Contemporary philippine arts from the regions_PPT_Module_12 [Autosaved] (1).pptx
Big Data in Multiple Sclerosis
1. E. Convolutional Neural Networks (CNN)
FLAIR
Fully connected layers Output
Cortical lesion98%
Periventricular lesion75%
Infratentorial lesion10%
Spinal lesion2%
Input Convolution + pooling layers
B. Additional supporting data
Multiple Sclerosis
patients
Robles®
C. Precision medicine D. Personalized medicine
Internet
•Social media
•Web-base applications
•Expert Pacient Programme
•Patient empowerment
Administrative
•Healthcare delivery
•Computer support
e-Records
•Electronic medical records
•Primary care records
•Registries (medical processes)
•National and international registries
•Self-reported patient surveys
•Personal databases
A. Healthcare-relevant Big Data in MS
Clinical
•Disability
•Cognitive impairment
•Relapses
•Symptoms related
•Age at onset?
•Gender?
Radiological
•Number of lesions
•Spinal lesions
•Cortical/yuxtacortical lesions
•Gadolinium enhancement
•Atrophy
Environmental
•Epstein Barr virus
•Cigarrete smoking
•Vitamin D
•Sun exposure
•Latitude gradient
•Microbiote?
Laboratory
•Oligoclonal bands (IgG/IgM)
•Chitinase 3-Like 1
•Light chain neurofilaments
Genetic
•HLA-DRB1
•Non-HLA variants (≈ 200)
Epigenetic
•Micro RNAs?
•DNA methylation?
•Patient stratification
•Improved diagnosis and prognosis
•Prediction of therapeutic outcome
•Novel therapeutic strategies
•Drug repositioning
•Personal
Big data
•Real-time
monitoring
•Biometrics
Improved
clinical
decision-making
Disease
prevention
Healthcare
quality
assurance and
cost-effectiveness
Improved
health
maintenance
Figure 1. A to D. Schematic representation of how the use of healthcare-relevant Big Data in MS could lead to the application of
Precision and Personalized medicine in the routine clinical practice. E. Schematic representation of a Deep Learning approach.
For example, from a FLAIR sequence, the Convolutional Neural Network is able to recognize the location of the lesion. This
same process merged with clinical, laboratory and genetic data, would result in a powerful prediction tool.
Environmental
•Epstein Barr virus
•Cigarrete smoking
•Vitamin D
•Sun exposure
•Latitude gradient
•Microbiote?
Genetic
•HLA-DRB1
•Non-HLA variants (≈ 200)
Epigenetic
•Micro RNAs?
•DNA methylation?
Catalonia: 400 new patients by year
Girona: 35 new patients by year
2.5 MILLIONS
Worldwide 700,000
Europe
47,000
Spain
7,000
Catalonia
900
Girona
Global MS prevalence
65-125 cases / 100,000 inhabitants
Predictive Models Based on Deep Learning using Radiological,
Clinical, Laboratory and Genetic Data
Big-Data in Multiple Sclerosis:
Robles-Cedeño, René1,2,3
; Perkal Rug, Héctor2,3
; Quintana Camps Ester2,3
; Gich Fullà, Jordi1,2,3
; Ramió-Torrentà, Lluís1,2,3
1
Department of Neurology, Dr. Josep Trueta University Hospital and Santa Caterina Hospital, Girona-Spain.
2
Girona Neuroimmunology and Multiple Sclerosis Unit, Neurology Department, Dr. Josep Trueta University Hospital. Biomedical Research Institute (IDIBGI), Girona, Spain.
3
Neurodegeneration and Neuroinflammation Group. Girona Biomedical Research Institute (IDIBGI), Girona-Spain.
uniem.girona.ics@gencat.cat @UNIEMGirona
1
Multiple sclerosis (MS) is a chronic and disabling disease
of the central nervous system (CNS) affecting young adults
and inducing a progressive deterioration in the motor
and cognitive spheres.
2 Pathogenesis of MS is not completely known.
The interaction of environmental and epigenetic factors
in genetically susceptible individuals would trigger the disease.
3 Many efforts are focusing on identifying biomarkers
that allow us to optimize the management and
improve the prognosis.
4
Our hypothesis here is that, by using the available computer power together with a new
set of Artificial Intelligence and Big Data techniques (Deep Learning), new biomarkers
and deep predictive models of the disease evolution can be proposed using not only
those from rich MRI descriptors but also fusing them with available environmental,
clinical, laboratory and genetic information.
5
In MS, the results of this project could give us the keys to start the use of such Deep Learning techniques in the routine clinical practice
Advances in this line will allow us to transfer such knowledge into predictive models about the evolution of the disease or the probability
of response to a particular drug or medical intervention which could markedly change the current paradigms of diagnosis, prognosis
and treatment of MS in the near future.