Imaging in Clinical Trials

Past, Present and Future

John Reilly, MD
University of Pittsburgh
 Disclosures
• Funding from NHLBI and NCRR of NIH
• Chair of the Clinical Trials Study Section of NHLBI
• 2004-7 funding from Aeris Inc for a clinical trial of lung
volume reduction
• Paid speaker in American Thoracic Society Educational
programs
• Sadly, that is it.
 Roles of Imaging in Clinical Trials

• Eligibility
• Outcome
• “Ancillary”
 Examples of Past
• Trials/Studies using imaging as an endpoint or
disease marker
– Streptomycin for tuberculosis 1948-50’s
– Scoring system for pneumoconioses
• Trials/Studies of imaging
– Studies of screening CXR’s for lung cancer 1980’s
 Examples of Present

• Eligibility
●
NETT: emphysema on CT
●
FORTE: emphysema on CT
●
Oncology trials: “evaluable” disease on imaging
●
Pulmonary embolism: demonstration on CT angio
• Outcome
●
FORTE: change in lung density
●
Alpha 1 AT trials: change in lung density
 Examples of Present (2)

• Trials of imaging
– NLST: role of serial CT scans in lung cancer detection
– PET-CT in lung cancer
NEJM 2009; 361: 32-9
NEJM 2009; 361: 32-9
 Future?

• Technological Changes
• Challenges
• Opportunities
• Change in the economics and paradigm
– 10 years ago: $ genotyping >> $ phenotyping
– Now: $ phenotyping >> $ genotyping
Define specific subsets within a diagnosis (‘phenotypes’)

Reilly J. N Engl J Med 2008;359:1616-1618

 COPD and Lung Cancer

• Emphysema on CT scan is a marker for an increased risk

of lung cancer.
– Independent of lung function
• Current theories are that the microenvironment in the
lung is important in the pathogenesis and behavior of
lung cancer.
●
Does CT scanning provide an assessment of the
microenvironment?
 From the NCI Lung Cancer Integration and
Implementation (I2) Team Business Plan:

• 3.4 To advance the science of imaging response assessment with

molecular imaging technologies that directly reflect response to
targeted therapies and by providing uniform, high quality imaging
acquisition, quality control, and analysis within the conduct of
clinical trials. A multi-pronged approach is proposed:
– Support the creation and operation of Lung Imaging
Response Assessment Team/s (L-IRATs) to establish
common operating standards
– Establish a "Lung Cancer Meta-directory" to link lung
cancer-related image data with associated metadata,
 Response Evaluation Criteria in Solid Tumors
(RECIST) vs. Volumetric Assessment

• Single slice measurement vs. volumetric measurement

12 April 2010 / Vol. 18, No. 8 / OPTICS EXPRESS 8151

 Technology as a Driver
• CT scanning
– Multidetector scanners
●
Large number of thin slices allow volumetric
reconstruction
– Short scan time
• MRI
– Stronger magnets
– Contrast Agents
– Functional studies
 Technology as a Driver (2)
• Data Processing
– Moore’s Law
●
Ability to perform vast numbers of calculations quickly &
cheaply
●
Need for modeling of complex branching systems with
many generations
 Deriving Function from Form
• Provides the opportunity to perform computationally
intense modeling of fluid flow and particle deposition.
• Data to date show that individual variation in anatomic
geometry has substantial effects on distribution of
ventilation and perfusion.
• Beginning an era of developing “personalized modeling”
that may predict specific risk and/or sites of risk in an
individual patient based on his/her anatomy.
– Examples from systemic circulation
●
Carotid bifurcation
●
Abdominal aortic aneurysms.
 Challenges

• The limitations of physics and biology

●
Resolution
●
Radiation dose
• Technology/”Capitalism”:
– Market is for “images”, not data
– Proprietary acquisition and processing algorithms
• Result: lack of equivalency among platforms
– Scanner/site becomes a major variable in studies
 Averaged trachea air median over all sections with lung
area greater than 2000 mm2.

-920
Median Tracher Air CT Number [HU]

-930
-940
GE/LS 1
-950 GE/VCT
-960 Philips/6
-970 Siemens
-980 Siemens
-990 Siemens

-1000
-1010
15 20 25 30 35 40 45 50
Data analysisBody Massby
and slides Index
Phil Judy, BWH
 Near Term Issues:
Develop Metrics
• Emphysema
– Perc 15 vs. %LAA -910 (-950)
– Also establish norms, reproducibility, effect of lung volume
• Airways
– Nomenclature of airway generations
– Correction for patient size
– WA% vs. Thickness
– Role of attenuation
– Is it possible to summarize the information from the entire tree in one
(or a couple) parameter?
• Bronchiectasis
 Near Term Issues:
High Throughput Data Analysis
• Current approaches require a fair amount of
human labor
– Semiquantitative reading and scoring
– Software Programs that require airway editing
• How do we convert the ‘radiologist’s eye’ to an
automatable process?
– Texture Analysis
– ‘Learning’ algorithms in software
– Reference data set (COPDGene workshop)
 Near Term Issues:
Cross-Platform Interchangability
• Convince manufacturers to focus on quantitation and to
share data and/or adopt common standards- unlikely
-or-
• Develop protocols that allow valid comparisons between
different scanners
– NLST
– COPDGene
• Develop a common phantom to facilitate
calibration/correction
 Conclusions

• Our challenge is to generate meaningful,

reproducible anatomic and functional data.
• Challenge #2 is to do so with sufficient
throughput to keep up with other technologies,
such as genotyping.
• The dramatic increase in available computational
power provides the opportunity to use these data
to do personalized modeling .