuroimage
LateralizaBon accuracy: Manual (78%)
Automated (74%)
Quan:ta:ve MR volumetry in hippocampal atrophy
Automated MR volumetry Hippocampal asymmetry (pa_ents & normal)
High discriminaBng power Sensi_vity 89.5%; Specificity 94.1%
LateralizaBon accuracy: 88% (visual inspec_on: 76-‐85%)
Farid N et al. 2012 Radiology
? Performance in 3T
In a study of 203 paBents with hippocampal sclerosis…
Coan AC et al. 2013 AJNR
Help ↑ detec:on rate in 28% of pa:ents with hippocampal
sclerosis
Pi\alls
reflects a combinaBon of
cor:cal thickness
& surface area measurements
Morphological analysis Average convexity (Fischl et al. 1999) Sharpness/ Curvedness/ Folding index (Pienaar et al. 2008) GyrificaBon index (Schaer et al. 2008) Sulcal paiern (Kim et al. 2008)
Shape parameter -‐ Jacobian matrix (Ronan et al. 2011) Surface area/geometric distorBon (Alhusaini et al. 2012)
Reflects changes in underlying connectivity and white matter tracts
So, they advocate…
Surface-‐based MRI morphometry post-‐processing surface reconstruc_on morphometric measures lesion tracing
Sn 92%, Sp 96%
Thesen T et al. 2011 Open Access
MR Spectroscopy
MRS – General Principles Molecules/ Metabolites
Func:on/ Clinical relevance
NAA Marker of neuronal density & viability
Cr Marker for energy–dependent system
Cho Marker of increased inflammatory/glioBc process & pathological changes in membrane turnover
Lactate Elevated aPer seizure & in hypoxia/ischemic injury Mitochondral disorder
Glutamate Major excitatory neurotransmiier, toxic if elevated leading to neuronal death
ml Marker of gliosis
Typical Spectra Short TE (35 ms)
Long TE (144 ms)
Very Long TE (288 ms)
Single vs Mul:-‐voxel Spectroscopy
Higher SNR Short acquisiBon Bme (~3 mins) Metabolic disease 1. 1 voxel at BG 2. 3 voxels at CS, LN, OP cortex Temporal lobe epilepsy 2 voxels at bilateral hippocampi
Single-‐voxel Mul:-‐voxel
Larger volume of informaBon Long acquisiBon Bme (~8 mins) Allow 3D acquisiBon
Technical Considera:ons • Higher magneBc field strength (higher SNR)
• MulBchannel (32-‐channel) receiver coils (higher SNR) Keil B et al. 2012 Magn Reson Med
• Shimming (maximise B0 homogeneity) Kanayanma S et al. 1996 Magn Reson Med; Hetherington HP et al. 2006 Magn Reson Med
• Fast spiral acquisiBon (allow fast spa_al encoding) Andronesi OC et al. 2012 Radiology
• AdiabaBc pulses (compensate for radiofrequency inhomgeneity) Garwood M et al. 1989 Magn Reson Med; Andronesi OC et al. 2010 J Magn Reson
Role of MRS
Screening of metabolic derangement Adjuvant in evaluaBon of medically refractory TLE CharacterizaBon of lesions/ masses ?Localizing techniques in extratemporal epilepsy
MRS Screening of metabolic derangement
↓ Cho in normal appearing cerebellar WM (80%) peritrigonal WM (67%) corBcal GM (60%)
↓ NAA/Cr in normal appearing cerebellum (93%) cortex (87%)
↑ Amino acids & Lactate
Bianchi C et al. 2003 AJNR
Mitochondrial disorders
MRS Screening of metabolic derangement
Cr deficiency Prevalence: 0.25% Inherited enzymaBc defects: AGAT GAMT SLC6A8
↓ Cr in normal brain
Arias A et al. 2007 Clin Biochem
EnzymaBc disorders
MRS Adjuvant in evaluaBon of medically refractory TLE
↓ NAA
↓ NAA/Cr raBo 86% agreement with EEG (c.f. 83% for volumetry with EEG)
12% in MR-‐negaBve TLE
Cendes F et al. 1997 Ann Neurol; Kuzniecky R et al. 1998 Neurology
TLE
MRS CharacterizaBon of lesions/ masses
FCD ↓ NAA/Cr raBo ↑ GABA, akanine, tyrosine, lactate, inositol
No elevaBon in Cho/NAA Pathology: type IIB FCD
Caruso PA et al. 2013 Neuroimag Clin N Am
FCD vs Neoplasm
MRS CharacterizaBon of lesions/ masses
Astrocytoma ↓ NAA, ↑ Cho
↑ Cho/NAA , ↓ NAA/Cr raBos Pathology: Angiocentric astrocytoma
Caruso PA et al. 2013 Neuroimag Clin N Am
FCD vs Neoplasm
MRS LocalizaBon in nonlesional epilepsy
Using MVS & subdural electrodes Areas of ↑ Cho/NAA & ↓ NAA/Cr
raBos overlapped with ictal zones
Krsek P et al. 2007 Eur Radiol
FCD
Diffusion Tensor Imaging
DTI – General Principle
Measurement of:
Magnitude & Direc:on Of
water diffusion
Indirect evalua:on of integrity of axonal microenvironment
anisotropic
Quan:fica:on
Frac:onal Anisotropy (FA)
Ranges from 0 – 1
0: isotropic diffusion 1: anisotropic diffusion
MD
λ1, λ2, λ3
Normal fiber tracts:
Anisotropic (high FA)
Degenerated fiber tracts:
↓ FA Wallerian degenera_on
Demyelina_on/ dysmyelina_on Maldevelopment
FCD Significant reducBon of FA in underlying subcorBcal WM
HypomyelinaBon ? SeneiBvity
Gross DW et al. 2005 Can J Neurol Sci
Technical consideration: Tesla; no of gradient…
MTLE
Widespread WM changes
↓ FA values
PosiBve correlaBon with hippocampal volume
Scanlon C et al. 2013 J Neurol; Oquz KK et al. 2013; AJNR
Arterial Spin Labeling
Noninvasive EvaluaBon of CBF
Interictal – hypoperfused; Ictal -‐ hyperperfused Wolf et al. 2001 AJNR; Madan N et al. 2009 Epilepsia
15 children 18F-‐FDG PET and DTI MRI Hypometabolism correlates with DTI indices
MR+ve & MR-‐ve pa_ents
Correla:on with Radionuclide Imaging
Lippe S et al. 2012 Epileptic disord
Our Preliminary Work…
T1 rho MR Imaging
Provides informaBon on slow molecular moBon – Transverse magneBzaBon of T1 is “locked” by spin-‐lock frequency – Made to decay slower – Followed by convenBonal imaging – GeneraBon of T1rho map
In neuroimaging, has been uBlized in: – Brain tumors – AD and Parkinson’s disease
Hypothesis & Aim
Hypothesis:
T1 rho imaging is able to reflect early neuronal loss in the epileptogenic zone
Aim: Determine the feasibility of noninvasive T1 rho MR
imaging in idenBficaBon & lateralizaBon of epileptogenic zone
Inclusion criteria MR-‐posi:ve MTL epilepsy
i) PaBents with established MTL epilepsy by EEG, seizure semiology and MR proven MTS
ii) Unilateral disease iii) No history of epilepsy surgery iv) No other epileptogenic focus
Normal subjects
i) No known epilepsy or any structural lesion idenBfied on MR brain imaging
ii) No history of cerebral disease iii) No history of brain surgery
Included: 15 normal subjects; 7 pa_ents 2 pa_ents excluded (significant mo_on ar_facts & bilateral MTS)
Scanning parameters 3.0 Tesla MR scanner, uBlizing a 8-‐channel head coil.
T2 relaxometry: • Sequence: TSE; TR/TE (ms): 1868/20; FOV(mm): 240*240; Matrix: 268*268; Slice
thickness: 3 mm; Gap: 0; Scan Bme: 6 min 42 sec.
T1rho: • Sequence: B-‐TFE; FOV(mm): 240*240; Matrix: 160 160; During of spin-‐lock pulse
(ms): 1, 10, 20, 30, 40; Spin-‐lock frequency: 500 Hz; TI (ms): 860; Slice thickness: 3
mm; Bandwidth: 130 Hz/pixel; Echo train length: 4; Scan Bme: 9 min 10 sec.
3D T1-‐weighted MPRAGE: • Sequence: MPRAGE; TR/TE: 7.0/3.1 msec; Flip angle: 8; FOV (mm): 250*250; Matrix:
256*256; Slice thickness: 1 mm; Gap: 0.
ROIs defini:on – on T2W
Manual drawing of ROI to contour:
Amygdala Hippocampal head Hippocampal body Hippocampal tail
Verified against automated ROIs -‐comparable results -‐ no significant differences Manual ROI is accurate
ROIs Coregistra:on
To obtain the mean values & SD of: -‐T1 rho value -‐T2 relaxaBon Bme
T2W T2 Relaxometry T1rho
Asymmetric RaBo
Sta:s:cal Analysis
• Gaussian distribuBon and homogeneity tests • Paired t-‐test between leP and right side for each group
• StandardizaBon of T2 relaxometry and T1rho values according to the corresponding values of the normal control group through Z-‐score transformaBon:
z = (X -‐ μ) / σ
• Abnormal if: >2 SD away from the mean of the normal group: z > 2 or z < -‐2
p<0.05 will be considered as sta_s_cal significant.
Results – Normal Subjects
Right (mean±SD)
Lek (mean±SD)
Asymmetric Ra:o
SD
T2 relaxometry
Amygdala 95.1. ± 3.12 96.20 ± 3.14 0.9863 0.0122
Hippo Head 96.51 ± 3.31 97.12 ± 3.49 0.9841 0.0210
Hippo Body 90.79 ± 4.72 91.27 ± 3.92 0.9876 0.0085
Hippo Tail 86.69 ± 6.33 87.83 ± 5.59 0.9849 0.0124
T1 rho
Amygdala 144.89 ± 35.22 144.74 ± 36.39 0.9888 0.0073
Hippo Head 140.26 ± 35.56 139.69 ± 36.27 0.9896 0.0076
Hippo Body 133.58 ± 34.55 134.12 ± 33.75 0.9888 0.0085
Hippo Tail 133.97 ± 34.8 134.49 ± 34.61 0.9880 0.0060
Note: Asymmetry = Min(L, R) / Max(L, R) SD = standard deviaBon
The respec_ve asymmetric ra_o were then used as reference for comparison in pa_ents’ group mean +/-‐ 2SD
Parametric Maps of Normal Subject
F/22 yrs old; Lek MTS
M/ 45 yrs old; Lek MTS
Accuracy of T2 relaxometry & T1 rho results Comparison against Volumetry
Amyg Hipp Head Hipp Body Hipp Tail T2R T1rho Volume T2R T1rho Volume T2R T1rho Volume T2R T1rho Volume
IYY y y y y y NKY y y y y y y y y WYY y y y y y y Y y y y y
CYY y y y y y y y y CHY y y y y y y y y y y FKY y y y y y y y y y y y y CYSA y y y y y y y y Y y
T2 relaxometry: Sn = 60.9% (14/23); Sp = 100.0% (4/4) T1rho: Sn = 100.0% (24/24); Sp = 50.0% (2/42)
Distribu:on of Asymmetric Ra:os
F/7 yrs old; GTC seizure; MR-‐ve
Potential role in detecting WM changes
Limita:ons/ Improving work Recruit more subjects (paBents and normal) to further validate diagnos_c value of T1rho
Lack of histopathological correlaBon Perform DTI analysis to test the feasibility in detec_ng WM changes
Plane of imaging -‐ coronal 3D whole brain imaging techniques Use of longer spin lock Bmes
T1 rho MR Imaging
• Feasible in idenBficaBon of epileptogenic zone
• A sensi:ve marker more sensi_ve than T2 relaxometry more sensi_ve than volumetry
• Can potenBally detect early molecular changes
Conclusion • Wide variety of eBologies
MCD, MTS
• Concept of MR-‐negaBve epilepsy Does it really exist?
• Availability of various advanced MR imaging techniques + limitaBon
Feasibility in clinical prac_se?
• Promising result of T1rho imaging
Thank you