The emerging role of quantitative imaging

By Andrew J. Buckler

Clinical radiology was previously limited to the subjective review of images, perhaps aided by simple screen tools like calipers. However, several innovations in quantitative analysis have emerged that yield objective data about system function, organ morphology and tissue processes. On the forefront of technology developments in the imaging contract research organization (CRO) industry, ACR Image Metrix™ integrated new imaging techniques and methodological advances to achieve more streamlined and efficacious clinical programs for the development of pharmaceutical agents and medical devices.

Quantitative imaging (QI) extracts quantifiable anatomical, functional and/or molecular features from medical images to assess the status of a disease, injury or chronic condition.  It encompasses the development, standardization and optimization of image acquisition, analysis, display and reporting protocols.  Standardization permits the validation of image-derived metrics with clinically relevant parameters, such as treatment response.  QI is used in research, drug development and clinical practice.

Driving factors in modern medicine include demands for prevention/early detection and more individualized, effective therapies.  These forces have also pressured the biopharmaceutical industry to streamline the drug/device development process.  This requires improved tools for the prediction and assessment of novel agents (Figure 1).  The FDA has called for “modernizing drug development by incorporating recent scientific advances, such as genomics and advanced imaging technologies, into the process.”

Figure 1

QI has applications throughout drug development that result in improved trial efficiency and reduced costs.  Detecting early phase response, QI can reduce the time required to conduct proof-of-concept studies for new drug candidates.  In early phase clinical trials, QI can help define the clinical dose range, improve patient selection and promote faster-paced programs.  If an early drug effect reliably predicts clinical outcomes, QI can serve as a surrogate endpoint in Phase III trials.  QI may also provide treatment response assessment when conventional techniques fail.

As the following example shows, diagnostic confidence in assessing treatment response depends not only on what is measured, but on how much of the available imaging data is utilized (Figure 2).

Figure 2

•    Before treatment, the 1D approach measures line length within a single image slice, whereas 3D volume interrogates all 3D pixels in the tumor.
•    Because tumors change asymmetrically, line length and volume measurements taken after treatment lead to different conclusions.
•    Whereas response evaluation criteria in solid tumors (RECIST) methodology (measuring change in line length) indicates tumor growth, the volume method indicates shrinkage.
•    This suggests added statistical noise in RECIST markup, leading to lower analytical power per subject than with the 3D volume method.
Most clinical trials evaluating solid tumor response to cancer treatment use RECIST, a set of published rules that define when cancer patients improve, stay the same or worsen during treatments.  However, relying solely on RECIST has its shortcomings.  

RECIST establishes categorical conclusions based on its inherent sensitivity, as determined from test-retest studies, to determine intra- and inter-reader variability.  While these wide categories ensure that crossing a category threshold reflects real biological change rather than measurement variance or bias, they result in a low analytical power per subject.  Both trial duration and patient enrollment (to achieve a given level of statistical power) are driven by the method's sensitivity and specificity.

Using all available pixel information factors out tumor morphology and heterogeneity to obtain a more accurate measurement of time-dependent change in size.  With less statistical noise, 3D volume measurement more quickly reflects biological change.  This can result in shorter trials and/or lower enrollment (Figure 3).

Figure 3.

Positron emission tomography (PET) has shown high predictive value in recent trials.  PET can assess early response in days rather than weeks, leading to more accurate prognosis and timely therapy adjustments.  18F-flurodeoxyglucose PET (FDG-PET) is a tool used to diagnose, stage and assess the treatment response of several solid tumor types.  An analog of glucose, FDG serves as a marker of tumor metabolic activity.  Cancers have an increased rate of glycolysis and glucose transport.  FDG is trapped intracellularly following injection, uptake and phosphorylation (Figure 4). 

Alterations and changes in FDG uptake after cancer treatment reflect the cellular response to therapy.  FDG-PET is used in animal models evaluating lead candidates for new treatments and is an established biomarker for several clinical indications in Phase III trials (e.g., Alzheimer's disease, lymphoma, infected joints).  FDG-PET can reveal certain physiological processes preceding tumor size change.  Early response measured by PET, coupled with mid-term CT volume measurement, may provide a more powerful indication of treatment response than RECIST.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) uses dynamic sequences to capture time-dependent behavior.  MRI's exquisite spatial resolution enables assessment of contrast agent accumulation in tumor neovasculature.  Because tumor growth depends on new blood vessel development, angiogenesis is considered a hallmark of cancer.  The time required for an agent to circulate through an enhanced area reflects tumor angiogenesis.  Thus, DCE-MRI can effectively assess the anti-vascular effect of therapeutic agents, sometimes within days of treatment initiation.  DCE-MRI may be superior to single image 3D volume measures for evaluation of anti-angiogenic agents.

Researchers from the University of Chicago have quantified the predictive capability of DCE-MRI in breast cancer.  Not only can DCE-MRI assess treatment response during drug development, its diagnostic systems have promising clinical applications.  The relative enhancement, especially the shape of the enhancement curve, is highly correlated with angiogenesis (Figure 5).

Figure 5

QI extracts objective phenotypic features covering a wide biological spectrum, from body systems to organs to tissues.  When these features are extracted with other quantitative features (e.g., cellular characteristics), it greatly increases the predictive power of pharmacogenomics (Figure 6). 

Figure 6

The Quantitative Imaging Biomarker Alliance (QIBA) is a coordinated effort involving clinicians, biopharmaceutical companies, imaging device manufacturers, government agencies and academicians seeking the qualification of imaging biomarkers.  Striving for consistent and reliable QI results across imaging platforms and clinical sites, QIBA has developed uniform procedures and processes.  Users like ACR Image Metrix and suppliers are working together to achieve biomarker qualification and standardization through FDA-approved validation processes. 

In summary, QI has promising applications in biopharmaceutical discovery and development.  These modalities can efficiently meet requirements for endpoints in trials, saving time and money.  With its ability to extract phenotypic features, QI is at the forefront of measurement science.  Whether applied in research, clinically or by organizations like QIBA, the future is bright for QI.

Figure 1 - Courtesy Stefan Ulzheimer of SIEMENS
Figure 2 – Adapted from one by Rick Avila of Kitware
Figure 3 - http://www.medicine.mcgill.ca/physio/joneslab/images/Fig1.jpg
Figure 4 – Hanahan and Weinberg, “The Hallmarks of Cancer,” 2009
Figure 5 – Giger, University of Chicago