Social networks and collaborative platforms are changing how radiology data is shared. The rise of online information sharing and cloud technology has led to a paradigm shift towards increased data sharing. This benefits research, enables second opinions, and advances precision medicine through collaborative care models. However, challenges remain around data storage needs, standardization across specialties, and ensuring patient privacy and control over their information as new players may enter healthcare using artificial intelligence.
Social networks and collaborative platforms for data sharing in radiology
1. Social networks and
collaborative platforms for
data sharing in radiology
ERIK RANSCHAERT, MD, PhD, CIIP
RADIOLOGIST, ETZ Hospital, Tilburg the Netherlands
EuSoMII President
2. Era of online information sharing
• We have become accustomed to sharing many types of
information via the Internet.
• Society finds this means of information exchange efficient
and desirable.
• Acceptance of this type of exchange of confidential
information indicates a reasonable level of trust by users of
this technology.
• Paradigm shift towards sharing of data
2
4. Technology shift in healthcare/radiology
Traditional image storage
“Siloed” architecture
o Intensive & expensive to manage
o Low utilization of HW & storage
o Vulnerable to failure & downtime
o Inefficient bandwidth
o Different for each site
Cloud-based image storage
“Virtualized” grid architecture
o Shared & cost-effective storage capacity
o Adaptive, self-healing, self-managing
o High speed
o Infrastructure is managed independent from application
o Single point of access: EPR, Portals
o Elimination of redundancy
Site 1 Site 2 Site 3
…
Site 1 Site 2 Site 3
…App 1 App 1 App 1 App 1 App 10 App 1 App 1 App 1 App 1 App 10
Cloud based storage platform
Source: SIIM Webinar Jan 28, 2016: “Real world challenges and benefits of cloud technologies” 6
5. Cloud-based archiving of data needed
• Storing requirements are gradually more demanding :
• Production of medical imaging will continue to increase
• New techniques are producing large data volumes
• Hybrid imaging: PET-CT, PET-MRI
• Other “-ologies” also want to store digital images
7
6. Any-ology
• Other specialties have needs for
both static & cine image storage
& retrieval
• Added value for collaboration
• More unified management of
data and images needed
• Yet not sufficiently addressed
by current methodologies, e.g.
problems of standardisation
Radiology Dermatology Ophthalmology
Oncology Cardiology Pathology
Otorhinolaryngology Neurosurgery Gastro-enterology
Source: SIIM Webinar Jan 28, 2016: “Real world challenges and benefits of cloud technologies” 8
12. Paradigm shift in healthcare
Current hospital-centric model Collaborative care model
14Source: SIIM Webinar Jan 28, 2016: “Real world challenges and benefits of cloud technologies”
Care
pathways
AI-based
solutions
Interoperable
data sources
Patient
Portal
2nd
opinion
s
Pharma
and HC
IndustryEpisodic
Departments
Proprietary
Data silos
Patients
Hospital centric
13. Why to share IMAGE & DATA?
• Patients are using the Web, mobile access to information
• Easy gathering of images from hospital
• Easy to obtain second opinions
Patient empowerment
• Broader market
• Increasing collaboration and specialty-oriented hospitals
• Personalized medicine
Telehealth
• Easy referral process to outside facilities
• Avoid redundancy
Easy access via cloud to
information
• Trauma & stroke patients: faster, treatment preparation
• Hospital as gateway to fluently organised patient-centric care
Liquid data, elegant
transfer of studies
• Principle of open science = sharing of data and knowledge
• Data needed for training, testing and validating of algorithms
• Labelled and validated data need to be shared among developers and institutions
Research and AI
15
Imaging data are only relevant when analysed in relevant context
14. Paradigm shift in society and healthcare
• Storing and viewing have
become commodity
• Steadily growing role of sharing
data and images in value chain
• Data are moving from storage
to intelligence
Integrate
Automate
Transfer &
collaborate
Visualise
Ingest & store images
15. Cloud storage – providers
• Amazon Web Services (AWS)
• Microsoft Azure (MA)
• Google Cloud Platform (GCP)
• lowest cost of storage
• most compliant with regulations
• API’s adapted to healthcare
services
16. CLOUD-based
image and medical
data management
platforms
Web-based secured imaging
Collaborative platforms
Preferably open-source SW
Interoperability is prerequisite
17. Cloud-based medical imaging platforms
• Cloud PACS
• Cloud VNA
• ZFP DICOM Web-based viewer
• Workflow analytics
• Provider network
• Scalable and secure
• Instant access to medical images and
data everywhere
• MEDICAL IMAGING ECOSYSTEM
18.
19. Google Cloud technology partners
• Works with Google Cloud Platform (GCP), but also AWS
and Microsoft as technology partners
• « Open source » TMIS (Translation Medical Imaging
Server) – only for clients
• Integration of « silos » from medical imaging,
pathology, clinical genomics -> precision medicine
• 360° view of patient: collaboration between
radiologists, pathologists, surgeons, oncologists by
sharing and visualizing patient data with relevant clinical
contaxt
• Present in 12 countries: USA, UK (+15 NHS trusts),
Spain, Mexico, Peru Jorge Cortell and his wife
23. AWS and medical data
• Agreement between Philips and
Amazon Web Services (AWS)
• Delivery of back-up service for
hospitals to store data in cloud, at
high tempo
• Improved protection of patient
data
• Increased need for storage
capacity due to complex imaging
techniques
25
24. Evolution in Healthcare
• PACS:
• Availability of cloud-based enterprise platforms in different formats
• Progressive move from hardware to virtual machines and services
• Digital infrastructure:
• Services offered by local providers (telco providers, data centers)
• Services offered by Google (GCP), Microsoft (MA) , Amazon (AWS)
• Progressive move from local machines to cloud machines
• Diagnosis:
• Move from local knowledge to automated and virtual knowledge
• Artificial Intelligence
26. Development process of AI algorithms
Concept
phase
Learning
phase
Testing
phase
Application
phase
Validation dataTraining datasets Test datasets
Goal of algorithm?
input
?
output
?
Training and test
datasets?
27. Input-output data
• The learning and testing steps require datasets in the form of pairs:
(input datai, output datai)i=1,N
• The learning dataset allows creation/adjustment of the decision model
• The test dataset evaluates the performance of the model
(because output datai is the truth about input datai )
• These datasets can be very diverse in nature
28. Radiology: early adopter of AI
• DL has exponentially grown from image analysis
• ImageNet
• Radiology is therefore at the forefront in medicine
• Abundance of data, mostly preserved in « silos »
• Data can be shared technically
33. Challenge to get data
• What datasources are currently accessible?
• Which image data are curated and labelled?
• What’s the role of the large cloud service providers?
• What is the role of hospitals and HC providers and radiological
societies?
34. Medical Image Databases and Biobanks
• UK Biobank
• Nationale Kohort
• The Cancer Imaging Archive (TCIA)
• Centre d’Acquisition et de Traitement des Images (CATI)
• FLI-IAM (France Life Imaging – Information Analysis and Management)
• NIH Image Gallery
• NCI Visuals Online
• CDC Public Health Image Library (PHIL)
• HON Media Gallery
• Health Education Assets Library (HEAL)
https://www.cna-aiic.ca/en/nursing-practice/overview-of-library-
resources/public-resources/images-and-image-banks
35. UK can lead in Radiology AI. Here’s how...Hugh Harvey, 2017, Medium - http://bit.ly/2CoKJE8
Industry partnerships
• Dominant industry vendors form
strategic partnerships with academic
institutions.
• GE, IBM Watson, Philips and many
others have created deals to allow data
access in return for funding research.
• In Europe we can see EU/industry
partnerships forming with the academy
36. Shift in healthcare market
• Shift of data sharing to higher level: artificial intelligence (AI)
• Where, how, at what price will the intelligence SW be
available?
• For what purposes will it be available?
• Discrimination?
37.
38. Zebra research platform
• Zebra has a database of +10 million de-identified imaging studies from
a variety of hospitals
• Ultimate vision: to create BLACK BOX in which a study comes in and
a diagnosis comes out.
• Goal is to develop hundreds of individual algorithms, for detection of
• breast cancer on mammography,
• malignant lung nodules on lung CT,
• brain tumours on MRI
• …
https://www.acpsem.org.au/documents/item/339
39. Clinical Research
Organisations
Health Insurance
companies
Healthcare providers
• Because Zebra can't build all of these applications on its own, the company published a
research platform on which it provides access to its large database of studies, as well as
storage, graphics processing unit (GPU) computing power and support for a variety of
research tools.
• They want to bring high-end research groups online to accelerate the pace and quality of the
algorithms that come out.
43. Relevance of protecting Health Data
• “FastMRI” partnership between NYU (CAI2R) and Facebook A.I. Research to make MRI
scans 10x faster
• Combination of domain-specific expertise from different fields and industries
• Train artificial neural nets to recognize underlying structures and construct MR-images
with less data
• NYU provides FAIR with 3 million MR-images (knee, brain, liver)
https://www.healthimaging.com/topics/artificial-intelligence/facebook-nyu-collaborate-make-mri-faster-ai
44. What’s next?
• Will there be new players in the HC market ?
• Data processors?
• Cloud service providers?
• Insurance companies as new payers?
• What business model will succeed?
• Will there be a shift in who owns and manages the data
(patient vs. hospital)?
• Is there a future for blockchain?
45. What business model?
• AppStore model, application on request
• Pay per analysis = fee for each analysis
• Subscription fee = fee per package of monthly analyses
• License = pay for whole platform with limited # analyses
• Included with modality
• Included with PACS/RIS
46. Hospital
Patient EHR
Physician A Physician B
Blockchain for
health data
In the future, the patient will be able to define the rights and obligations associated with the data.
48. The patient blockchain
• Embleema (NY, 2018)
• 1st blockchain network to give patients complete control over health data
• HIPAA compliant
• $2/month subscription fee
• Patients can grant access to their medical data for trials and are compensated for doing so
Our patient surveys have shown
that 53% would agree to it if they
had a proper way to give their
consent to accelerating research.
https://healthcareweekly.com/embleema-blockchain-healthcare/
https://support.embleema.com/hc/en-us/articles/360007096714-Embleema-Launches-the-First-Personal-Health-Records-Blockchain-To-Give-Patients-Complete-Control-Over-Their-Health-Data-and-Be-at-the-Center-of-Clinical-Research
49. Google and others?
• Google and others believe that
delivering AI through the cloud
will be a big, lucrative trend in
computing in coming years.
BSR annual meeting 2018
50. New HC market players?
• Amazon sees a new grow
market in HC and will offer HC
services
• Own health insurance – first for
Amazon employees (500.000 !)
• “We have the size for influencing
the insurance companies. The
problem of the rising costs in HC is
only becoming worse. Innovation is
necessary”.
Chief
Technology
Officer Amazon
NRC, 6 april 2018