Arterial Spin Labelling: Non-Invasive Measurement of Perfusion Michael A

Arterial Spin Labelling: Non-invasive measurement of perfusion Michael A. Chappell [email protected] www.ibme.ox.ac.uk/QuBIc

Institute of Biomedical Engineering & Wellcome Centre for Integrative Neuroimaging University of Oxford. Perfusion • Perfusion is a measurement of delivery of blood to capillary bed ➡ Related to nutrient delivery to cells and waste removal. ➡ Altered by task activity. ➡ Changes in disease. • Quantity of blood delivered per unit of tissue per unit of time ➡ ml blood / 100g tissue / min ➡ (Dimensions of [T]-1) ➡ Grey matter ‘magic’ number: 60 ml/100g/min • Cerebral Blood Flow (CBF) is a misleading name! • To image perfusion we need a tracer ➡ ASL uses blood-water as an endogenous tracer.

Arterial Spin Labelling : M.A. Chappell Perfusion

Why use ASL? • 50 ➡ A direct measure of perfusion changes -

physiological response. 40 ➡ (Potentially) fully quantitative - possible to 30 calculate absolute perfusion. ml/100g/min

➡ Good for low frequency or ‘one-off’ 20 designs. ➡ Large ‘effect size’.

• What are the challenges? ➡ SNR ➡ Temporal sampling - TR and the need for label and control scans.

Arterial Spin Labelling : M.A. Chappell Perfusion • ASL is not BOLD! ➡ CBF change is a component of the BOLD signal. ➡ ASL can make absolute measurements under different conditions. ➡ You DONT need a interleaved design with ASL. FMRI Experiments ➡ ‘Rest’ and ‘task’ don’t even need to be in the same session. Stimulus

• ASL and BOLD can be combined A B A B A ➡ Dual (mutli-) echo ASL/BOLD (baseline) (stimulus) (baseline) (stimulus) (baseline)

e.g. flashing 0 1 2 3 4 5 chequerboard time (TRs) • Simple paradigm design:Arterial Spin Labelling : M.A. Chappell - stimulus vs baseline - constant stimulus “intensity” - constant block lengths - many repetitions: ABABA • Need baseline (rest) condition to measure change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°'+61#706A'%67'#6+'>SX°'+61#706A.'A0K06A'%'9#9%,'#I'-S'+61#706A'1J1,+&M''=+&&+,>+61#7+7'4+:I?&0#6'0$%A06A'O%&'4+:I#:$+7'?&06A'%' &06A,+>&"#9'+1"#'4,%6%:'0$%A06A'BC8QH':+%7#?9'%67'9"+'I#,,#O06A'4%:%$+9+:&]'!C'-S$&.'!\'RMRX&.'^_`9"&'4%:90%,'Y#?:0+:.'9%A'7?:%90#6'-MS&.'4#&9>,%@+,06A'7+,%J' -&.'K#T+,'&0<+'RMSTRMSTRMR$$.'5Z'&,01+&.'-SZ'K#,?$+&.'%1U?0&090#6'90Perfusion$+'a]XZM''8:+&%9?:%90#6'%67'9O#'A,#@%,'"J4+:@#,01'&+1%69'06K+:&0#6'4?,&+&'O+:+'?&+7'I#:' @%1*A:#?67' &?44:+&&0#6' F^GM' ' =+&&+,>+61#7+7' 7J6%$01' 5[' %6A0#A:%4"J' B&0$0,%:' 9#' F5GH' O%&' 4+:I#:$+7' 06' %' 9"01*' 9:%6&K+:&+' #@,0U?+' &,01+' ?&06A' 07+6901%,' +61#706A'1J1,+&'@?9'O09"'%'1%:70%1'A%9+7'90$+>:+&#,K+7'@%,%61+7'&9+%7J>&9%9+'I:++'4:+1+&&0#6'B@DDY8H':+%7#?9'FaG'O09"'9"+'I#,,#O06A'4%:%$+9+:&]'!C'5$&.'!\' SMRX$&.'I,04'%6A,+'RZ• But ASL° .'K#T+,'&0<+'-MZT-MZTXZ$$.'^_`9"&'4%:90%,'Yisn’t as funky as rs-fMRI/tractography/etc#?:0+:.'b\2882'I%19#:'5.'-S'&+A$+69&.'9+ - where is the (spurious)$4#:%,':+&#,?90#6'^ZMc$&.'-5'1%:70%1'4"%&+&. colour coding? '9%A' 7?:%90#6'`ZZ$&.'%1U?0&090#6'90$+'d5]XZM'D+4%:%90#6'#I'K%&1?,%:'1#$4#6+69&'06'9"+'=C832DE'7%9%'O%&'4+:I#:$+7'?&06A'%'$%T0$?$'0!/29+),52,5!&#,?90#6'F`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rterial Spin Labelling : M.A. Chappell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c`'B5ZZXH' F-G'(#6A.'/\/'X`]'-Z`^>-Zc-'B5ZZaH'' FSG'g%6&%A:%.'QD/\/'B5ZZcH]'fR^X5' FaG')*+,,.'QD/\/'B5Z--H]'fSZRS'' F5G')*+,,.'/\/'^S]'^cèaZ^'B5Z-ZH' FXG'3"%44+,,.'/\/'^S]-X5c>Rc'B5Z-ZH' F`G'3"%44+,,.'QD/\/'B5Z--H]'fR^^' ' FSL for Arterial Spin Labelling • BASIL: a toolset for resting ASL quantification: ➡ CBF quantification. ➡ Calibration / M0 estimation ➡ Registration. ➡ Partial volume correction.

➡ Command line tools oxford_asl, basil, asl_reg, asl_calib ➡ Graphical User Interface asl_gui

• Make sure you are using FSL v 6.0.1 at home!

Arterial Spin Labelling : M.A. Chappell What Have I Got Here!? • What I have...

• What I want...

• What should I do? I just want to do something simple/easy!

I must have absolute perfusion (ml/100g/min)

Command line instructions here for future reference...

Arterial Spin Labelling : M.A. Chappell Outline • Acquisition • Keep it simple! ➡ Perfusion weighted images. • Quantitative perfusion: ➡ Kinetics: A short course in tracer kinetics. ➡ Calibration: Measuring arterial blood magnetization. • Preparing for group analysis. • Advanced quantiﬁcation: ➡ Distortion Correction ➡ Macro vascular contamination ➡ Partial Volume Correction

Arterial Spin Labelling : M.A. Chappell Arterial Spin Labelling • A tracer experiment with an endogenous tracer - blood water. LABEL CONTROL

Aquire image of brain Wait for blood to reach brain Label blood by magnetic inversion

Arterial Spin Labelling : M.A. Chappell ASL Acquisition • Spot the difference? LABEL CONTROL

Perfusion is ~60 ml/100g/min = 0.01 s-1 Signal is ~ 1-2% Arterial Spin Labelling : M.A. Chappell ASL Acquisition • Nuts & bolts: Labelling cASL: Continuous ASL pASL: Pulsed ASL pcASL: psuedo-continous ASL

Label blood by magnetic inversion

Label a region in a single Label blood ﬂowing through a pulse plane for some time pcASL uses pulses and is more widely available

Arterial Spin Labelling : M.A. Chappell ASL Acquisition

Nuts & bolts: Inﬂow time • pcASL

Post-labeling imaging delay (PLD) cASL label Wait for blood to time reach brain Label blood by pASL

magnetic inversion pASL label imaging

time Inversion time (TI)

Arterial Spin Labelling : M.A. Chappell ASL Acquisition

Nuts & bolts: Bolus/label duration • pcASL

Label duration (τ) Post-labeling imaging delay (PLD) cASL label Wait for blood to time reach brain Label blood by pASL

magnetic inversion pASL label QUIPSS II imaging TI1 pASL time • Label duration is undeﬁned in pASL. Inversion time (TI) • QUIPSSII pulses ‘cut off’ the end of the labeled bolus. Arterial Spin Labelling : M.A. Chappell ASL Acquisition • Nuts & Bolts: Readout 3D: GRASE/RARE Higher SNR Long echo-train: Aquire image blurring of brain Muti-shot/segmented approaches Wait for blood to reach brain 2D: EPI (Multi-slice) Label blood by Different PLD for magnetic inversion each slice

Arterial Spin Labelling : M.A. Chappell ASL Acquisition

Nuts & Bolts: Background Suppression • pcASL

Background suppression pulse(s) imaging

cASL label Wait for blood to time reach brain Label blood by ➡ Suppress signal from static tissue magnetic inversion ➡ Reduce subtraction artefacts ➡ Reduce sensitivity to motion and physiological noise

Arterial Spin Labelling : M.A. Chappell ASL Acquisition • The ASL ‘white paper’ - a good place to begin: ➡ Use pcASL where possible ➡ Ideally 3D readout. Label duration 1800 ms 2D EPI an acceptable alternative. Post labeling delay ~1800 ms ➡ Resolution: ➡ Otherwise pASL with QUIPSSII 3-4 mm in plane. Inversion time ~1800 ms 4-8 mm through plane. TI1 of 800 ms Slab thickness 15-20 cm ➡ Use background suppression.

Recommended Implementation of Arterial Spin Labeled Perfusion MRI for Clinical Applications: A consensus of the ISMRM Perfusion Study Group and the European Consortium for ASL in Dementia Magnetic Resonance in Medicine - 73 (1) p102-116, 2015. Arterial Spin Labelling : M.A. Chappell Outline • Acquisition • Keep it simple! ➡ Perfusion weighted images. • Quantitative perfusion: ➡ Kinetics: A short course in tracer kinetics. ➡ Calibration: Measuring arterial blood magnetization. • Preparing for group analysis. • Advanced quantiﬁcation: ➡ Distortion Correction ➡ Macro vascular contamination ➡ Partial Volume Correction

Arterial Spin Labelling : M.A. Chappell Example (simple) Perfusion What I have... Control Tag (labelled) • (difference) ➡ ASL data!

• What I want... ➡ A perfusion image (in this - = subject). • What should I do? ➡ Label-control subtraction Tissue + blood Tissue - blood 2 * blood ➡ Average

asl_file --data={ASLdata.nii.gz} --ntis=1 --iaf=tc --diff --out={diffdata.nii.gz} asl_file --data={ASLdata.nii.gz} --ntis=1 --iaf=tc --diff --mean={diffdata_mean.nii.gz}

Arterial Spin Labelling : M.A. Chappell Example • What I have... ➡ ASL data ➡ (calibration images) • What I want... - = ➡ Perfusion in ml/100g/min

• What should I do? ➡ Label-control subtraction. ✓ ➡ Kinetic model inversion. ← ➡ Calibration

Arterial Spin Labelling : M.A. Chappell Introduce Kinetic Model Inversion tracer Tissue (voxel)

Arterial Input Residue Function Function

ΔM(t) = F ⋅ AIF(t) * r(t)

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion

Tells us about the Arterial Input delivery of the Function tracer Tracer concn

Time

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion LABEL Tissue T1 decay (voxel) Parameters: Arterial Transit Time Label duration T1 decay AIF

pcASL Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion

Residue Function Tells us what happens to the tracer after is 100% has arrived.

Tracer Remaining

‘time’ Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion

‘Well mixed’ T1 decay Parameters: ‣ Rapid exchange r(t) Arterial Transit Time ‣ Single well mixed Label duration compartment T1 decay ‣ No spins leave the compartment ‣ Decay with T1 ‘Stays & decays’

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion LABEL

T1 decay ‘Well mixed’ Parameters: Perfusion - F Arterial Transit Time ∆M(t) AIF r(t) Label duration T1 decay * =

ΔM(t) = F ⋅ AIF(t) * r(t)

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion

Arterial Input Residue Function Function Concentration time curve

time

Arterialtime Spin Labelling : M.A. Chappell Kinetic Model Inversion LABEL

T1 decay ‘Well mixed’ Parameters: Perfusion - F Arterial Transit Time ∆M(t) AIF r(t) Label duration T1 decay * =

ΔM(t) = F ⋅ AIF(t) * r(t)

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion LABEL T decay 1b ‘Well mixed’

T1t decay pASL pcASL 2 ∆t 2

1.5 0.3 1.5 The ‘simple’ model 0.5 • 0.7 M(t) 1 0.9 1 ∆ 1.1 ➡ Only one T1 value (blood) 0.5 0.5 0 0 0123401234 ➡ Spins never leave tissue r(t) 2 τ 2 1.5 0.6 1.5 0.8 1

M(t) 1 1.2 1 The ‘standard’ model: ∆ 1.4 • 0.5 0.5 0 0 0123401234 ➡ Separate T1 for blood and tissue T 2 1 2 1.5 1.2 1.5 (T1t < T1b). 1.4 1.6

M(t) 1 1.8 1 ∆ 2 ➡ Spins leave voxel at rate 0.5 0.5 0 0 determined by perfusion and 0123401234 t (sec) t (sec) partition coefﬁcient. Arterial Spin Labelling : M.A. Chappell Example • What I have... What you need to know about your data: ➡ ASL data pASL (pulsed) or pcASL (continuous) ➡ (calibration images) Inversion time(s) • What I want... Post-labeling delay(s) ➡ Perfusion in ml/100g/min Bolus duration Label duration Labeling (if QUIPSS/Q2TIPS) • What should I do? ➡ Label-control subtraction. 3D/2D (slice timing)

✓ out ➡ Kinetic model inversion.← Read ➡ Calibration. T1 (tissue and blood) Arterial Transit Time Model oxford_asl -i {asl_data} -o {output_dir} --iaf={tc} {--casl} --tis={list_of_TIs} —bolus={bolus_duration} —slicedt={time_per_slice} {model/analysis options}

Arterial Spin Labelling : M.A. Chappell Example

What I have... pcASL with Assume • labeling duration: 1.8 s ‘white paper’ ➡ ASL data post-label delay: 1.8 s T1 : 1.65 s ➡ (calibration images) 2D readout ATT : 0 s • What I want... 45.2 ms per slice ➡ Perfusion in ml/100g/min

• What should I do? ➡ Label-control subtraction. ✓ ➡ Kinetic model inversion.← ➡ Calibration

Arterial Spin Labelling : M.A. Chappell Example

pcASL with labeling duration: 1.8 s post-label delay: 1.8 s 2D readout 45.2 ms per slice Assume ‘white paper’ T1 : 1.65 s ATT : 0 s

#Do label control subtraction > asl_file --data={ASLdata.nii.gz} --ntis=1 --iaf=tc --diff --out={asldiffdata.nii.gz} \ --mean={asldiffdata_mean.nii.gz}

Arterial Spin Labelling : M.A. Chappell Example

pcASL with labeling duration: 1.8 s post-label delay: 1.8 s 2D readout 45.2 ms per slice Assume ‘white paper’ T1 : 1.65 s ATT : 0 s

#Do label control subtraction > asl_file --data={ASLdata.nii.gz} --ntis=1 --iaf=tc --diff --out={asldiffdata.nii.gz} \ --mean={asldiffdata_mean.nii.gz}

Arterial Spin Labelling : M.A. Chappell Example

• What should I do? ➡ Label-control subtraction. ✓ ➡ Kinetic model inversion.← ➡ M0 calculation.

#Do label control subtraction asl_file --data={ASLdata.nii.gz} --ntis=1 --iaf=tc --diff --out={asldiffdata.nii.gz} \ --mean={asldiffdata_mean.nii.gz}

Arterial Spin Labelling : M.A. Chappell Example

pcASL with labeling duration: 1.8 s post-label delay: 1.8 s 2D readout 45.2 ms per slice Assume ‘white paper’ T1 : 1.6 s ATT : 0 s

Do motion correction

# Do the analysis using oxford_asl > oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --casl --tis=3.6 --bolus=1.8 / --slicedt=0.0452 --wp --mc

Arterial Spin Labelling : M.A. Chappell Example Perfusion (arbitrary units)

oxasl/native_space/perfusion.nii.gz

Arterial Spin Labelling : M.A. Chappell Example • What I have... ➡ ASL data ➡ (calibration images) • What I want... - = ➡ Perfusion in ml/100g/min

• What should I do? ➡ Tag-control subtraction. ✓ ➡ Kinetic model inversion. ✓ ➡ Calibration ←

Arterial Spin Labelling : M.A. Chappell Calibration

LABEL ‘Concentration’ of T1 decay ‘Well mixed’ the label M0a -M0a Magnetization of arterial blood M0a Imperfect inversion ∆M(t) AIF r(t) Inversion efﬁciency α

∝ 2⋅α⋅M0a * =

ΔM(t) = 2⋅α⋅M0a⋅F ⋅ AIF(t) * r(t)

Arterial Spin Labelling : M.A. Chappell Calibration • Cannot measure M0a Calibration image: directly. ➡ Proton Density weighted • indirect via brain ‘tissue’ ➡ ‘Long’ TR: > 5 seconds magnetization. ➡ No labelling or background ➡ Calculate M0t. suppression (M0 of ‘tissue’) ➡ M0t to M0a. Account for relative proton densities: M M = 0t 0a λ Partition co-efﬁcient λ (relative concentration of water) oxford_asl ... -c {calibration_image.nii.gz} --tr={TR}

Arterial Spin Labelling : M.A. Chappell Calibration

• Cannot measure M0a Perfusion (ml/100g/min) = (Perfusion / M ) × 6000 directly. 0a • indirect via brain ‘tissue’ magnetization. Voxelwise Calibration ➡ Calculate M0t. (M0 of ‘tissue’) ➡ M0t to M0a. = ÷ × 6000 • Practicalities ➡ Voxelwise

• Cannot measure M0a Perfusion (ml/100g/min) = (Perfusion / M ) × 6000 directly. 0a • indirect via brain ‘tissue’ magnetization. Reference Tissue ➡ Calculate M0t. CSF or WM (M0 of ‘tissue’) ➡ M0t to M0a. • Practicalities ➡ Reference Tissue = ÷ M0a × 6000

oxford_asl ... -c {calibration_image.nii.gz} --tr=[TR] asl_calib --mode longtr ... asl_calib --mode satrecov ... fslmaths {perfusion.nii.gz} -div [M0a] -mul 6000 {perfusion_calib.nii.gz} Arterial Spin Labelling : M.A. Chappell Calibration • Cannot measure M0a Voxelwise Reference ‘tissue’ directly. Calculate M0t • indirect via brain ‘tissue’ magnetization. Reference tissue mask (CSF or WM) ➡ Calculate M0t. (M0 of ‘tissue’) M0t → M0a ➡ M0t to M0a. voxelwise M0a value Single global M0a value Practicalities • (coil sensitivity correction) ➡ Reference ‘tissue’? ➡ Voxelwise? Perfusion (ml/100g/min) = (Perfusion / M0a) * 6000 oxford_asl ... -c {calibration_image.nii.gz} --tr=[TR]

Arterial Spin Labelling : M.A. Chappell Example Assume What I have... pcASL with • labeling duration: 1.8 s T1 (blood) : 1.6 s ➡ ASL data post-label delay: 1.8 s T1 (tissue) :1.3 s ➡ (calibration images) 2D readout ATT : 1.3 s • What I want... 45.2 ms per slice α : 0.85 ➡ Perfusion in ml/100g/min

• What should I do? ➡ Label-control subtraction. ✓ ➡ Kinetic model inversion.← ➡ Calibration

Arterial Spin Labelling : M.A. Chappell Example Calibration image No background-suppression TR: 4.8 s

✓ ←

Arterial Spin Labelling : M.A. Chappell Example Calibration image TR: 4.8 s

Calibration mode Voxelwise

# Do the analysis using oxford_asl > oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --casl --tis=3.6 --bolus=1.8 / --slicedt=0.0452 --wp --mc —c {calibration_image.nii.gz} --tr=4.8

Arterial Spin Labelling : M.A. Chappell Example Perfusion (ml/100g/min)

Correct for ‘edge effects’ (and distortion)

oxasl/native_space/perfusion_calib.nii.gz

Arterial Spin Labelling : M.A. Chappell Example

Calibration image TR: 4.8 s

Calibration mode Reference region CSF (ventricles)

Calibration mask (derived automatically from structural image)

# Do the analysis using oxford_asl > oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --casl --tis=3.6 --bolus=1.8 / --slicedt=0.0452 --wp --mc —c {calibration_image.nii.gz} --tr=4.8 / --fslanat=T1.anat Arterial Spin Labelling : M.A. Chappell Example Perfusion (ml/100g/min) Voxelwise Reference region Ventricular mask (automatically generated)

oxasl/native_space/perfusion_calib.nii.gz

Arterial Spin Labelling : M.A. Chappell 1. The entire labeled bolus is delivered to the target tissue. This is the case when PLD>ATT for PCASL, or (TI-TI1)>ATT for QUIPSS II PASL. 2. Labeled blood water is well mixed with tissue before outflow occurs Because the tissue water pool is much larger than the blood water pool, and water exchange between blood and tissue is rapid, this is typically a good assumption (Zhou J , Wilson D, et al. JCBFM, 21:440, 2001).

3. The relaxation of the labeled spins are governed by blood T1. While this assumption is not likely to be strictly true, the errors introduced by this assumption, whichSummary are related to the difference in T1 between blood and tissue, are typically relatively small. The ASL ‘white paper’ quantiﬁcation formula (pcASL): • Under these assumptions CBF in each voxel can be calculated for PCASL using (10): !"# Subtraction !!,!"##$ !"""⋅!⋅(!"!"#$%!"!!"!"#$!)⋅! CBF = ! ![ml/100!g/min] ! !!,!"##$ Kinetic Model Inversion !⋅!⋅!!,!"##$⋅!"!"⋅(!!! ) and for QUIPSS II PASL using (37): M0 Calculation (Calibration) !" !"""⋅!⋅(!" !!" )⋅!!!,!"##$ Values:CBF = !"#$%"& Assumptions:!"#$! ![ml/100!g/min] T1blood = 1650 ms (3T) !⋅!⋅!"!⋅Voxelwise!"!" calibration (M0t = SIPD) α =where 0.85 λ is the brain/bloodT1tissue partition = T1blood coefficient in ml/g, SIcontrol and SIlabel are the time- λ = 0.9 ml/g ATT = 0 averaged signal intensities in the control and tag images respectively, T1, blood is the longitudinal relaxation time of blood in seconds, α is the labeling efficiency, SI is the Recommended Implementation of Arterial Spin Labeled Perfusion MRI for PD Clinical Applications:signal intensity A consensus of a ofproton the ISMRM density Perfusion weighted Study Group image, and and τ is the label duration. PLD, TI the European Consortium for ASL in Dementia and TIMagnetic1 are Resonance as defined in Medicine above. - 73 (1) p102-116, The factor 2015. of 6000 converts the units from ml/g/s to ml/(100g)/min, which is customary in the physiologicalArterial Spin literature. Labelling : M.A. ChappellSee Table 3 for a summary of parameters for use in CBF quantification. Single TI PASL without the QUIPSS II modification cannot be reliably converted into CBF.

To scale the signal intensities of the subtracted ASL-images to absolute CBF units, the signal intensity of fully relaxed blood spins is needed. Although several approaches can yield estimates of this value, we recommend using a separately acquired proton density

(PD) image (represented by SIPD in the above equations) to obtain this scaling factor on a voxel-by-voxel basis. The factor λ scales the signal intensity of tissue to that of blood. The use of a PD image for this scaling serves two additional important functions. By dividing by this image, signal variations caused by RF coil inhomogeneity, as well as differences in transverse relaxation are largely corrected as well. The PD image should have an identical readout module as the ASL-scan, with a long TR to provide proton density weighting. If TR is less than 5s, the PD image should be multiplied by the factor !!"/! 1/ 1 − e !,!"##$% , where T1,tissue is the assumed T1 of gray matter, in order to

compensate for T1 relaxation. No labeling or BS should be applied for this scan. Care should be taken that the absolute scaling between the signal intensities in this acquisition and the ASL scans is known. Note that since this PD image goes in the Arterial Spin Labelling: Non-invasive measurement of perfusion Michael A. Chappell [email protected] www.ibme.ox.ac.uk/QuBIc

Institute of Biomedical Engineering & Wellcome Centre for Integrative Neuroimaging University of Oxford. EXAMPLE • What I have... pcASL with ➡ ASL data - multi-PLD label duration: 1.4 s ➡ (calibration images) post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s • What I want... ➡ Perfusion in ml/100g/min • What should I do? ➡ Label-control subtraction. ➡ Kinetic model inversion. TI: 1.65 1.9 2.15 2.4 2.65 2.9 ➡ Calibration

# Label control subtraction for each PLD individually > asl_file --data={ASLdata.nii.gz} --ntis=6 --iaf=tc --ibf=rpt --diff --split \ --mean={asldiffdata_mean_at_each_PLD.nii.gz}

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion LABEL

T1 decay ‘Well mixed’ Parameters: Perfusion - F Arterial Transit Time ∆M(t) AIF r(t) Label duration T1 decay (in blood) * = T1 decay (in tissue)

ΔM(t) = F ⋅ AIF(t) * r(t)

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion

Parameters: Model Data Perfusion - F white noise Arterial Transit Time Label duration ⊕ T1tissue T1blood

Single-TI/PLD Multi-TI/PLD Analytic solution Non-linear ﬁtting (least squares)

Bayesian inference (BASIL) Chappell et al., IEEE TSP 57(1), 2009. Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion • Perfusion ➡ Want to know this - variable • Arterial Transit Time ➡ Want to correct for this - variable Variability but limited to a sensible range • Label duration ➡ Set by sequence - fixed (might not be that well fixed, pASL?) • T1 tissue ➡ 1.3 s at 3T - fixed

• T1 blood Doesn't T1 vary a bit? ➡ 1.65 at 3T - ﬁxed

Arterial Spin Labelling : M.A. Chappell Kinetic Model Inversion Priors: Spatial prior: Spatial Prior distribution for perfusion in Perfusion voxel deﬁned over its neighbours

Arterial Transit Time σ 1.3 s

1.3 s

σ - spatial scale of prior (determined from the data)

Arterial Spin Labelling : M.A. Chappell EXAMPLE • What I have... pcASL with ➡ ASL data - multi-TI/PLD labeling duration: 1.4 s ➡ (calibration images) post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s • What I want... ➡ Perfusion in ml/100g/min • What should I do? ➡ Label-control subtraction. ➡ Kinetic model inversion. TI: 1.65 1.9 2.15 2.4 2.65 2.9 ➡ Calibration.

# Label control subtraction for each PLD individually > asl_file --data={ASLdata.nii.gz} --ntis=6 --iaf=tc --ibf=rpt --diff --split \ --mean={asldiffdata_mean_at_each_PLD.nii.gz}

Arterial Spin Labelling : M.A. Chappell Example

pcASL with labeling duration: 1.4 s post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s

> oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --ibf=rpt --casl \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 --bolus=1.4 --slicedt=0.0452 \ --fixbolus --artoff --mc \ -c {calibration_image.nii.gz} --tr=4.8 Arterial Spin Labelling : M.A. Chappell Example Data: RatioPerfusion (ml/100g/min) Arrival (seconds) • (multi/single) Multi-TI pcASL Single-TI Multi-TI ➡ Single-PLD label duration: 1.8 s post-label delay: 1.8 s Assume ATT of 1.3 s

➡ Multi-PLD label duration: 1.4 s PLDs: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s oxasl/native_space/perfusion_calib.nii.gz oxasl/native_space/arrival.nii.gz

Arterial Spin Labelling : M.A. Chappell Single- vs. Multi-PLD • Which is better in a ﬁxed scan duration? Single-PLD ∆M(t)

x6 ✹

Multi-PLD ∆M(t) ✹ ✹ ✹ ✹ ✹ Woods et al. MRM ✹ 2018 in press Arterial Spin Labelling : M.A. Chappell Outline • Acquisition • Keep it simple! ➡ Perfusion weighted images. • Quantitative perfusion: ➡ Kinetics: A short course in tracer kinetics. ➡ Calibration: Measuring arterial blood magnetization. • Preparing for group analysis. • Advanced quantiﬁcation: ➡ Distortion Correction ➡ Macro vascular contamination ➡ Partial Volume Correction

Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis • Group analysis and quantitative comparisons between individuals requires consistent representation • Consistent geometry: ➡ ‘Spatial’ normalization (registration) ➡ Transform perfusion map to a common space, e.g. MNI152

• Consistent intensity: ➡ Quantitive maps - perfusion in ml/100g/min. ➡ Intensity normalization to a reference.

Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis • Registration to ‘standard’ space ➡ ASL → Structural linear - 6 DOF ➡ Structural → Standard linear - 12 DOF non-linear

oxford_asl ... --s {structural_image.nii.gz}

See also: asl_reg, flirt, fnirt

Arterial Spin Labelling : M.A. Chappell Example Run fsl_anat on structural image

BASIL will then do registration and transformation to: structural space standard space

> fsl_anat {T1.nii.gz} > oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --casl --tis=3.6 --bolus=1.8 / --slicedt=0.0452 --wp --mc —c {calibration_image.nii.gz} --tr=4.8 / --fslanat=T1.anat Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis • Quantitative maps • Intensity normalization: ➡ requires estimate of M0a - ‘calibration’ ➡ requires a ‘reference’. data. e.g. a brain structure: thalamus • Pros: e.g. a ‘global’ value: mean in GM or WM ➡ An absolute scale - can potentially Pros: relate to physiology • ➡ No need for calibration. ➡ Ought to be able to set consistent ➡ Removes inter subject variability in thresholds ‘global’ perfusion. e.g. perfusion < 20 ml/100g/min is ischaemia Cons: Cons: • • ➡ Relies on a consistent reference. ➡ Requires calibration information. ➡ Loose information on ‘global’ perfusion ➡ Global perfusion appears to be quite changes. variable between individuals.

Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis • Intensity normalization: • Transform ROI into perfusion space or vice versa? ➡ Pick a ROI: ➡ ROI in high-res → perfusion space Manually Interpolation on ROI mask: sharp boundaries in high-res From atlas become ‘soft’ requiring thresholding - possible bias. From a segmentation ➡ Perfusion image → high-res ➡ Calculate mean within ROI. Interpolation occurs on perfusion values, ROI untouched. ➡ Scale perfusion maps. • Exception is ‘soft’ segmentations e.g. GM/WM on a structural image. ➡ Transform ‘soft’ segmentations ﬁrst and THEN threshold to create ROI. oxford_asl ... --norm oxford_asl ... --report

Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis GM PVE Threshold Transform at 0.7

Threshold at 0.7 Threshold Transform at 0.7

High res GM mask Arterial Spin Labelling : M.A. Chappell Preparing for Group Analysis • ROI Absolute perfusion: ➡ GM / WM(?) A direct physiological measurement e.g. Asllani et al., JCBFM, 28, 2008. partial volume issues A consistent baseline (c.f BOLD) ➡ Structures e.g. Wang et. al, MRM, 49, 2003. • Voxelwise Inter subject and inter session variability e.g Gevers et al., JCBFM, 31, 2011. Petersen et al., NeuroImage, 49(1), 2011. • Designs ➡ Group mean Arterial Transit Time (multi-TI/PLD): ➡ Group differences/paired Potential confound differences An extra quantitative measurement e.g. Bokkers et al., AJNR, 29(9), 2008. MacIntosh et al, AJNR, 33(10), 2012. Feat (higher-level analysis) Randomise

Arterial Spin Labelling : M.A. Chappell EXAMPLE • What I have... pcASL with ➡ ASL perfusion in multiple labeling duration: 1.4 s sessions/subjects post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s ➡ Structural images • What I want... ➡ Perfusion change/ difference (in ml/100g/min) • What should I do? ➡ Registration. ➡ GLM

> oxford_asl -i {ASLdata.nii.gz} -o {oxasl} —iaf=tc --ibf=rpt --casl \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 --bolus=1.4 --slicedt=0.0452 \ --fixbolus --artoff --mc --fslanat=T1.anat\ -c {calibration_image.nii.gz} --tr=4.8 Arterial Spin Labelling : M.A. Chappell EXAMPLE • Data: pcASL, Multi-PLD label duration: 1.4 s PLDs: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s

8 individuals task - ﬁnger tapping and visual stimulation

• Paired t-test

> flameo --cope=perfusion_study.nii.gz \ --mask=${FSLDIR}/data/standard/MNI152_T1_2mm_brain_mask.nii.gz \ --dm=design.mat --tc=design.con --cs=design.grp --runmode=ols --ld=flameout

Arterial Spin Labelling : M.A. Chappell EXAMPLE • Data: pcASL, Multi-PLD Perfusion change (effect size) label duration: 1.4 s PLDs: 0.25, 0.5, 0.75,1.0, 15.0 1.25, 1.5 s

8 individuals 10.0 task - ﬁnger tapping and visual stimulation ml/100g/min 5.0 • Paired t-test

Arterial Spin Labelling : M.A. Chappell Example • Data: pcASL, Multi-PLD label duration: 1.4 s PLDs: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s

• Paradigm Monitoring response to painful stimulus

Segerdahl et al. Nat Neuroscience 2015 Arterial Spin Labelling : M.A. Chappell Outline • Acquisition • Keep it simple! ➡ Perfusion weighted images. • Quantitative perfusion: ➡ Kinetics: A short course in tracer kinetics. ➡ Calibration: Measuring arterial blood magnetization. • Preparing for group analysis. • Advanced quantiﬁcation: ➡ Distortion Correction ➡ Macro vascular contamination ➡ Partial Volume Correction

Arterial Spin Labelling : M.A. Chappell Advanced: Distortion Correction • EPI readout will include distortion in regions of ﬁeld inhomogeneity ➡ cf BOLD fMRI • Need to correct for: ➡ geometric distortion AP encoding PA encoding Corrected ➡ AND intensity • Need: ➡ ﬁeld map OR ➡ phase encode reversed image oxford_asl ... --cblip={ASL_calibration_phase_reversed} pedir=[direction] \ --echospacing=[value] oxford_asl ... --fmap={fieldmap_image} --fmapmag={fieldmap_magnitude_image} \ --fmapmagbrain={brain_extracted_fmapmag} --pedir=[direction] --echospacing=[value] Arterial Spin Labelling : M.A. Chappell Example

pcASL with labeling duration: 1.8 s post-label delay: 1.8 s 2D readout 45.2 ms per slice Calibration images TR: 4.8 s (1) AP encoding (2) PA encoding echo spacing (dwell time): 0.06 ms

# Do the analysis using oxford_asl > oxford_asl -i {ASLdata.nii.gz} -o {oxasl} --iaf=tc --casl --tis=3.6 --bolus=1.8 / --slicedt=0.0452 --wp --mc —c {calibration_image.nii.gz} --tr=4.8 / --cblip={calibration_PA.nii.gz} --pedir=y --echospacing=0.06 Arterial Spin Labelling : M.A. Chappell Example Perfusion (ml/100g/min)

oxasl/native_space/perfusion_calib.nii.gz

Arterial Spin Labelling : M.A. Chappell Advanced: Macro Vascular Contamination • Early PLDs may contain label still within larger arteries. ➡ perfusion overestimation

• Use long PLD(s) • Use ﬂow suppressing gradients • Include in model - multi- PLD data aBV and TOF MIP ➡ provides estimate of arterial blood volume oxford_asl: MV component included by default, use --artoff to turn off Ye et al., MRM 37(2), 1997. Chappell et al., MRM 63(5), 2010.

Arterial Spin Labelling : M.A. Chappell Advanced: Macro Vascular Contamination

ARD prior: ~N(0,v) ΔM(t) = CBF ΔM (t) + aBV ΔM (t) v determines the tiss IV relevance of the prior. v is determined from the Perfusion/ Spatial ARD data. Restrictive aBV v → 0 prior: Arterial parameter Transit Time forced to 1.3 s 0.8 s prior mean

T1 v → ∞

1.3 s 1.65 s

Liberal prior: parameter free to be estimated from data

Arterial Spin Labelling : M.A. Chappell EXAMPLE • What I have... pcASL with ➡ ASL data - multi-TI/PLD labeling duration: 1.4 s ➡ (calibration images) post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s • What I want... ➡ Perfusion in ml/100g/min ➡ Arterial blood volume in ml/ml. • What should I do? TI: 1.65 1.9 2.15 2.4 2.65 2.9 ➡ Tag-control subtraction. ➡ Kinetic model inversion. ➡ M0 calculation.

> oxford_asl -i {ASLdata.nii.gz} -o {oxasl} —iaf=tc --ibf=rpt --casl \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 —bolus=1.4 --slicedt=0.0452 \ --fixbolus --artoff --mc -c {calibration_image.nii.gz} --tr=4.8 Arterial Spin Labelling : M.A. Chappell Example Perfusion ml/100g/min Arterial cerebral blood volume % (ml/ml * 100)

middle slice lower slice ~ Circle of Willis oxasl/native_space/perfusion_calib.nii.gz oxasl/native_space/aCBV_calib.nii.gz

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Partial voluming of grey and white matter inevitable. • Leads to GM perfusion underestimation ➡ WM perfusion < GM ➡ WM blood arrival > GM

• Correction ➡ PV estimates from segmentation of structural image. Note: partial volume estimates NOT a hard segmentation or probabilities. ➡ Make separate GM and WM perfusion estimates in every voxel. An under determined problem.

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Does it matter that much? ➡ Resolution of ASL ~ 3 x 3 x 5 mm ➡ Cortical thickness ~ 2 - 4 mm • Unlikely to have many pure GM or WM voxels in the cortex Structural resolution ASL resolution Partial Volume Partial Volume Threshold at 90% Threshold at 90% Estimate Estimate

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • What do we mean when we report GM or WM perfusion? GM WM Whole brain Perfusion ml/100g/ min

GM mask threshold at Threshold on % PVE 90% oxford_asl ... --report

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Does it matter that much? ➡ Resolution of ASL ~ 3 x 3 x 5 mm ➡ Cortical thickness ~ 2 - 4 mm • What is this?

60 * PVEGM + 10 * PVEWM Estimated perfusion from ASL Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Partial volume correction exploiting kinetic data: T1 decay

GM WM ATT

➡ Perfusion: GM > WM ➡ ATT: WM > GM ➡ T1: WM < GM

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Multi-component model: ΔM(t) = PVGMΔMGM(t) + PVWMΔMWM(t) + PVCSFΔMCSF(t) + aBV ΔMMV(t) Grey matter White matter CSF Macro vasc.

ΔMCSF = 0

• Spatial priors on CBF for GM and WM

Arterial Spin Labelling : M.A. Chappell Advanced: Partial Volume Correction • Example from simulated data

Simulated true Conventional PVEc GM PVE perfusion perfusion map perfusion map

Chappell et al., MRM 65(4), 2011.

Arterial Spin Labelling : M.A. Chappell EXAMPLE • What I have... pcASL with ➡ ASL data - multi-TI/PLD labeling duration: 1.4 s ➡ (calibration images) post-label delays: 0.25, 0.5, 0.75,1.0, 1.25, 1.5 s • What I want... ➡ Grey matter perfusion in ml/100g/min • What should I do? ➡ Tag-control subtraction. TI: 1.65 1.9 2.15 2.4 2.65 2.9 ➡ Kinetic model inversion. Segmented structural image, e.g. fsl_anat output ➡ M0 calculation. ➡ Partial volume correction

> oxford_asl -i {ASLdata.nii.gz} -o {oxasl} —iaf=tc --ibf=rpt --casl \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 —bolus=1.4 --slicedt=0.0452 \ --fixbolus --mc --pvcorr --fslanat=T1.anat \ -c {calibration_image.nii.gz} --tr=4.8 Arterial Spin Labelling : M.A. Chappell EXAMPLE Perfusion (uncorrected) Grey matter perfusion White matter perfusion ml/100g/min ml/100g/min ml/100g/min

oxasl/native_space/perfusion_calib.nii.gz oxasl/native_space/pvcorr/perfusion_calib_masked.nii.gz oxasl/native_space/pvcorr/perfusion_wm_calib_masked.nii.gz

Arterial Spin Labelling : M.A. Chappell Need to know more…

Oxford Neuroimaging Primers: Introduction to Perfusion Quantiﬁcation using Arterial Spin Labelling ➡ Cover material in this lecture and more. ➡ http://www.neuroimagingprimers.org Examples using BASIL (extended from the course) FSL: The FMRIB Software Library ➡ BASIL: www.fmrib.ox.ac.uk/fsl/basil User guide & tutorials for FSL v6.0+ Follow the link for the ‘pre-release’ and updated user guide/tutorials Quantiphyse ➡ www.quantiphyse.org User guide & tutorials for ASL perfusion quantiﬁcation Arterial Spin Labelling : M.A. Chappell Acknowledgements • QuBIc, Engineering Science, Oxford ➡ Martin Craig • Brad MacIntosh (Univ. Toronto) ➡ Moss Zhao • Manus Donahue (Vanderbilt) ➡ Flora Kennedy McConnell • Xavier Golay (UCL, London) ➡ Tom Kirk • Esben Petersen (Utrecht) WIN/FMRIB, Oxford • Marco Castellaro (Padova) • Ilaria Boscolo Galazzo (Verona) ➡ Peter Jezzard • ➡ Tom Okell ➡ Joe Woods ➡ Michael Kelly ➡ James Meakin ➡ Matthew Webster ➡ Mark Jenkinson

Arterial Spin Labelling : M.A. Chappell