Imaging Data in the NEST: Stroke and the DAISI...
Transcript of Imaging Data in the NEST: Stroke and the DAISI...
DavidSLiebeskind,MD
Imaging Data in the NEST: Stroke and the DAISI CRN
Professor of NeurologyDirector, Neurovascular Imaging Research Core
Director, Outpatient Stroke and Neurovascular ProgramsAssociate Neurology Director, UCLA Stroke Center
Consultant to Stryker and Medtronic
October20,2017
Overview
Precision stroke imaging and real-world data
Imaging and angiography core lab perspective
Integration of imaging in registries & evidence generation
Precision stroke imaging
Prototype of stroke as ideal neurological disorder to cultivate precision medicine
Acute presentation
Systematic use of imaging characterization
Expansive dimensions of multimodal & serial imaging
Rapid interaction of therapy, pathophysiology, outcomes
Extensive data collection, utilizing common data elements
Defined infrastructure & multidisciplinary coordination
Common disorder across the globe, easier to study and generalizability
Typical sequence of imaging
CT (CT/CTA/CTP)
DSA
24 hr CT
AE
Local imaging & decision-making
ASPECTS
Occlusion
Go/no-go?
“TICI 3”
local TICI versus central…
bias for “quality” and financial implications
Imaging data to verify measures and time metrics
Core lab perspective - imaging and angiography
Differentiating centralized imaging versus core lab adjudication
History in acute ischemic stroke studies of endovascular therapy
Merci IDE
Trials: IMS III/TREVO EU/TREVO2/DAWN/RESILIENT
Registries: Multi MERCI/MERCI Registry/STAR/societal registries/STRATIS/Trevo Retriever Registry/numerous investigator-initiated international studies – ENDOSTROKE and many others
Others: SWIFT/SWIFT PRIME/HERMES (7)
VISTA-Endovascular/TREAT/Imaging Workgroup of NIH StrokeNet
No imaging – GWTG, Coverdell, NIH SPOTRIAS, SNIS NVQI, NIH StrokeNet
StrokeCloud & Million Brains Initiative™
Concerning core labs
Biases of internal core labs in current trials
Unfamiliarity, lack of infrastructure or experience by most CROs
Pitfalls of commercial core labs
Ideal framework of imaging & angiography core lab integration
Role as separate entity removed from clinical or other data
Blinded
Statistical analyses conducted by independent body
Systematic battery of imaging & angiography CDE
Data collection battery - standard CDE, hierarchies
Routine collection of all elements
Extraction of all time points from DICOM tags (utilized as verification of all key time intervals)
Lessons learned - acquisition, transfer, quality
Variability in angiographic hygiene
Technique/data acquisition varies
Transmittal
Least burdensome approach at sites
Costs - routinely acquired imaging, part of data
Optimized models for multicenter endovascular studies
© DSL
Innovative imaging and interventional registries
rationale for innovative, imaging-intensive interventional registries as pivotal step in realizing precision medicine for cerebrovascular disorders
enhanced registries may serve as a model for expansion of our translational research pipeline to fully leverage the role of phase IV investigations
scope and role of registries in precision medicine
review on the history of stroke and interventional registries
data considerations
critiques or barriers to such initiatives
potential modernization of registry methods into efficient, searchable, imaging-intensive resources
simultaneously offer clinical, research and educational added value
Registries and the DAISI CRN
Need generalizability on a very large scale
Imaging triage and management - OH or MSU and beyond
What is most valuable data variable in a clinical study? Added value?
Imaging and angiography at top of hierarchy of list – before novel markers, prehospital, discharge or rehab variables or economic costs
Imaging before genomic infatuation
Quality of data is essential
Digitally preserved, can be verified, automated, must have expertise
Value of imaging
Imaging routinely acquired
Extension of clinical examination
Real-time and real-world impact
Currently, siloed and inefficient
Pivotal in decision-making
Reflects pathophysiology of effectiveness and safety
Data hierarchy - ASPECTS, collaterals, TICI, FIV over age, NIHSS
Embedded in all trials, digital perseverance
Unique verification capabilities
Failure of omission
Future of big data and precision medicine with stroke imaging
100,000s
Global sites on all continents
StrokeCloud model of anonymized and encrypted CDE of imaging data (baseline, angiography, post)
Automated eASPECTS, RAPID/Olea, CTA/MRA CFD modules, PerfAngio
Machine learning with oversight by expert readers with long track record of experience
Conclusions
Imaging data - data are king, imaging>clinical, need imaging and angiography to characterize patient pathophysiology and specific treatments beyond blackbox description of endovascular therapy
Technology potential exists now, for least burdensome approach, most efficient model, learn maximal amount regarding effectiveness and adverse events (look at modifying variables), learn early why trials fail, how to redirect without abandoning large-scale research studies
Simple linking of existing registries is fruitless and stroke imaging complexity is unlike cardiac or vascular registries
Real-time and real-world
Separation for scientific integrity
Data quality and verification
Investment - examples of how relative expense is trivial
Failure to integrate imaging and angiography ignores pathophysiology of stroke and real-world practice, will culminate in guaranteed failure in research progress from earliest trials to registries