Large Scale Health Telemetry and Analytics with MQTT, Hadoop and Machine Learning DSLs

Large Scale Health Telemetry and Analytics with MQTT, Hadoop and Machine Learning DSLs

Murali Kaundinya & Gopinath Janakiraman

July 11, 2016

Agenda

• Overview

• Technologies in scope

• Platform Architecture– Context for demos

• Telemetry & Visualization with DSL (Demo)

• Regression/Confidence on Allergies with DSL (Demo)

• Predictive Analytics with DSL (Demo)

• Summary

2

3

Murali Kaundinya

Applied Technology @ Merck

Gopinath Janakiraman

Applied Technology @ Merck

Acknowledge contributions from …

• Jakub Kotowski

• Semion Andreevich Piskarev

4

5

Overview

Digital Health Trends

6

Accele

ration o

f D

igital Technolo

gy

Wireless

Sensors

Mobile

Connectivity

Social

Networking

Genomics

Internet

Imaging

Data Universe

Health Information

SystemsDisease

Diagnosis

Management

Prevention

Prediction

Time

Digital Health – Trends

7

Predictive Analytics can identify at-risk patients

8

Study found 55% of patients predicted as “highest risk” were admitted within 6 months

Opportunities with Healthcare Wearables

• Devices that drive better outcomes will thrive.

• The KPIs are:

– Increasing quality of care.

– Lowering cost.

– Decreasing hospitalizations.

• Chronic diseases can benefit the most.

• FDA regulation increases reliability and quality.

9

Technologies in scope

10

Sensors

Devices

Wearables

PHRs

EMRs

Aggre

gato

rs/D

ata

Stre

am

s

Ingestion

Stratification

Analytics

Query

ing / V

isualiz

atio

n / A

PIs

DSLs

Meta-Programming

System

M2Ms

WEKA

HBase

Patient

Portal

Provider

Portal

Payer

Portal

Care

Coordinator

Portal

11

2009 Flu Pandemic

12

Demos

Wearable Devices

• Experiences with

– Fitbit

– Misfit

– Apple’s Research Kit

– Google Fit

– Microsoft Band

• Experience with device portals

– Validic, Data Minded Solutions, Human API, REDOX

13

Telemetry and Visualization with DSL - Demo

14

Regression/Confidence with Allergies - Demo

15

Predictive Analytics with DSLs- Demo

16

In closing …

• IoT – Telemetry - Easier to embed, integrate

– More devices, generating more non-standard data.

• Discoverable data sources (internal and external)

– Machine toolable

• Domain Specific Languages

– Declarative programming w/ projectional editors

• Abstract away complexity

– Compute, Analytics/Machine Learning

• Visualize data

17

What’s next…

• Internet of Medical Things

– More devices, generating more non-standard data.

• Consent and sharing

– Privacy, Compliance

• Interoperability with EMRs

– FHIR

• Precision Medicine

– Genomic Sequencing, Personalized Medicine

• Population Health

– Longitudinal data++, disease models, preventive care

18

Next Steps

• Welcome community development!

• [email protected]

• [email protected]

19