System for Quantifying Treatment-induced Image Change,

and for Integrating Quantified Change into Cancer

Treatment-related Decision-making and Diagnosis

Timothy Sawyer, MD (presenter)

ImQuant, Inc.
Richard Robb, Ph.D.
Mayo Clinic Biomedical Imaging Resource
Robert Foote, M.D.
Vice Chairman of Radiation Oncology, Mayo Clinic
Val Lowe, M.D.
Director, PET Imaging, Mayo Clinic
Shigeru Yokoyama, Ph.D.
Radiation Physics, Idaho Quantitative Oncology Consortium / Saint
Alphonsus Regional Medical Center, Boise, ID

Edited version of presentation given at RSNA

November 30, 2005
 Introduction
Cancer treatment is not tailored to individual patients -- less
than optimal outcome, greater than necessary toxicity
and expense

Traditional advanced medical images are of limited value to

the present-day practice of oncology

Hypothesis:
A novel, quantitative imaging format that preserves
spatial integrity -- in combination with the right software-
based processes -- can facilitate tailoring of cancer
treatment to individual patients
 Introduction
Example: Adjuvant chemotherapy

Adjuvant chemotherapy (breast, colon, non-small cell

lung, and other cancers):

Set agents given for a set number of

cycles, based on empiric evidence
from randomized trials

10 % improvement in long-term survival:

100 patients are treated to benefit 10.

 Introduction
Example: Adjuvant chemotherapy
Statements that are probably true:

A -- Of the 10 patients that benefited, some did not need all of the (toxic and
expensive) cycles administered

B -- Of the 90 patients who did not benefit:

Some would have been cured even without chemo
Some will die, despite receiving it

C -- Of those that died despite receiving chemo:

Some would have lived with additional cycles of the same chemo
Some would have lived with different chemo

All received toxic, expensive treatment, without cure

 Introduction
Example 2: Prostate cancer irradiation

Cancer in anterior prostate 81 Gy to entire prostate, +

Cancer in posterior prostate margin, including anterior
Cancer very radiosensitive rectum and posterior
bladder
Cancer very radioresistant
 Imaging

The ultimate in vivo system for measuring

anatomy, physiology and function, and
molecular concentration
 Ideal image quantification / treatment tailoring system

• Applicable to all major imaging sources -- PET, SPECT, MRI, MRSI,

DCE-MRI, diffusion MRI, perfusion MRI, etc.
• Applicable to all major cancer applications -- systemic therapy, radiation
therapy, surgery, etc.
• Considers each voxel, yet in quantifying response and making
predictions, maintains spatial integrity of entire tumor, and spatial
relationships between voxels
• Detects and quantifies changes in sub-regions of the tumor that are
physiologically or molecularly heterogenous, or that respond to therapy
at different rates or to different degrees
• Compares changes in sub-regions of the tumor to other sub-regions,
and to a data bank of sub-regional changes for which the outcome is
known. Goals: a) to gain early warning signs from a particular sub-
volumetric region, even when an overall tumor appears to be
responding well to therapy, and b) to facilitate differential radiation
therapy dosing to different regions of the tumor
 Ideal image quantification / treatment tailoring system
(continued)

• Does not rely on simple registration / fusion and digital subtraction,

since the pre-therapy versus the mid-therapy tumor is of a different
shape and size
• Results not only in change quantification, prediction, and treatment
recommendations, but also in automatic mid-therapy display of key
sub-regional contours, for dynamic (mid-therapy) changes in intra-
tumor radiation dose-painting (delivery of higher dose, each fraction,
to the sub-regions responding less, and / or less likely to continue to
respond)
• Sensitive and specific enough to detect and quantify changes very
early in treatment, so that early changes in treatment can be made
(resulting in better outcomes, less toxicity, and less expense)
• Simple to use
 Methods

Tools and Code from AVW library Developed

by Mayo Clinic Biomedical Imaging
Resource

New code with customized graphical user

interface from Mayo BIR and ImQuant
PET, MRI, SPECT, volumetric MRSI, etc.
Voxels

Intensity Value X

Z
 Methods
• Voxels of like or similar intensity value (physiology,
molecular concentration, etc.) “connected” to form
isonumeric contours
• Collections of isonumeric contours, each representing a
different intensity value, form 3D “functional, physiologic,
or molecular profiles”
• 3D molecular / functional profiles analyzed for
quantifiable features
• Images, and image changes, represented as numbers,
sets of numbers, graphs, or equations
• Series of processes developed, to integrate quantified
change into clinical decision-making
Potential Methods for Isonumeric Contour
Determination

• Uniform Intensity Spacing

• Histogram Distribution
• Multispectral Classification
• Watershed (Level Sets)
• Distance From Center/Edge
• Brightness-Area Thresholding
 Results

Intensity Elevation
ROI Map in ROI

Isonumeric contours
in ROI based on
intensity gradients
Results Thin Sagittal cuts
through ROI
“Functional Topography” (3D +
alpha)
Partial list of quantifiable features
measurable by software-based tools
• Volume of each contour
• Surface area of each contour
• Shape characteristics
• Median, peak intensity values for
voxels within a contour-defined
volume
• Distance of contours from each
other
• Distance of contours from a point
• Volumes of “elevations” (analogy
from conventional
topography)
• Volumes of “depressions”
• Max or min intensity level within an
elevation or depression
• Numbers, or locations, of
elevations / depressions
 Results

Images represented as numbers, sets of

numbers, graphs, or equations

Treatment-induced image change represented

as numbers, sets of numbers, graphs,
equations
 Results
Approximately 60 processes developed, to integrate change into clinical decision-making
in medical oncology, radiation oncology, diagnosis, surgery, and other interventions

Example: Dynamic Chemotherapy

• Image
• Administer systemic therapy
• Re-image
• Compare images or imaging data
• Express volumetric change as numbers, sets of numbers, graphs, or equations
• Express sub-volumetric changes as numbers, sets of numbers, graphs, or equations
• Compare volumetric change to volumetric data bank of changes for which outcome is
known
• Compare sub-volumetric changes to sub-volumetric data bank of changes for which
outcome is known
• Express relative volumetric change
• Express relative sub-volumetric change
• Predict ultimate likelihood of favorable volumetric change, or other clinical endpoints,
assuming no change in plan
• Rules engine-based recommendation for next cycle (change interval, dose, agents)
Conclusions and Future Directions
Relevant to multiple imaging sources
Relevant to multiple applications

Display
Diagnosis
MRI Chemotherapy tailoring
Perfusion MRI Tailoring radiation therapy
Diffusion MRI dose

DCE-MRI Tailoring radiation therapy

targeting and intra-tumor
MR Spectroscopy dose-painting
PET Dynamic and change-based
SPECT targeting
Others Surgical planning
Surgical targeting