Using Computational Back-ends for Artificial Intelligence in Childhood Cancer Research

Use of Computational Back-ends for Artificial Intelligence in Childhood
Cancer Research
H2020 EU PROJECT | Topic SC1-DTH-07-2018 | GA: 826494
Ignacio Blanquer | UPV-I3M
16/07/2019

PRIMAGE is one of the largest and more ambitious European research projects in
medical imaging, artificial intelligence and childhood cancer.
The project is funded with 10 million euros by the European Commission, 16 European
institutions are participating in the consortium and has an implementation duration of
4 years. Internationally recognized researches in in-silico technologies and clinical
experts in pediatric cancer are part of the staff of PRIMAGE.
2
What is PRIMAGE?

PRIMAGE proposes a
cloud-based platform
to support decision
making in the clinical
management of
malignant solid
tumors, offering
predictive tools based
on the use of novel
imaging biomarkers, in-
silico tumor growth
simulation and
machine-learning.
3
What is PRIMAGE?
PRIMAGE project is devoted at developing methods of computational analysis of medical images
applied to child cancer.

4
PRIMAGE architecture
Application
Manager
Application
templates
Application
containers
Run an
HTC/MPI/AI
job
Job Id & Job
access if
interactive
Autobuild
Application
code
Unity tests
Provide local access
to data
Error Report
Get Output
Manage jobs
Provide access
to data
API Call
API Return
Internal Interaction
PRIMAGE Service
External Service
Processing Objects
Storage Objects
Job exec
service
GPUs
StorageId,
Access Token
Certification
Local storage StorageId,
Access Token
CertificateId
Repository
Certificate
Request
Data
Manager
Access history
PROMETHEUS
Job exec
service

5
PRIMAGE Use Case
Job exec
service
Distributed
Tensorflow
Classifier Unity tests
Autobuild and
test Service
Private
Registry
Storage
Input
Data
API Call
Internal
Interaction
PRIMAGE Service
External Service
Processing Objects
Storage Objects
1
2
4
5
6
7
38
Trained
Model
GPUs GPUs
● Model training is computationally
intensive and requires GPUs.
● A Data Scientist (DS), writes the code
for the training and error estimation of
the classifier and stores it (1) in a
private repository.
● The commit triggers (2) the autobuild
system to build a container image (3),
which is stored in the private registry.
● The DS submits the training job (4).
● The application runs on the tagged
resources provided of GPUs (5) which
pull the images (7) and gets the data
on a volume mounted from the shared
storage (6).
● At the end of the process, the output
model is stored in the shared storage,
accessible from the user’s console (8).
Data Scientist

THANK YOU
Ignacio Blanquer | iblanque@dsic.upv.es
PRIMAGE aims at developing state-of-the-art and cutting-edge research tools for
building up clinical decision support systems.
Such research tools will leverage HPC resources and cloud infrastructures for their
development and operation.
PRIMAGE is a clinical-led project technically coordinated by an SME leading the
application of AI technologies in medicine.
More information:
www.primageproject.eu/

Using Computational Back-ends for Artificial Intelligence in Childhood Cancer Research

Related slideshows

More Related Content

Using Computational Back-ends for Artificial Intelligence in Childhood Cancer Research