Big data analytics in healthcare: promise and potential for rare disease

Big data analytics in healthcare: promise and potential for rare disease Chedy Raïssi, INRIA - France

10th CRISU-CUPT International Conference

29 October 2015

Context

10th CRISU-CUPT International Conference

29 October 2015

Context

10th CRISU-CUPT International Conference

29 October 2015

Experimental platforms

10th CRISU-CUPT International Conference

29 October 2015

Motivation There is an increasing use of the Web in events of overall interest such as politics, sports and health. Smartphones and connected objects are almost personal medical devices. Use it to monitor heart rate, diet, exercises. Technology will only get smarter.

What are the implications of this rise of digital technology in healthcare? Relation to climate change?

Our goal: qualify, quantify, understand and summarize content being exchanged, stored in various ways and evaluate the impact on specific healthcare events.

10th CRISU-CUPT International Conference

29 October 2015

Outline Today: 2 case studies

Dengue Web Observator y

10th CRISU-CUPT International Conference

Orphan Disease Analytics Platform

29 October 2015

Background on Dengue • Dengue: a mosquito-borne infection - causes a severe flu-like illness - sometimes a potentially lethal complication - approximately 2 billion people at risk (> 100 countries), 50 million infections

• Outbreaks tend to occur every year during the rainy season - but there is a large variation of the degree of the epidemic in areas with similar rainfall

• Current strategies for prediction of dengue epidemics - surveillance of insects - outbreaks detection may take a few weeks - loss of precious time to address the epidemic

10th CRISU-CUPT International Conference

29 October 2015

Dengue Web Observatory

Analyze dengue epidemics manifestations in Twitter for surveillance.

Design and implement an active surveillance framework - analyzes how social media reflects epidemics - based on a combination of four dimensions

volume, location, time and public perception. Predict?

10th CRISU-CUPT International Conference

29 October 2015

Dengue Web Observatory Methodology steps • Content analysis (NLP) • Correlation analysis • Spatio-temporal analysis • Surveillance

Determine the sentiment categories • Personal experience: “You know I have had dengue?” • Ironic/sarcastic tweets: “My life looks like a dengue-prone steady water” • Opinion: “the campaign against dengue is cool” • Resource: “Dengue virus type 4 in circulation” • Marketing: “Everybody must fight dengue. Brazil relies on you”

10th CRISU-CUPT International Conference

29 October 2015

Dengue Web Observatory Methodology steps • Content analysis (NLP) • Correlation analysis • Spatio-temporal analysis • Surveillance

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine A rare disease, also referred to as an orphan disease, is any disease that affects a small percentage of the population. « Rare diseases are rare, but rare disease patients are numerous » - Estimation: 3 million patients in France

• The project is based on the analysis of the Orphanet ontology.

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine Simple and intuitive visualisation for each disease

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine Network visualisation and analysis

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine PubMed search

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine Differential diagnosis • based on a log linear model • takes into account medical observation of symptoms

10th CRISU-CUPT International Conference

29 October 2015

OrphaMine

10th CRISU-CUPT International Conference

29 October 2015

Terima Kasih, ขอบคุณครั บ, Merci, ً‫شكرا‬

10th CRISU-CUPT International Conference

29 October 2015