CHIPS - A Service for Collecting, Organizing, Processing, and Sharing Medical Image Data in the Cloud
Rudolph Pienaar, PhD
Technical Director
Fetal-Neonatal Neuroimaging and Developmental Science Center
Staff Scientist
Boston Children's Hospital
introduction
Healthcare informatics is at an inflection point
This talk considers a particular perspective on the problem and proposes a solution
Cloud
Healthcare
Information
Processing
Service
CHIPS
overview
Some trends confronting computing in healthcare and shaping the landscape
An introduction to our web-based system called CHIPS
Logical topology / design
UI considerations
Containerization
Healthcare informatics landscape
Clouds and untethered computing
Historical legacy
overview
you have to know the past
to understand the present
Carl Sagan
overview
overview
Information explosion...
Annual global IP traffic will pass the zettabyte (1000 exabytes) threshold by the end of 2016, and will reach 2 zettabytes per year by 2019.
-
By 2016, global IP traffic will reach 1.1 zettabytes per year, or 88.4 exabytes (nearly one billion gigabytes) per month, and by 2019, global IP traffic will reach 2.0 zettabytes per year, or 168 exabytes per month.
Global IP traffic has increased fivefold over the past five years, and will increase threefold over the next five years.
-
Overall, IP traffic will grow at a compound annual growth rate (CAGR) of 23 percent from 2014 to 2019.
http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/VNI_Hyperconnectivity_WP.html
Information explosion...
Two-thirds of all IP traffic will originate with non-PC devices by 2019.
-
In 2014, only 40 percent of total IP traffic originated with non-PC devices, but by 2019 the non-PC share of total IP traffic will grow to 67 percent.
Traffic from wireless and mobile devices will exceed traffic from wired devices by 2016.
- By 2016, wired devices will account for 47 percent of IP traffic, and Wi-Fi and mobile devices will account for 53 percent of IP traffic. In 2014, wired devices accounted for the majority of IP traffic, at 54 percent.
http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/VNI_Hyperconnectivity_WP.html
Information explosion...
Global Internet traffic in 2019 will be equivalent to 66 times the volume of the entire global Internet in 2005.
-
Globally, Internet traffic will reach 37 gigabytes (GB) per capita by 2019, up from 15.5 GB per capita in 2014.
The number of devices connected to IP networks will be more than three times the global population by 2019
-
There will be more than three networked devices per capita by 2019, up from nearly two networked devices per capita in 2014. Accelerated in part by the increase in devices and the capabilities of those devices, IP traffic per capita will reach 22 GB per capita by 2019, up from 8 GB per capita in 2014.
http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/VNI_Hyperconnectivity_WP.html
Information explosion...
Information explosion...
Healthcare vs Financial
- Financial
- Number of accounts: 10,000 to 300 million
- Storage per Account: GB
- Total Storage: 10s of TB to 300 PB
- Healthcare
- Number of patients: 10,000 to 300 million
- Storage per Patient: GB today, TB tomorrow
- Total Storage: 20 PB to 600 EB
Information explosion...
Healthcare
Information explosion...
Healthcare
- 1.65M tumors measured in various modalities
- > 4 exabytes of data in various localities
- 1 petabyte of data (0.001EB) to transfer
- On current Internet, would take about 21 days to transfer this data around.
an operating theatre...
healthcare... philosophy shapes informatics
historically, healthcare is differential, not integrative
informatics is poorly understood and not seen necessarily as fundamental
classically information relationship is best described as 1:1
the practice of medicine is less integrative
healthcare "compute"...
provided by vendors with very specialized skills and market
thin, vertical stacks
Typical hospital "web" apps...
Typical hospital "web" apps...
Typical hospital "web" apps...
healthcare informatics landscape...
Healthcare informatics is at an inflection point
Need: a platform that combines multiple data and multiple compute on useful hardware
explosion of sensors
complex analytics
multiple possible data sources
lack of local compute resources
Forces acting on healtcare workflows
Compounding Complexity
- Data is distributed in multiple places
- EMRs
- Modality Databases
- Department databases
- Different formats
- Data is structured and unstructured
- "Voice" recordings
- "Free form text"
Forces acting on healtcare workflows
Compounding Complexity
- Inconsistent/variable definitions
- Quantitative interpretations
- Qualitative measurement variability
- T(R) on Siemens might be different to GE
- Data itself is complex
- genomics
- radiomics (images)
- combinatorial explosion
Typical hospital "web" apps...
Current App Philosophy
- Old-style "web" apps subsumed to look like desktop apps.
- Reflect an "app" centric model, not "data" centric.
- Not really "web" apps:
- Typically limited to obsolete versions of Internet Explorer
- No support for mobile or other platforms.
- No cross-app integration on data level.
some trends in informatics...
disparate, disconnected computing
the primacy of the browser as client platform
the rise of the ...
silos...
Multiple, deep data sets exist in the healthcare dataverse
Largely disconnected and independent of each other
clinical
billing
health records
sparse dataverse...
data points, rich as they are, can be insufficient...
the more connections between points...
data needs "connectedness"
the better!
patterns of information flow in healthcare
Current
- Unidirectional
- Isolated
- Non-integrative
- Static
Future
- Multi-directional
- Collaborative
- Integrative
- Dynamic
clouds and clouds and clouds...
what are "clouds"?
remember above all others that a "cloud" is really just...
someone else's computer out on the network
what do clouds do?
well... they rain...
essentially, we often think of "cloud" computing as less of "computing" and more as storage
we use "clouds" as hard drives in the sky
pull data (i.e. rain) and consume it locally
medical compute... in the cloud
definite need for data mobility and centralization...
with compute being "free" to process data where-ever it might naturally collect...
CHIPS...
CHIPS is a bio-medical data workflow manage that allows easy and intuitive collection, analysis, and sharing of data between parties.
- allows for advanced view rendering
- allows for sharing of images
- allows for immersive real time collaboration
- allows for post-processing of images
many data sources... many compute sources...
via here
connect data here
to here...
and compute!
system topology
Main web server local data repository
remote compute nodes
data source (e.g. in hospital PACS)
distributed design...
web server
coordinator
data handler
compute handler
distributed design...
Red Hat collaboration
containerization
containerization
compute and io...
disparate, disconnected computing
data needs to be transferred from server host to remote location
data needs to be processed on remote location
io
local web server
data descriptor
data handler
io
pfurl --verb POST --raw --http 172.17.0.2:5055/api/v1/cmd --msg \
'{ "action": "pushPath",
"meta": {
"remote": {
"key": "someKey"
},
"local": {
"path": "/home/data"
},
"transport": {
"mechanism": "compress",
"compress": {
"encoding": "base64",
"archive": "zip",
"unpack": true,
"cleanup": true
}
}
}
}' --quiet --jsonpprintindent 4
compute
local web server
exec descriptor
exec handler
compute
pfurl --verb POST --raw --http 172.17.0.2:5010/api/v1/cmd \
--jsonwrapper 'payload' --msg \
'{ "action": "run",
"meta": {
"cmd": "cal 7 1970",
"auid": "rudolphpienaar",
"jid": "cal-job-1234",
"threaded": true
}
}' --quiet --jsonpprintindent 4
compute... using a container
pfurl --verb POST --raw --http 10.17.24.163:5010/api/v1/cmd
--jsonwrapper 'payload' --msg '
{ "action": "run",
"meta": {
"cmd": "$execshell $selfpath/$selfexec --prefix test- --sleepLength 0 /share/incoming /share/outgoing",
"auid": "rudolphpienaar",
"jid": "simpledsapp-1",
"threaded": true,
"container": {
"target": {
"image": "fnndsc/pl-simpledsapp",
"cmdParse": true
},
"manager": {
"image": "fnndsc/swarm",
"app": "swarm.py",
"env": {
"shareDir": "/home/tmp/share",
"serviceType": "docker",
"serviceName": "testService"
}
}
}
}
}
'
coordination
timing can be complex...
PUSH data
separate coordinator service
WAIT!
EXEC on data
WAIT!
PULL data
WAIT!
coordination
security
- Medical data is constrained by many external factors
- Logging and control
- Single shutoff point
- Communication pathways
- HIPAA
- Local regulatory
- IRB
secure architecturally
secure communication
big data
- Imaging
- Genomics
- Pathology
- Input DICOM space
- Processed result space
- Structured clinical reporting
big data...
UI concepts
UI concepts
UI concepts
conclusion
- Healthcare compute is at an inflection point
- Lags larger computing -verse
- Many opportunities and challenges exist
Untapped opportunity to integrate cutting edge approaches to inform healthcare
references... CHIPS/ChRIS
-
https://github.com/FNNDSC/ChRIS_ultron_backEnd
-
https://github.com/FNNDSC/ChRIS_ultron_frontEnd
references... services
-
https://github.com/FNNDSC/viewerjs
-
https://github.com/FNNDSC/gcjs
-
https://github.com/FNNDSC/fmjs
-
https://github.com/FNNDSC/toolbarjs
-
https://github.com/FNNDSC/rboxjs
-
https://github.com/FNNDSC/rendererjs
-
https://github.com/FNNDSC/thbarjs
-
https://github.com/FNNDSC/pman
-
https://github.com/FNNDSC/pfioh
-
https://github.com/FNNDSC/pfcon
references... viewers
-
https://goxtk.com
-
https://github.com/FNNDSC/ami
team
- Jorge Bernal
- Nicolas Rannou
- Rudolph Pienaar
- Yangming Ou
- Daniel Haehn
- Daniel Ginsburg
- Ellen Grant
BCH Team
BU Team
- Orran Krieger
- Ata Turk
- Aditya Awalker
RedHat
- Dan McPherson
- and others!
fin!
Thank you!
FTC_2017
By Rudolph Pienaar
FTC_2017
Future Technology Conference 2017
- 1,096