IBM Haifa Research Lab: Let's Build a Smarter Planet

lavishgradeΛογισμικό & κατασκευή λογ/κού

25 Νοε 2013 (πριν από 3 χρόνια και 10 μήνες)

88 εμφανίσεις

IBM Haifa Research Lab




S
PA


a Platform for

Hereditary Disease Management

and
P
edigree
A
nalytics

Rizzoli FOAK


Alex Melament, melament@il.ibm.com

July 2010

© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

1

IOR
-

Rizzoli
Orthopaedic
Institute


IOR
-

is the main Italian institute of orthopedics

and has a status of a 'Scientific research hospital’

-
About 150,000 patients examined every year

-
Over 18,000 orthopedic operations every year

-
Nine laboratories at the institute employing a staff

of 250 including doctors, biologists and technicians.


Medical Genetic Unit specializes in Rare Skeletal Hereditary diseases
such as:

-
Multiple Osteochondromas (MO)


Osteochondroma is a cartilage capped bony projection arising on

the external surface of bone containing a marrow cavity that is

continuous with that of the underlying bone

-
Osteogenesis Imperfecta (OI)


It is frequently caused by defect in the gene that produces type

1 collagen, an important building block of bone. There are

many different defects that can affect this gene
.


© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

2

Project Goals


Understanding the relationships between hereditary
diseases and their genetic background

-
Analysis of inherited diseases and their associated
phenotypes is of great importance to gain knowledge of
underlying genetic interactions

-
Discovering and defining a correlation between Phenotype
and Genotype data will enable


Fit adequate treatment protocol


Ensure appropriate clinical follow
-
up


Improve patient’s quality of life


Build Healthcare Platform that will enable efficient
treatment and productive research in Hereditary
Diseases


© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

3

IT for understanding of Hereditary Diseases


Collects and Integrates medical images, clinical and genomic data of
each patient and his family

-
Leverages standards such as DICOM and HL7

-
Maintains content and context aware associations to support clinical and
research usage


Enables secure cross Hospital data and knowledge sharing

-
Leverages Industry’s Best Practices, Standards and IHE Profiles

-
Enables cross Hospital Patient ID correlation through PIX&PDQ interfaces


Enables data insights discovery

-
Supports federated queries

-
Provides on demand pedigree visualization

-
Provides a platform for data analytics & knowledge extraction


Enables to host third party analytics



© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

4

Pedigree Analysis


The pedigree documents biological
relationships in families and the
presence of diseases.


-
Pedigree includes number, gender and
closeness of affected relatives, their ages at
disease onset, and associated health
conditions


Pedigree is needed to

-
Assess disease risk


BRCARPO risk model can indicate a
chance of having BRCA1 or BRCA2
mutations

-
Investigate correlation between Phenotype
and Genotype data

©
2009
IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

5









Solution Architecture

Healthcare BUS (WAS7/WPS 6.1, HL7 v2&3, WS, IHE profiles)

Hospitals (PACS, Clinical and Genetic Labs)

Cross River/Rimon

(
Phenotype,Genotype)


DICOM

HL
7
v
3
, v
2
.x

WAS 7, IBM CM


PIX/PDQ

Server


CMO

WAS 7, DB2 Data Warehouse

WAS 7

Data Federation



WAS 7, DB2 Data Warehouse

+

EHR,

Knowledge

Systems

WAS 7

WS

Pedigree Analytics Cognos Excel BO

Physicians and

Researchers

WS









Partners (Disease risk assessment models, Pedigree clustering, classification and visualization tools)









Cube Services

&

IBM BI

Infrastructure

© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

6

Pedigree Visualization


Dynamic pedigree visualization


Presentation of all available
information for the persons in the
pedigree

-
Clinical, genomic data and medical
images


Standard pedigree representation

-
HL7 v3 Family History

-
Enables standard based pedigree
interoperability


Enables disease risk assessment




©
2009
IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

7

On Demand Analytics Platform


Currently Available Algorithms


Decision Trees



Explanation of one dimension by others; Clustering

-
Random ID3

-
C4.5 (entropy based)


Bayesian Networks




Data Cleansing

-
Naive Bayes

-
Chu
-
Liu Trees


K
-
Means



Clustering

-
Hamming distance for category dimensions only

-
Continuous dimensions only

-
Discretization of Continuous dimensions by Entropy and application of
Hamming distance on all the dimensions.


Statistical Analysis

-
Chi
-
Square


Association between two category dimensions

-
Spearman


Association between two category dimensions



© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

8

CMO


SOA based Medical Imaging Repository


Supports DICOM, WADO, HL7,

XDS
-
I, PIX and PDQ



Provides secure cross
-
enterprise
sharing



Complies with regulations for
storing and managing data



Enables extensibility and reuse of
existing legacy assets through an
open and flexible architecture



Leverages IBM's market
-
proven
middleware to ensure scalability,
high availability and disaster
recovery


Data Sources

(CT, MRI, US…, PACS)

Image

Consumers

Fine
-
Grained

Authorization

Data

Analytics

Image

Processing

Imaging

Archive

© 2009 IBM Corporation

IBM Haifa Research Lab: Let’s Build a Smarter Planet

9

9