standardizing image acquisition jeffrey l. evelhoch, phd director medical sciences (imaging)

Standardizing Image Acquisition

Jeffrey L. Evelhoch, PhDDirector

Medical Sciences (Imaging)

2

Why standardize acquisition?

Best Image Processing

Ever

3

Issues to consider

•Manufacturer-based differences

•Technology changes

•Not all sites are created equal

•QC for quantitative v. diagnositic imaging

•Appropriate method depends on:– Exact question– Image analysis methods

•Who will ‘mind the shop’

4

An Example – Consensus Protocol•DCE-MRI

•Early phase clinical trials for drugs targeting tumor blood supply      

•Low molecular weight gadolinium chelates

•1.5 Tesla

5

A Brief History

•October 1999: NCI Workshop consensus recommendation– Included both treatment response and

breast diagnosis

•March 2002: CRUK PTAC Workshop consensus recommendation– Treatment response in early clinical trials

•Nov 2004: NCI Workshop consensus recommendation– Treatment response in early clinical trials

6

NCI Workshop Participants

• Jeffrey Abrams

• Thomas Chenevert

• Laurence Clarke

• Jerry Collins

• Jeffrey Evelhoch

• Susan Galbraith

• Michael Knopp

• Jason Koutcher

• Martin Leach

• Nina Mayr

• Daniel Sullivan

• Edward Ashton

• Peter Choyke

• Patricia Cole

• Gregory Curt

• Milind Dhamankar

• Michael Jacobs

• Gary Kelloff

• Adrian Knowles

• Lester Kwock

• Peter Martin

• Teresa McShane

• Anwar Padhani

• Stefan Roell

• Mark Rosen

• Gregory Sorensen

• Charles Springer

• Michael Tweedle

• Donald Williams

• Antonio Wolff

7

Specific recommendations for

• Type of measurement

• Requirements for contrast agent injection

• Primary endpoints

• Secondary endpoints

• Nomenclature

• Data reduction

• Region of interest

• Images acquired prior to contrast injection

• Dynamic acquisition protocol

• Measurement requirements for primary endpoints

• Trial design

• Image analysis

Let’s consider issues & approaches adopted to address those issues

9

Manufacturer-based differences

e.g., Sequence names

Sequence GE Philips SiemensGradient

echo GRASS FFE GRE

Spoiled Gradient

echoSPGR T1FFE FLASH

Inversion Recovery MPIR IR-TSE TurboIR

10

Approaches

•Need access to experts on all systems

•Use basic sequences (differences smaller)

•Tweak parameters to match as closely as possible

•Make certain reconstruction methods are as close as possible

•Same reference standard on all systems

11

Technology changes keep it simple•1.5-T (for near future)

•Basic sequences (less likely to change with upgrades)

•Parameters (at least) one step back from cutting edge

•Use analysis with considerable experience within community, most forgiving

•Revisit recommendations periodically

12

Not all sites are created equal •Keep it simple

•Training, training, training, …

•Routine QC

•Careful monitoring

•Immediate feedback

13

QC for quantitative v. diagnositic imaging•Requires higher performance

– MR system– Sequences– Set-up procedures– Radiological processes

•Better here or here?

14

Appropriate method

•For breast cancer diagnosis want dynamic temporal-spatial resolution trade-off

M. Schnall,MRM 2001

15

Who will ‘mind the shop’?

Summary of key recommendations for

data acquisition

17

General Issues

• Entry criteria should consider tumor size in relation to pharmacological mechanisms, MRI resolution, sensitivity to motion and potential confounding factors from previous treatment (e.g., radiation) or rapid tumor growth rates

• Tumors in a fixed superficial location should be at least 2 cm in diameter; other tumors should be at 3 cm in diameter

• Adjust orientation so that motion is in-plane when motion effects cannot be avoided (e.g., liver, lungs)

18

Pre-injection

•Acquire high quality clinical images of entire anatomic region (preferably in two orthogonal planes)

•Acquire T1- and T2-weighted images registered in the same planes as the dynamic data

• If possible, measure T1 (using same resolution and field of view for dynamic data)

19

Contrast agent injection

•Use power injector to minimize variation

• Injection dose should be standardized by weight

•15-30 sec for total injection, at least 20 cc saline flush

•Document injection site, use same site for subsequent studies in same subject

•Minimum of 24 h between studies

20

Dynamic study (I)

•For first 150 sec after bolus injection, use fastest sampling possible consistent with spatial resolution/anatomic coverage requirements, but not slower than 20 sec temporal resolution

21

Dynamic study (II)

•Acquire data out to at least 8 min (continual sampling is optional)

•If possible, include in imaging volume a normalization function (e.g., arterial or other tissue)

•For serial studies, imaging volume of interest should be adjusted to sample same region of tumor

22

Summary

• Development & application of appropriate protocol requires close academic-industry interaction

• For multi-center studies– Keep it as simple as possible while still

getting the required info– Clarity & education up front is essential– Non-standard methods require

substantial continuous support