1) An AI system implemented at Johns Hopkins Hospital helped optimize hospital operations and bed assignment. It allowed beds to be assigned 30% faster.
2) This reduced the need to keep surgery patients in recovery rooms longer than necessary by 80% and cut wait times for ER patients to receive beds by 20%.
3) The efficiencies also allowed the hospital to accept 60% more transfer patients from other hospitals.
Axa Assurance Maroc - Insurer Innovation Award 2024
Bio IT World 2019 - AI For Healthcare - Simon Taylor, Lucidworks
1. To AI or Not to AI,
That Is the Question
Simon Taylor, VP Global Partners & Alliances,
Lucidworks
2. Concerns about AI in Healthcare?
1. Bad data in = bad recommendation
out. False positives reducing
confidence in machine learning
2. Longer term complacent
dependencies on “infallible” machine
learning
3. In ability for AI to take account of the
wider context of patient needs /
treatments
HEALTHCARE IT NEWS - JANUARY 2019
3. Narrow “Weak” vs. General “Strong” AI
Artificial Narrow Intelligence (ANI)
is the AI that exists in our world today,
programmed to perform a single
task — whether it’s checking the
weather, being able to play chess, or
analyzing raw data to write journalistic
reports.
Artificial General intelligence (AGI) or
refers to machines that exhibit human
intelligence. In other words, AGI can
successfully perform any intellectual task
that a human being can being conscious,
sentient, and driven by emotion and self-
awareness.
ANI systems are able to process data
and complete tasks at a significantly
quicker pace than any human being
can, which has enabled us to improve
our overall productivity, efficiency, and
quality of life e.g. assist doctors to
make data-driven decisions, making
healthcare better
AGI is expected to be able to reason,
solve problems, make judgements
under uncertainty, plan, learn,
integrate prior knowledge in decision-
making, and be innovative,
imaginative and creative.
4. …agile organizations
insist on full transparency of
information, so that every
team can quickly and easily
access the information
they need...
MCKINSEY, JANUARY 2018
Organizations possess
lots of data...siloed inside
disconnected applications, and
unavailable to employees at
their moments of need.
FORRESTER, JULY 2018
85% of employees
are not engaged or actively
disengaged at work.
GALLUP, 2017
5. Administrative and operational
inefficiencies account for nearly one
third of the U.S. health care system’s
$3 trillion in annual costs.
HARVARD BUSINESS REVIEW, HARVARD BUSINESS SCHOOL
NOVEMBER 2018
6. 6 out of every 10 people who work in
health care never interact with patients.
Even those who do can spend as little as
27% of their time working directly with
patients. The rest is spent in front of
computers, performing administrative
tasks.
HARVARD BUSINESS REVIEW, HARVARD BUSINESS SCHOOL
NOVEMBER 2018
7. The Time it Takes to Get things Done
BigDataManagement
&Analytics
Data Scientist Driven Activity Business Alignment
❶ ❷ ❸ ❹ ❺ ❻ ❼ ❽
Meet the “Last Mile Problem”
10. Let’s Focus on Where AI is Helping Healthcare
• MRI / CIT Image Diagnosis
• Hospital Operations Decision Making
• Exploring Clinical Pathways
• Patient Risk Management
• Automation & Exploration of Clinical Documentation
• Ontology Base Search
• Fraudulent Claim Detection
• Competitive Drug Go-to-market
MEDIACITY NEWS, JONATHAN MUSE
JUNE 2018
11. Complex
Oncology solutions for
decision based personalized
medicine & patient care
LARGE VOLUMES OF LEGACY DATA OVER LONG TIMELINES
Easier
Optimization of hospital
operations to predict
demand for additional ED
capacity
CORRELATION OF REAL-TIME INPATIENT DATA
AI Healthcare Decision Paths
12. Machine learning is a method of data analysis
that automates analytical model building. It is a
branch of artificial intelligence based on the
idea that systems can learn from data, identify
patterns and make decisions with minimal
human intervention.
SAS INSTITUTE INC. 2019
13. Many AI tools’ designs
mainly focus on data, not
human. Thus, a great AI
platform is usually solution
based leveraging search rather
than just tool based.
AI AUTHORITY
CHAO HAN, HEAD OF DATA SCIENCE, LUCIDWORKS
SEPT 2018
15. • Open source tech at its
core: Apache Solr & Apache
Spark
• Personalizes work with
applied machine learning
• Hardened on the biggest
corporate & government
information systems
16. Integration requires:
• Comprehensive data access
• Integrated end-to-end security
• Classification & categorization
The Integration Challenge:
Securely access, ingest & synchronize data
in real-time, and at massive scale
17. How to Eliminate data silos and reducing tribal wisdom
FEATURE BENEFIT IMPACT
Entity Extraction
Classify and categorize items
such as people, places, and
things from unstructured text
Improve search precision so people can do their
jobs better with less error
200+ connectors and
advanced Connector SDK
Real-time access to billions of
documents across storage
systems and file formats
Increase: access to precise information and
employee productivity. Decrease: frustration,
time to market, and time to value.
Integrated security
Reduce risk by meeting global
compliance requirements
Securely administer access controls by
integrating AD, SSO, and Kerberos with support
for field, document, and user level restrictions
1
2
3
18. Curate the experience by:
• Understanding the way your analysts
access and pivot around data
• Automatically tuning relevancy through
ML
• Personalizing predictive outcomes and
insights with AI
19. FEATURE BENEFIT IMPACT
Self-learning, real-time
recommendations engine
Help users when they don’t
know what they are looking for
Increases employee productivity. Decreases:
frustration, time to value and time to market
Natural Language Processing
and Rules based NLP suite
Parse and process normal
language queries so users don’t
have to learn complex
operators or functions
Allows users to search like they speak
augments individual and organizational
intelligence
Content and signal-driven
automatic relevancy
Personalized results improve
with each click as the system
learns from user activity
Connects people to insights, content and
each other, promoting collaboration,
reducing frustration and increasing
innovation
What you need to improve user engagement
1
2
3
20. The Exploration Challenge:
Connect users to insights
when they need them most
Explore with:
Superior user experience
Real-time indexing
Hyper-personal relevance
Performance at scale
21. How to Accelerate organizational agility
FEATURE BENEFIT IMPACT
Modern distributed
architecture, built on Apache
Solr and Spark
Reliable performance
at scale in real-time
Deliver highly accurate, relevant results
from billions of documents and
thousands of queries per second (QPS)
Signal-driven relevancy Personalized results increase
productivity, reduce frustration
Tailor results in real time with user info
like role, dept, location and expertise
Fusion App Studio
Let developers quickly create apps
in mins with prebuilt templates
Rapid time to value
1
2
3
22.
23.
24. advanced connectors and AI enrichment,
surfaced by one or more applications via App Studio
AIAI AI
Fusion
Connect users to
insights precisely at
their moment of need
any format, any platform
System Generated
Human Generated
Application Generated
Data
Digital Workplace
Solution
25. The Time it Takes to Get things Done
BigDataManagement
&Analytics
Data Scientist Driven The Last Mile Problem
Direct Alignment with Business NeedsQuantifiable & Faster ROI
Reduced Time to Value
Search,Discovery&
OperationalAI
❶ ❷ ❸ ❹ ❺ ❻ ❼ ❽
❶ ❷ ❸ ❹
26. Surface the insights that matter most,
with ML & NLP
Recommenders give every user
a customized experience
Machine learning models pre-
tuned and ready to use
Classifiers for precise
understanding of intent
Clustering and anomaly detection
for discovery
Experiment more, code less
Bespoke, data-anywhere search and
discovery apps for all devices
Data-driven apps created in hours, not
weeks or months
Modular, pre-built components for
repeatable, predictable outcomes
Tried, tested & proven modules allow
iterative, prototype-led production
Build applications more rationally by
starting with real data
Supports over 25 data platforms
Highly scalable search engine and
NoSQL datastore
Trillions of data objects - any
source, any type
1000s of queries per second from
1000s of concurrent users
Full text search, SQL capabilities
End-to-end inherited and embedded
security
Extensible
28. Don’t like math?
Out of the box, automated clickstream relevance tuning
Best-in-class recommendations simplify
personalization, reduce system latency and simplify
operations
Data quality algorithms preemptively find issues in your
data before your customers do
Built-in query analytics automatically identify poor
results and bad queries and suggest fixes
Automated synonym generation simplifies
management
Let us do it…
29. Like math?
Fusion users have full and complete access
to core search and ML algorithms
Built-in support for most popular data
science tools like Jupyter, Zeppelin, SQL,
Python, R and SAS
One stop shop for building and deploying
machine learning models
Common data prep functions for data
science and engineering are out of the box,
significantly speeding up model building
Be our guest…
30. Being Data Driven is not
optional
Experiment management that goes beyond A/B
testing and is optimized for search
Relentlessly track and measure relevance
automatically
Dissect the “why” through our in-depth query
analytics workbench
Capture, search, analyze and leverage user
feedback all from within Fusion
32. [Using a new AI system], Johns Hopkins
can assign beds 30% faster. This has
reduced the need to keep surgery
patients in recovery rooms longer than
necessary by 80% and cut the wait time
for beds for incoming ER patients by 20%.
The new efficiencies also permitted
Hopkins to accept 60% more transfer
patients from other hospitals.
HARVARD BUSINESS REVIEW, HARVARD BUSINESS SCHOOL
NOVEMBER 2018