Interagency Working Group on Medical Imaging National Science and - - PowerPoint PPT Presentation

Interagency Working Group on Medical Imaging

National Science and Technology Council

Renée Cruea, MPA

Executive Director Academy of Radiology Research

Thank you

The Academy of Radiology Research is an alliance of 27 professional imaging societies and 50 academic research departments, which together, represent the

scientific community advocating for more

sustainable medical research budgets. The Academy serves as the umbrella organization to the Coalition for Imaging and Bioengineering Research (CIBR). CIBR was established in order to foster collaboration among other important stakeholders in the imaging research community: imaging

equipment manufacturers, and patient advocates.

Organizational mission:

• Stop this damaging and unfortunate paradigm by

urging policymakers to invest more in federally funded R&D

• Use the public health AND economic power of

imaging science as a surrogate for federally funded R&D

• Explore new ways to convey to policymakers that a

national re-commitment to high-impact R&D can fuel the US innovation economy in the same way that made Apple the largest company in the world

Outline

§ Science to Meet a National Need § Demand for Imaging Science § Teeing Up Subsequent Speakers

Outline

§ Science to Meet a National Need § Demand for Imaging Science § Teeing Up Subsequent Speakers

Medical Imaging Research Initiative - The Committee believes there is potential in the near future to accelerate revolutionary new imaging technology for medical professionals and researchers to combat disease and support high-skilled manufacturing jobs in the United States. Such advances will require inter-agency coordination of Federal medical imaging research and development initiatives to accelerate the transfer of new technologies into commercial products manufactured in the United States and strengthen innovative research programs. Since many Federal agencies have existing and complementary roles on medical imaging research, there is a strong need for a Federal strategy that will coordinate and accel-erate such research. The Committee directs OSTP, in cooperation with the National Institutes of Health as the lead agency, to establish, through the National Science and Technology Council's Committee on Science, a Medical Imaging Subcommittee [MIS] to coordinate Federal investments in imaging

• research. The MIS should be required to develop a roadmap for the full scope of imaging

research and development, including: basic STEM science and technology creation, medical and translational research, evidence generation, clinical implementation, workforce and training support, and export-oriented manufacturing incentives.

“This memorandum outlines the Administration's multi-agency science and technology priorities for formulating FY 2016 Budget submissions to the Office of Management and Budget (OMB). The priorities covered in this memo require investments in R&D; support for activities, such as science, technology, engineering, and mathematics (STEM) education, technology transfer, R&D facilities, and scientific data collection and management, that enable a robust science and technology enterprise; and cooperation among multiple Federal agencies.

“Innovation in life sciences, biology, and neuroscience: Agencies should give priority to programs that support fundamental biological discovery research that could generate unexpected, high-impact scientific and technological advances in health, energy, and food security, particularly in platform technologies.”

“Within research portfolios, Federal agencies are encouraged to identify and pursue clearly defined "Grand Challenges” - ambitious goals that require advances in science, technology and innovation to achieve and to support high- risk, high-return research”

“Agencies should consider, where appropriate and authorized, supplementing traditional R&D "push" mechanisms (e.g., grants and contracts) with "pull" mechanisms - results-based market incentives designed to overcome market failures, engage a wide range of solvers, and catalyze innovation, such as incentive prizes and advanced market commitments.”

“Great promise for developing such technologies lies at the intersections of nanoscience, imaging, biological engineering, informatics, and other rapidly emerging fields of science and engineering.”

Outline

§ Science to Meet a National Need § Demand for Imaging Science § Teeing Up Subsequent Speakers

!$1,338!! !$1,652!! !$1,791!! !$1,889!! !$2,020!! !$2,148!! !$2,848!! !$3,052!! !$3,154!! !$3,257!! !$3,319!! !$3,772!! !$3,506!! !$3,634!! !$520!! !$696!!

6.5%% 7.1%% 6.6%% 6.7%% 7.1%% 7.5%% 9.8%% 10.3%% 10.3%% 10.4%% 10.7%% 12.2%% 12.0%% 12.1%%

$0! $500! $1,000! $1,500! $2,000! $2,500! $3,000! $3,500! $4,000! $4,500! 2001! 2002! 2003! 2004! 2005! 2006! 2007! 2008! 2009! 2010! 2011! 2012! 2013! 2014!

Total Imaging Research at NIH

"Positron-Emission Tomography" ultrasound tomography radiology MRI "magnetic resonance" "molecular imaging" "optical imaging" "imaging biomarkers" "MR elastography" "quantitative imaging" "image analysis" "image informatics" "image-guided"

!$#!!!! !$100!! !$200!! !$300!! !$400!! !$500!! !$600!! !$700!! 2005! 2006! 2007! 2008! 2009! 2010! 2011! 2012! 2013! 2014! Millions' NCI! NHLBI! NIMH! NINDS! NIBIB! NIA! NIDDK! NICHD! NIDA! NEI! NIAMS!

Total Imaging Research, by IC

0% 5% 10% 15% 20% 25% 30% % dollars imaging FY08 % dollars imaging CG % dollars imaging FY11

NIA NIAMS NEI NIDCD NIMH NHLBI NICHD NCI NIDDK NIDA

The Demand for Imaging Research

Percent of an IC’s Portfolio Imaging

Imaging at DoD, in nominal dollars

$63,854,335 $80,093,162 $88,410,896 $78,449,154 $101,160,327 2008 2009 2010 2011 2012

"Positron-Emission Tomography" "computed tomography" "computerized tomography" radiology "MRI" "magnetic resonance" "molecular imaging" "optical imaging" "imaging biomarkers" "MR elastography" "quantitative imaging" "image informatics" "image-guided"

Imaging at FDA, nominal dollars

"Positron-Emission Tomography" "computed tomography" "computerized tomography" radiology "MRI" "magnetic resonance" "molecular imaging" "optical imaging" "imaging biomarkers" "MR elastography" "quantitative imaging" "image informatics" "image-guided"

$1,049,115( $181,882( $715,173( $1,792,645( $1,763,519( $2,383,308( $4,698,469(

2008( 2009( 2010( 2011( 2012( 2013( 2014(

Imaging at NSF, by projects

453$ 689$ 721$ 459$ 420$ 460$ 457$ 2008$ 2009$ 2010$ 2011$ 2012$ 2013$ 2014$

"Positron-Emission Tomography" "computed tomography" "computerized tomography" radiology "MRI" "magnetic resonance" "molecular imaging" "optical imaging" "imaging biomarkers" "MR elastography" "quantitative imaging" "image informatics" "image-guided"

Imaging at VA, by projects

"Positron-Emission Tomography" "computed tomography" "computerized tomography" radiology "MRI" "magnetic resonance" "molecular imaging" "optical imaging" "imaging biomarkers" "MR elastography" "quantitative imaging" "image informatics" "image-guided"

19# 1# 44# 85# 89# 171#

2009# 2010# 2011# 2012# 2013# 2014#

Outline

§ Science to Meet a National Need § Demand for Imaging Science § Teeing Up Subsequent Speakers

NIBIB NCI NSF DOD/VA NASA NIST FDA CMS

Thank you

Academy of Radiology Research Renée L. Cruea Executive Director rcruea@acadrad.org

Interagency Working Group on Medical Imaging October 5, 2015

Richard L. Ehman, M.D. Professor of Radiology Blanche R. & Richard J. Erlanger Professor of Medical Research Mayo Clinic Chair , Board of Directors Radiological Society of North America

Head Injury

Impact On Health Care…

?

Diagnostic Burr Holes Advanced Medical Imaging Bleeding?

Abdominal Pain

Exploratory Surgery Pancreatic Cancer?

Advances in medical imaging have had an unparalleled impact on health care. The medical imaging research initiative proposes an intensified strategic focus

• n this field, through integrated, multi-agency federal initiatives, to yield

innovative science that addresses critical national needs. Research in the field of medical imaging is:

• technology-intensive,
• and often uniquely applies disciplines such as physics, mathematics, and

engineering to critical medical needs The advances achieved by use-inspired medical imaging research are historically:

• rapidly translated,
• and have a substantial impact on patient care,
• and frequently contribute to the medical innovation sector of the economy.

Medical Imaging Research Initiative

Inventions, embodied in patents, are a major driver of long-term regional economic performance, especially if the patents are of higher quality. In recent decades, patenting is associated with higher productivity growth, lower unemployment rates, and the creation of more publicly-traded companies.

Average 5.9 citations per patent

NIBIB NIGMS NIAID NCI NHGRI NHLBI NIDDK NINDS NIA NICHD NIMH NIDCR NIDCD NEI NIDA NIAAA NIEHS NIAMS NIMHD NINR NCCAM DARPA NSF DOE

1 4 16 64 5 10 15

Citations per Patent

NIH Nobelists

Patents per $100 million

NCCAM Complementary and Alternative Med NCI Cancer Institute NEI Eye Institute NHGRI Human Genome Research Institute NHLBI Heart, Lung, and Blood Institute NIA Institute on Aging NIAAA Alcohol Abuse and Alcoholism NIAID Allergy and Infectious Diseases NIAMS Arthritis, Musculoskeletal, and Skin Dis NIBIB Biomedical Imaging and Bioengineering NICHD Child Health and Human Development NIDA Drug Abuse NIDCD Deafness and Communication Disorders NIDCR Dental and Craniofacial Research NIDDK Diabetes and Digestive and Kidney Dis NIEHS Environmental Health Sciences NIGMS General Medical Sciences NIMH Mental Health NIMHD Minority Health and Health Disparities NINDS Neurological Disorders and Stroke NINR Nursing Research

Average 4.2 patents per $100M

Over the past 50 years, the Pentagon’s Defense Advanced Research Projects Agency (DARPA) has produced an unparalleled number of breakthroughs. Arguably, it has the longest-standing, most consistent track record of radical invention in history. A central reason DARPA has been so successful over time is its unwavering commitment to work in what the late political scientist Donald E. Stokes, of Princeton, described as “Pasteur’s Quadrant”. It entails pushing the frontiers of basic science to solve a well-defined, use-inspired need.

Donald E. Stokes 1927-1997

• Raised questions about the

productivity of a model that strictly separates basic and applied research

• Also questioned the linear

model of progress, starting from basic research, to applied research, to development, production, and finally

• peration.
• Proposed a different, 2

dimensional view of scientific progress.

Pasteur Quadrant

Emphasis on Fundamental Understanding Relevance for Immediate Application

High Low High Low

Bohr Quadrant Pure Basic Research Use-Inspired Research Edison Quadrant Applied Research

NIBIB NIGMS NIAID NCI NHGRI NHLBI NIDDK NINDS NIA NICHD NIMH NIDCR NIDCD NEI NIDA NIAAA NIEHS NIAMS NIMHD NINR NCCAM DARPA NSF DOE

1 4 16 64 5 10 15

Citations per Patent

NIH Nobelists

Patents per $100 million Pasteur’s Quadrant?

Cooley-Tukey Butterfly

0.0

5.0 10.0 15.0 5 10 15

Patents per $100 M of R&D Funding Mean Citations per Patent

NIH DOE NSF DOC/NIST

Patents as Proxies Revisited: 2015 Battelle Report

What Gaps and Opportunities in Medical Imaging Science and Technology Can be Advantageously Addressed by Integrated, Multi-Agency Federal Initiatives?

• Essential tool in precision medicine
• Provides objective endpoints in clinical trials (biomarkers)
• Covers a wide range of biological scale, molecule to organism
• Reflects histology, while providing a vastly larger volume of tissue

characterization and classification

Quantitative Imaging:

• Standardize and optimize medical

imaging acquisition protocols, technology standards, data analysis, display methods, and reporting structures.

• Validate image-derived metrics with

anatomically and physiologically relevant parameters, such as treatment response and outcome.

• Professional Societies
• Academic Institutions
• Pharmaceutical companies
• Imaging device companies
• Imaging informatics companies
• Imaging core labs (iCROs)
• Government agencies
• Clinical trialists and clinicians

• CT Volumetry
• Lung Density
• FDG-PET/CT
• Functional Neuro MRI
• Magnetic Resonance Elastography
• Perfusion, Diffusion, and Flow MRI
• PET-Amyloid
• SPECT
• Ultrasound Shear Wave Speed
• Ultrasound Flow Volumetry

QIBA Progress

• Multiple profiles are being developed
• QIBA Metrology Working Group publications
• Quantitative Imaging Data Warehouse (QIDW)
• Over 216 users, 14 communities, 127,000

data sets

• Three NIBIB contracts, ~$7.5M, have funded

55+ groundwork projects

• New international extensions of QIBA in

Europe, Japan, and Brazil

• QIBA profile specifications are being

incorporated into industry and federally- funded clinical trials

• Marketing of imaging technologies is starting

to emphasize quantitative ability

• Patients are primary beneficiaries: improved care + accelerated clinical trials
• Imaging manufacturers encouraged to develop technology with reduced

variability, increased accuracy, and substantial capability improvements

• Medical device and drug trials facilitated by improved performance of

quantitative imaging biomarker measurands => smaller sample sizes

A Promising Approach Opportunity: Scale-Up and Accelerate the Approach

• A large proportion of key stakeholders have been successfully engaged
• Methodology has been established and tested
• With appropriate resources and prioritization, there is now an opportunity to

efficiently scale-up and accelerate the scope and impact of the QIBA process to realize the great promise of quantitative imaging in health care

• At the federal agency level, both “push” and “pull” measures may be

appropriate

CAD in Mammography "Ambient Medical Image Analytics" in Medical Imaging Workflow

Aneurysm "sniffer" system B Erickson, MD Mayo Clinic

Challenge:

Fully-Automated Medical Image Interpretation

Realizing this capability will likely require:

• Use-inspired multidisciplinary efforts
• Data curation
• Large databases
• Deep machine learning technology
• Re-engineering image acquisition
• Re-imagining the entire diagnostic process

Benefits: Standardized representation of information, eliminating inter-observer variability, potential for higher diagnostic performance and reliability

Why aren’t we using our most powerful medical imaging technologies more proactively?

Avascular necrosis of the femoral head

• Telephone
• Refrigerator
• Computer
• AED

ca 1900 ca 1969

Key Issues:

• Technology requirements
• Information management
• Medical economics

Why aren’t we using our most powerful imaging technologies more proactively?

General Electric Co. Global Research chief engineer, Trifon Laskaris, stands next to a 3-tesla Magnetic Resonance Imaging (MRI) scanner, used for scanning heads only, which he is process of building, Monday May 6, 2013, at GE Global Research in Niskayuna, N.Y. Laskaris was recently awarded his 200th U.S. patent. The 3-T scanner would mark 201 patents, and he has another 25

• pending. (Will Waldron/Times Union)

Low Cost Compact 3T MRI

Motivation

• Global shortage of MRI access with wide disparities
• Helium is a non-renewable resource in short supply
• A low-cost, compact system for heads, extremities,

and infants could significantly expand access

• A Public/Private/Academic Partnership
• NIH Grant: BRP-R01-EB010065 (2010)

2015: First Human Images from Low-cost Compact 3T System

Comparison to Standard 3T MRI

< ½ space ~ ¼ weight < 1% helium

10 20 30 40 50 60

Open MRI Standard 3T MRI High Performance 3T Low Cost Compact 3T Connectome 3T

MRI System Performance: Slew rate * Gmax

Challenge:

Disruptive Transformation of Our Most Powerful Medical Imaging Technologies

Key Issues that need to be addressed:

• Technology requirements
• Information management
• Medical economics

An opportunity for an integrated, multi-agency federal initiative to provide much-needed strategic leadership Benefit: High value imaging. Ready, low cost, access to the most effective imaging technologies for health care.

Three “Challenge” Opportunities

• The promise of Quantitative Imaging in Healthcare
• Fully-Automated Medical imaging Interpretation
• Disruptive Transformation of Medical Imaging

An Infrastructure to support translational and clinical research in Medical Imaging

Mitchell Schnall MD, PhD Chair ACR Research Commission

Eugene Pendergrass Professor of Radiology The Perelman School of Medicine at the University of Pennsylvania

ACR Research and Innovation

11/2/15 ¡ 2 ¡

• Spans more than 40 years

– Over 500 clinical trials – 2 million images processed annually

• Established research infrastructure in Philadelphia

– 16,000 square feet of office space – 120 researcher professionals – ACR Imaging Core Laboratories – Can accommodate up to 50 readers simultaneously – Health Policy Institute – Socio-Economic Research – 8 Reston Employees

Neiman ¡ Health ¡Policy ¡ Ins2tute ¡ IROC ¡ NRG ¡

Opera2ons ¡ Head ¡Injury ¡ Ins2tute ¡ Cardiovascular ¡ Commi=ee ¡

ACR ¡RESEARCH ¡COMMISSION ¡

Health ¡Services ¡ Research ¡

Neuro ¡ Commi=ee ¡

ECOG-­‑ACRIN ¡

ACR ¡Peds ¡ Commission ¡

ACR ¡Research ¡and ¡Innova2on ¡ ¡

RTOG ¡ Founda2on ¡

Industry ¡

ACRIN ¡ ¡ Industry ¡

Pediatric ¡ Commi=ee ¡ ECOG-­‑ACRIN ¡ Opera2ons ¡ Industry ¡ ¡ Data ¡ Management ¡

Radia2on ¡ Oncology ¡

NRG ¡Data ¡

Management ¡

and ¡ ¡

Sta2s2cs ¡

Serves as infrastructure to serve the broad imaging research community

ACRIN

Imaging

Science Community

Imaging Standards Organizations

Imaging

Industry

Clinical Providers Multicenter translational and clinical imaging research to translate innovation to practice Training a generation of imaging clinical trialists Contribution of methods and standards for imaging clinical trials

Government

Organizational accomplishments

§ Accrued over 75,000 patients to imaging related clinical trials § Developed a network of over 200 sites to participate in multicenter imaging trials § Deployed a clinical trials image management IT infrastructure (TRIAD) across 7000 sites.

§ Integrated with Medidata RAVE

§ Integrated program with the NCI Quantitative Imaging Network program (QIN) § Organized a distribution network for experimental radiopharmaceuticals.

Some important trials/results

• DMIST: Digital mammography has higher

sensitivity than film in young, premenopausal and women with radiographically dense breasts

§ NLST: CT lung cancer screening reduces lung cancer specific mortality by 20% in a selected high risk population

Some important trials/results

Arm Person Years (py) Lung cancer deaths Lung cancer mortality per 100,000 py Reduction in lung cancer mortality (%)

CT 144,097.6 354 245 20.3 CXR 143,363.5 442 308

First Major ACRIN Cardiovascular Publication

ER discharge Rate CT: 49.6%; Control: 22.7% Length of Stay CT:18 hours; Control:24.6 hours Dx of CAD CT:9%; Control:3.5%

No difference in rate of MI or Death

Collaboration with CMS

IDEAS Registry

§ PET Amyloid Imaging for Alzheimer's Disease § CMS approved

§ Activates in early 2016. § Created by and funded by the Alzheimer’s Association § Additional funding by Industry

§ ACR Operations Center

11

A Phase II Study of 3′-Deoxy-3′-18F- Fluorothymidine PET in the Assessment of Early Response of Breast Cancer to Neoadjuvant Chemotherapy: Results from ACRIN 6688

Lale Kostakoglu1, Fenghai Duan2, Michael O. Idowu3, Paul R. Jolles3, Harry D. Bear3,4, Mark Muzi5, Jean Cormack2, John P. Muzi5, Daniel A. Pryma6, Jennifer M. Specht5, Linda Hovanessian-Larsen7, John Miliziano8, Sharon Mallett9, Anthony F. Shields10 and David A. Mankoff6 on behalf of the ACRIN 668 Investigative Team behalf of the ACRIN 668 Investigative Team

Core Lab Support Process: Meets GCP standards

Protocol development Site qualification Site management QA review Analysis Image management Protocol development Analysis QA review Site qualification Site management Image management

TRIAD ¡Site ¡Server ¡Auto-­‑Popula=ng ¡a ¡Registry ¡

Scanner PACS

Localizer

• r Scout

Image DICOM Dose Structured Report Dose Sheet DICOM

1. De-identification

2. Normalization 3. Authentication 4. Transmission

TRIAD Site Server TRIAD Central Services

Web Services

Dose Registry

Imaging Facility ACR

Post Processing NRDR Portal Facility and Benchmark Reports

• Over 750 facilities active across the US
• Over 1 million scans per month

Active TRIAD Sites

1000 2000 3000 4000 5000 6000 7000 8000 4 8 12 16 19

Accreditation Clinical Trials Registries 2011 2012 2013 2014 2015 Q3

• The TRIAD network continues to expand across sites that span

the public and commercial sectors, academic and private practices, urban and rural settings, and civilian and military institutions.

• With over 7,000 active sites, TRIAD’s footprint facilitates the

widespread movement of imaging data.

Integrating images across scales with metadata in the data warehouse

Rave Pt Reports

Records Claims data Registries Digital Pathology

Images

Laboratory Data

Data Warehouse

High Throughput Genomics

Data exploration interface

Summary

§ The ACRIN is the hub of an extensive infrastructure to support translation and validation of imaging technology § This effort is collaborative with NIH, industry, academia and other professional societies § Although the primary focus is radiologic imaging, there are efforts to leverage this infrastructure more broadly to include pathology and other disciplines that use imaging.

Thank You

17

James H Thrall MD

Chairman Emeritus, Department of Radiology Massachusetts General Hospital Distinguished Juan M Taveras Professor of Radiology Harvard Medical School International Society for Strategic Studies in Radiology (IS3R)

International Priorities and Trends In Medical Imaging

 Precision Medicine

 Radiogenomics

 Molecularly targeted imaging  Deep Learning– Big data, data mining

 Data into knowledge  Radiomics– quantitative tissue signatures

(phenotypes)

“I want the country that eliminated polio and mapped the human genome to lead a new era of medicine — one that delivers the right treatment at the right time. …Tonight, I'm launching a new Precision Medicine Initiative to bring us closer to curing diseases like cancer and diabetes — and to give all of us access to the personalized information we need to keep ourselves and our families healthier.” President Barack Obama, State-of-the-Union speech, Jan 20, 2015 $215 million in proposed budget specifically for precision medicine initiatives

Toward Precision Medicine: Building a Knowledge Network and A new Taxonomy of Disease

 “The tailoring of medical treatment to the individual

characteristics of each patient”

 “Classification of patients into subpopulations that differ

in their susceptibility to a particular disease, in the biology and/or prognosis of those diseases, or in response to a specific treatment”

 Subpopulations defined by genotype and phenotype

National Research Council of the National Academies, White Paper, 2011

Radiogenomics:

The study of the linkage between genotype and imaging phenotype

When Genotype is known:

What are the image findings associated with gene expression? (I.e. imaging phenotype)

When Imaging Phenotype is known (I.e. image findings):

What is the genotype?

Patient with NF1

Whole body MRI with image segmentation

Genotype is known– what are the manifestations of gene expression? What is the imaging the phenotype?

Texture energy maps in colon cancer

Texture energy maps from a cecal tumor Selected ROI

S Do, C Cruz Romero, MGH

Energy values per tumor genotype

• Each genotype has a different

signature

• BRAF tumors demonstrated the

lowest average texture energies

• NRAS tumors demonstrated the

highest average texture energies

S Do, C Cruz Romero, MGH

Biomarkers and Phenotypes

ADC DCmin

in is

is a Prog

• gnos

nostic ic Bio iomar arker in in Ovar aria ian C n Cance ancer (A

(ADC DC—Apparent D Diffusio ion Coefficient)

Nakamura K, Imafuku N, Nishida T, et al. Gynecol Oncol 2012;124:335–339.

111 patients with preoperative DWI

CA--125 (Cancer Antigen 125) ADCmin

ECOG—ACRIN

Cancer Research Group

Gallam ini et al. JCO, 2 0 0 7

International Prognostic Score

F/U 18FDG PET is a predictive biomarker for progression free survival

Gallam ini et al. JCO, 2 0 0 7

* *

Chemosensitive PET

FDG PET changed the IPS prediction for survival

Non Chemosensitive PET

18F-FDG PET/CT

Glycolysis

18F-FDHT PET/CT

Androgen Receptor

Prostate Cancer

Revealing Heterogeneous Biology

• f Tumor Metastasis

CT

Hricak H.: Oncologic Imaging: A Guiding Hand

• f Personalized Cancer Care ; Radiology 2011

Zr-89 J591 PSMA mAb*

18F-FDG PET/CT

Glycolysis CT

*Pandit-Taskar, Eur J Nucl Med Mol Imaging: 2014

Prostate Cancer

Revealing Heterogeneous Biology

• f Tumor Metastasis

DEEP LEARNING

WHAT IS IT?

A TYPE OF MACHINE LEARNING USING AI AND NEURAL NETWORKS

WITH MINIMAL HUMAN INPUT (UNSUPERVISED)

REQUIRES LARGE DATA SOURCES FOR SYSTEM TRAINING

K Dreyer, MGH

DEEP LEARNING OF OBJECTS

TRAINING PROCESS

Untrained Neural Network DOG CAT FISH HUMAN Trained Neural Network

K Dreyer, MGH

Radiomics– An application of deep learning

 Extraction of large numbers (hundreds)

• f quantitative parameters (biomarkers)

from medical images

 Quantitative imaging phenotypes

 Radiomic image analysis can be applied

to every image voxel

 Tissue heterogeneity (proteins, cells,

microenvironment) limits usefulness of biopsy based molecular assays

Radiomics Work Process

Segmentation of tumor Analysis– size, shape, intensity, texture Analysis

From www.radiomics.org

Radiomics Heat Map

440 quantitative features

Strategies for Maintaining US Leadership in Medical Imaging

 Exploit US leadership in medical

imaging by fostering new interdisciplinary and interagency programs at the confluence of trends in biomedical science and imaging science

Opportunities for Interdisciplinary and Interagency Research

 Precision Medicine and patient phenotype:

 Validation of imaging biomarkers  Correlation of image data with other information sources to

better classify patients into more precise subpopulations (sub phenotypes)

 Disease biology:

 Imaging phenotypes provide clues to disease biology– gene

expression

○ Differences in presence, location, extent and behavior of

diseases

○ Differences in response to treatment

 Molecularly targeted imaging

○ Probe development ○ Probe application

 Cancer heterogeneity

Opportunities for Interdisciplinary and Interagency Research

 Imaging and Treatment:

 New imaging criteria for selection of therapy  New criteria for monitoring response to therapy  Imaging informed adaptive therapy  New criteria for establishing prognosis

 Imaging and Clinical trials:

 Use of imaging inclusion criteria in patient

selection for clinical trials

 Use of imaging biomarkers as surrogate

endpoints in clinical trials

Strategies for Maintaining US Leadership in Medical Imaging

 Exploit US lead in computing capabilities

and Deep Learning:

 Build data bases available for data mining  Support research in multi- information source

Deep Learning

○ Mobilize the Deep Learning toolkit ○ Establish reference data sets to compare methods

 Criteria  Standards

 Establish registries:

○ Imaging biomarkers ○ Phenotyping systems-- radiomics ○ Radiogenomic linkages