38
ARTIFICIAL INTELLIGENCE IN MEDICAL EPIDEMIOLOGY PREDICTING DISEASES. SAVING LIVES. DR DHESI BAHA RAJA (MD, MPH, DrPH) Twitter: @DhesiMD

Predicting diseases. Saving lives. - UNCTAD | Home · Predicting diseases. Saving lives. Author: DHESI BAHA RAJA Subject: Presentation Keywords: Predicting diseases. Saving lives

  • Upload
    others

  • View
    10

  • Download
    0

Embed Size (px)

Citation preview

ARTIFICIAL INTELLIGENCE IN MEDICAL EPIDEMIOLOGY

PREDICTING DISEASES.

SAVING LIVES.

DR DHESI BAHA RAJA (MD, MPH, DrPH)

Twitter: @DhesiMD

MALARIA

DENGUE

MEASLES

FLUHIV/AIDS

CHIKUNGUNYA

EBOLA

CVD

17.3M

deaths

TB

GONORRHEA

CHOLERA

ZIKA

POLIO

DENGUE 2.5B

at RISK

TB

1.5 Million

Deaths

HIV/AIDS

40M

Infected

FLU

100

Million

Deaths

EBOLA

90%

fatality

GONORRHEA

498M Cases

MALARIA

219

Million

CHOLERA

4 Million

Cases

400M denguecases

ANNUALLY

DENGUE & ZIKA WORLDWIDE

2.5 BILLION

AT RISK >1MZikacases

SO FAR

Image Source: HealthMap

BRAZIL

$>3.5 B

DENGUE & ZIKA ECONOMIC IMPACT

$1.3 BDENGUE

ZIKA

CURRENT RESPONSE

PASSIVE 76.90%

SENTINEL

19.20%

ACTIVE

3.90%

OUTBREAK MANAGEMENT

UNKNOWN

OUTBREAK

AREAS

UNPLANNED

MANAGEMENT

HEALTH

CAMPAIGNS

FOGGING &

LARVICIDING

GENETICALLY

MODIFIED

MOSQUITOES

Tech Approach: Dengue-ZIKA EPIDEMIOLOGY + AI

WEATHER DATA

ENVIRONMENT DATA

HUMAN MOVEMENT

DATA

EPIDEMIOLOGY DATA

COMPARATIVE DATA

272 VariablesLakes

Parks

Constructions Sites

Weather temperature

Sunrise time

Dengue & Zika cases

First Day of Symptoms

Humidity

Wind speed & direction

Cloud coverage

Dew Point

Pressure

Visibility

Feature Selection

30 Days Prediction

14 Days Prediction

7 Days Prediction

GOVERNMENTS &

MULTILATERAL ORGANIZATIONS

PUBLIC HEALTH OFFICERS

(VIA ANALYTICS INTERFACE)

USERS OF THE COMMUNITY (VIA

APP)

BACKED BYUSED BY

PUBLIC HEALTH

OFFICE

HEALTH

CENTER

[SYSTEM]

Remote data input

interface (REDINT)

CENTRALIZED

DATABASE [SYSTEM]

- Data Analytics Interface

(DAI)

- Data Prediction Interface

(PINT)

- User mgmt and operation

mgmt interfaces (UMI and

OMI).

AIME MAIN SYSTEM

Within 23 Seconds We Tap 272 Variables For Prediction

PREDICTION REAL

TIME

LOCATION &

MAGNITUDE

ANALYTICS

EPID CURVE &

REPORT

3. Addition of a Reward System.For each validated and confirmed

breeding site uploaded, users would

be able to obtain “Webe Points”.

These points could then later by

exchanged by rewards set by telco /

local / state government to the

Community.

Targeted Health Campaign

PICTURE UPLOADED

REWARD A

“POINT”DO

NOTHING

IS IT

VALID?

Y

E

S

N

O

AIME MOBILE SYSTEM

AIME’S ACHIEVEMENTS GLOBALLY

GLOBAL TRACTION

Weather HotspotsDengue Construction

Bayesian Network

Bayes Model

Evidence Instantiation

GLM

MapReduce

Regression Coeffs

Table

GLM

Predict

Single

Tree

Drive

Decision Tree

Table

Single

Tree

Predict

Predicted Cases Outbreak Likelihood Hotspot Likelihood

Prediction

DISEASE ANALYTICS

93.0%

Sensitivity

79.3%

Specificity

VALIDATED

FIELD

TESTED88.6%

Accuracy

84.11%5 days of Data Analysis

With Limited Resources

Accuracy of

DISEASE ANALYTICS

30 Days Prediction

84.87%

MANILA, Philippines

ZIKA

PREDICTION

87.45%

AIME PILOT WITH RIO CITY HEALTH DEPARTMENT

MANILA, PHILIPPINES

JULY, 2016

AIME Inc WITH THE GOVERNEMT OF PHILIPPINES

CALIFORNIA, USA

MARCH 2016

AIME AWARDED FOR ITS EFFORT FROM

THE CLINTON FOUNDATION

LONDON, UK

APRIL 2016

AIME, WINNER OF KING’S COLLEGE LIFE

SCIENCE AWARD

MASSACHUSETTS, USA

JULY, 2016

AIME AS THE GLOBAL

TOP 8 HEALTH INNOVATOR AWARD

Top 10 Exceptional Scientist & Innovations of the United Nations Solution

Summit during the United Nations General Assembly 2016, New York

AIME Inc invited as EXPERTS

to lead Vaccine Deployment site in United States of America & Latin America

FORBES TOP 40 WORLD CHANGERS

LARVA: 75.5%

AEDES: 92.2%

@AIMEforLIFE

FOLLOW US