Machine
Learning (ML) and Artificial Intelligence (AI): Impact of AI/ML in Healthcare: Part-Eight
by
Dr. RGS Asthana
Senior Member IEEE
Figure 1: Some of the ML/AI driven
healthcare companies [27]
Summary
What really marks healthcare
different from other disciplines [36]?
Healthcare may often have very little labeled data (e.g., clinical NLP).
This may prompt the use of semi-supervised learning algorithms i.e. keeping
human in the loop (HITL) [38]. Sometimes, we have only small numbers of samples
(e.g., for a rare disease) and we need to learn as much as possible from other
data (e.g. EHR data of healthy patients). We may have lots of missing data that
too at varying time intervals and may only get censored labels. Other more
important problem which we need to solve is that ML base algorithms do not give
reason for arriving at a particular decision. Therefore, it is pertinent to model
the problem keeping these aspects in view and may be reason for HITL in the
solution.
ML based
solutions are good at prediction and diagnosis too is a prediction in a way. We,
therefore, describe ML based diagnosis and treatment systems. The only thing necessary for systems to give
better prediction is training on substantial data. The areas where ML/AI based systems have impact in healthcare are:
on-line consultations, Health assistance and medication
management, Personal genetics, development of drugs of the future, discovering
new diseases, persistent care, discovering new clinical pathways and last but
not the least Robotics and Healthcare.
Keywords
Machine
Learning (ML) Tools, Artificial
Intelligence (AI), Neural
Networks, Internet
of Things (IoT), Deep Mind, IBM’s Watson
Prelude
It has always been our
goal to catch health-related issues early, before they pose some real problems,
for instance, tracking the risk factors that are related to diabetes before a
person actually develops diabetes.
Big Data and ML/AI will transform our lives as today
we have a lot of data which is partially made
available for research in anonymized form and it looks that plowing it into ML
algorithms to build predictive models should be easy. However, first major
challenge is that this data is raw and needs to be pre-processed. The second major challenge is in combining
various databases such as shopping behavior, daily exercise and diet info,
maybe personal genomes and also occasional blood test data.
ML/AI in healthcare and medicine can optimize
patient treatment plans as well and also provide physicians with the
information to make a good decision.
IBM’s Watson is exploiting cognitive computing for healthcare in a big
way.
The
quality of the solution is judged from the facts whether the solution is able
to transform medicine not only by making it accessible to the poorest medical
institutions or also to a handful of experts because they are too difficult to
use. AI already found several areas in healthcare to transform starting from
the design of treatment plans through the assistance in repetitive jobs
to medication management or drug creation. Dell, Hewlett-Packard, Apple,
Hitachi Data Systems, Luminoso, Alchemy API, Digital Reasoning, Highspot,
Lumiata, Sentient Technologies, Enterra, IPSoft , Next IT are a few leading names
in healthcare.
The
most obvious application of ML/AI in healthcare is data management which
includes processes like collecting, storing, normalizing and tracing family history of data. Google recently acquired
Deepmind [11] – a AI company - to begin with they have aligned with the Moorefield’s
Eye Hospital NHS Foundation Trust [29].
IBM
Watson [12] provides oncologists/clinicians evidence-based treatment options. It has advanced ability for oncology to
analyze context of structured and unstructured data in clinical notes and generate
reports that may be useful to picking a treatment pathway.
Medical
Sieve [15] is an ambitious project of IBM to build a “cognitive assistant” with
analytical, reasoning capabilities and a wide range of clinical knowledge. The
idea is that routine images are seen and analyzed only by Medical Sieve and
Radiologists in the future should only look at the most complicated cases where
human supervision is useful. This medical assistant will handle both Radiology
and Cardiology cases. Deep learning can readily handle a broad spectrum of
diseases in the entire body and all imaging modalities (X-rays, CT scans, etc.)
On-line consultations
In
US and UK, you cannot meet your GP when you need him but you have to seek an
appointment first. In UK, an online medical consultation and health service
called Babylon [16] was
launched which offers medical AI consultation considering personal medical
history and common medical knowledge. This app reminds patients to take their
medication and does necessary follow up by finding its patient’s current
medical status.
Health
assistance and medication management
The world’s first virtual nurse, Molly was developed
by the medical start-up Sense.ly [17].
The virtual nurse has a smiling,
amicable face with a pleasant voice and is designed to help people with
monitoring their condition and treatment. This interface uses ML to provide
care to patients with chronic conditions in-between doctor’s visits. It
provides proven, customized monitoring and care, with a strong focus on chronic
diseases.
Personal genetics
The most significant application of AI and ML in genetics
understands how DNA impacts life and also what influences life and biology; we
need to first understand the language that is DNA. This is where ML algorithms
play a big role and big players, such as; Google’s Deep Mind [19] and IBM’s
Watson [20] are in the fray. Technologies like big data detect patterns
from enormous amounts of data (e.g. patient records, clinical notes, diagnostic
images, treatment plans).
Deep Genomics [18] is developing the capability to interpret
DNA by creating a system that predicts the molecular effects of genetic
variation. Their database is able to explain how hundreds of millions of
genetic variations can impact a genetic code. This will provide personalized
insights to individuals based on their biological dispositions.
This trend is indicative of a new era of “personalized genetics,” whereby individuals
are able to take full control of their health through access to unprecedented
information about their own bodies. As is the case with any application
of AI/ML and also Big Data, the technology must have access to vast amounts of
data in order to better curate lifestyle changes for individuals.
Develop drugs of
the future [10]
Application of AI/ML in healthcare is not only reshaping the
industry and making what was once was considered impossible task but also
helping in the reduction of both cost and time in drug discovery.
Atomwise [11]—a San Francisco-based startup—is considering to
use supercomputers instead of test tubes in the drug development procedure. The
company uses ML and 3D neural networks that search through a database of
molecular structures to find out therapies, discover the effectiveness of new
chemical compounds on diseases and identifying existing medications that can be
used to cure another ailment.
Berg Health [21]—a Boston-based Bio-Pharma company—uses a
different approach for drug discovery and works on patient biological data
using AI to determine why some people survive diseases, and then applies this
insight to improve current therapies or create new ones.
Figure 2
Cloud technology used for drug discovery
Cloud Pharmaceuticals [26] is using AI design innovations to
discover and generate new novel drugs using cloud technology (see figure 2).
There are many new companies working on drug discovery
exploiting ML/AI to the hilt to discover drugs and the future of this field is
encouraging. It may be worth mentioning that the cost and time of drug
development in the usual way varies between USD 0.5 to 3 Billion and may take
more than a decade to reach the user.
Discovering new
diseases
Genomics aims at identifying patterns in huge
data sets of genetic information and medical records, looking for mutations and
linkages to disease. A new generation of computational technologies are being
invented that can inform doctors when DNA is altered by genetic variation,
whether natural or therapeutic. Most diseases are far more than just a simple gene mutation. Although the healthcare system generated
enormous amounts of (unstructured) data—which is gradually refining and
improving in quality—but we did not possess the necessary hardware and software
in place to investigate it and produce telling results.
Disease diagnosis involves a number of factors, like the texture
of a patient’s skin to the amount of sugar level in his blood. So far, we
have worked on symptomatic detection, e.g. if one has a fever and stuffy nose,
he is diagnosed most likely to have the flu. Systems can now predict the
molecular effects of genetic variation [18], opening a new and exciting path to
discovery for disease diagnostics and therapies.
But often the arrival of detectable symptoms is too late,
especially when dealing with diseases such as cancer and Alzheimer’s. With ML,
the hope is that faint signatures of diseases can be discovered well in advance
of detectable symptoms, increasing the probability of survival and/or treatment
options.
Freenome [22, 25] — a San Francisco-based startup—is a technology
company developing proprietary algorithms and novel methods to enable accurate
diagnosis of clinical conditions. Freenome platform — an Adaptive Genomics Engine —dynamically
detects disease signatures in your blood. The company uses your Freenome —
the dynamic collection of genetic material floating in your blood and is
constantly changing over time and it’s like a genomic thermometer of who you
are as you grow, live and age. As per CEO of the company, ‘our aim is to bring
accurate, accessible, and non – invasive screenings to doctor to proactively
treat cancer and other diseases at their most manageable stages’.
Enlitic [23] uses deep learning techniques to empower doctors
with necessary data to not only become faster but also accurate. This is
possible as doctors look at disease diagnosis and treatment plans by coupling
deep learning with medical data and extract actionable insights from billions
of clinical cases. IBM’s Watson is working with Memorial Sloan Kettering in New
York to analyze data on cancer patients and treatments used over decades to
present and suggest treatment options to doctors in dealing with unique cancer
cases.
Google’s Deep Mind will scan 1 million medical records of
Moorfield’s Eye Hospital [24] and will analyze these digital scans of the eye will
be used to train an AI system to help doctors better understand and diagnose
eye disease. Google has estimated that up to 98% of sight
loss happens due to diabetes and can be prevented with timely and early
treatment based on the
needs and habits of patients.
CBI [27] insights located number
of companies involved in new programs for imaging and diagnostic using ML and
deep learning algorithms are more practiced at recognizing patterns. An IBM supported group called Pathway
Genomics [28] provides
physicians and their patients with actionable and accurate genetic information
to improve or maintain health and wellness.
It is developing a simple blood test to
determine if early detection or prediction of certain cancers is possible. Lumiata [30] is
using ML based predictive analytics tools to find accurate insights and make
predictions related to symptoms, diagnoses, procedures, and medications for
individual patients or patient groups.
Treatment
IBM’s
Watson [12, 31] helps oncologist make the best care decisions for their
patients. Watson provides clinicians with
evidence-based treatment options and also helps patients understand specific
care options available to them.
In a completely
different field, Ginger.io [32] is developing a mobile app to deliver
mental health treatments. Using this app
a person can analyze his moods and
learn strategies developed by doctors.
Persistent care
Persistent care means that quality of care as applicable remains
consistent irrespective of patient location, i.e., IPD in hospital, ICU in
hospital or at home. This can only be possible, in my opinion, when care is
automated and human being is not involved in actual care.
ML
based algorithms are best at prediction if trained properly not only on
internal data, within the control of enterprise, but also on external data that
is available and accessible on the internet [55]. With the advent of the
Internet of Things (IoT) [6], the number of sources of data has significantly
increased. As this is the only way to
get vast amount of data. Diagnosis and
treatment also fall in this domain.
Hospital re-admittance is the biggest difficulty
a doctor faces in healthcare as they do not know whether a patient is taking
prescribed medication when he goes home.
A surveillance tool from Ayadsi [57, 58]
analyzes patient’s EHR and gives clinicians’ real-time information whether a
patient experiencing shortness of breath may meet the criteria for sepsis,
chronic heart failure, pneumonia or any other condition. This acts as an early
warning system, enabling clinicians to identify patients that may be trending
toward a bad outcome. Being able to have crucial information sent to staff
enables them to save time and doctors can get involved earlier in-patient care.
AiCure [33] exploits AI for
continuous patient monitoring using mobile technology and facial recognition
technologies. It determines if a patient
is taking the right medications at the right time to help doctors confirm that
the patient is taking their medications and alert them if something goes wrong.
NextIT [34] uses conversational AI system based bots,
assistants, and any other endpoint going beyond streamlining processes for
innovating and empowering Fortune 5000 enterprises for over a decade to serve
their customers and employees better. It
has developed a digital health coach, similar to a
virtual customer service rep on an ecommerce site. The assistant can prompt
questions about the patient’s medications and remind them to take the medicine,
ask them about symptoms, and convey that information to the doctor.
The
Caféwell Concierge [35] app uses IBM’s Watson’s [12] natural language
processing (NLP) power and guides and inspires a
user to be healthier. It gives personalized recommendations to help a user to attain
optimal health and get the most out of health plan benefits. It is available on Android as well as on Apple
platforms.
Clinical Pathways
Last
year, I happen to visit a hospital as my sister was admitted there but nurses
there were not very nice. They did not care about the patients and only were
present in strength during the visiting hours in ICU. The Nurses were of
opinion that people who have money leave their patients in hospital to die. In
India 90% of nurses come only from one state called Kerala and they in general
do take care of the patients. I came in
the above hospital where nurses were local there was problem of service quality.
Doctors were good at handling problem
but service quality did not improve despite best efforts put by doctors in
front of patients. Intelligent
automation, i.e., use of virtual or ML/AI based robotic nurses is a solution to
ensure uniform and persistent care. This may have number of other indirect advantages
like reduced infections, cost reduction, less re-admissions and also reduction
in length of stay in the hospital.
Using EHR data and machine
learning algorithms, researchers have developed a cost-based framework for
identifying clinical pathways for chronic disease patients.
By creating step-by-step clinical pathways based on a
patient’s anticipated
disease development
[37], big data analytics techniques could help
providers achieve accurate predictions of anticipated future events and costs
following different clinical and cost pathways for improved shared decision
making.
Figure
3: Clinical Performance Measurement and
Standardization within units and across sites enhancing patient safety and
clinical outcomes.
The first step in
creating these algorithms is to build out Multi-stage approach for
patients:
·
Identify
the condition and factors including environmental and historical data may
affect treatment?
·
Predict risks
and the conditions of unknown variables?
·
Activate
resources needed for treatment?
·
Monitor
what clinical data can be accumulated during the treatment process? and also
·
Learn
what findings after discharge can be applied into algorithms?
Deontics pathway and
decision solutions [39] integrate individual patient data from the EMR with
guideline/protocol knowledge. If any input data items changes, the system will
instantly reevaluate the relevant evidence and patient data and render new
outputs.
All clinical data items are functionally linked with standard ontology
codes e.g. SNOMED CT– provides the core general terminology electronic health records (EHRs) [40],
RxNorm – a standard drug terminology [41], LOINC [42] - nomenclature for
clinical laboratory tests etc., offering communication and interoperability
benefits. Deontics [43] - a
world-leading AI company with Clinical Pathway and Clinical Decision Support
Systems (CDSS) products - providing the capability to fully ‘measure’ clinical
performance at both an individual and population level and offering benchmarks
for improvement (see figure 3).
Deontics systems [39, 43] have been
implemented in major US and UK hospitals.
Mercy hospital – the
seventh largest healthcare system in US - is using Ayasdi Care [60] to increase the quality of
patient care by optimizing the care pathways which are about 80 and amounts to
about 80% of care provided in most hospitals.
With Ayasdi Care's analytics and unique data-driven approach, physicians
determine optimal pathways by using reverse experimentation to identify
patterns in data derived from their local patient data, and automatically
recognize best practices that deliver the best patient outcomes.
For example, using Ayasdi analytics, Mercy Hospital identified
that the administration of a specific analgesic after arthroscopic knee
replacement surgery correlated with lower costs and shorter hospital stays
[59].
The benefits of this approach will be higher patient satisfaction, lower
readmissions, shorter length-of-stays, and also lower costs. However, each
outcome of the AI automation is analyzed by a group of physicians before being
selected for implementation/modification (if required), i.e., human is kept in
the loop.
Robotics and
healthcare
Medical robots [44] do not only exist in sci-fi movies but
they are coming to healthcare in a big way as Robots can support, assist and
extend the service health workers are offering. In jobs of repetitive and
monotonous nature robots can even completely replace humans.
Here are some of the
interesting medical robot facts:
·
Due to use of Xenex [45] Robot there is at least 70% drop in hospital
acquired infections.
·
Two Belgian Hospitals have “Hired” Pepper Robots as Receptionists [38].
·
Surgical robotics sales are expected to almost double by 2020, i.e., it
would reach up to 6.4 billion USDs.
·
Cuddly animal-shaped PARO Robot [52] reduces stress for patients. PARO is
an advanced interactive robot. It allows
the documented benefits of animal therapy to be given to patients in hospitals
and extended care facilities.
·
Mechanical aides can not only pick up groceries and take out the trash
[53] for the elderly people but also can easily be programmed to remind and
give medicine.
·
The elderly, a group indifferent to
new technology, now finds themselves as a target market for the growing
robotics industry. Intuition Robotics [54], an Israel-based startup
that makes robots, has received funding from Toyota Research Institute. The funding will be used to
develop a social companion robot, ElliQ. This robot
will be designed to reduce caregiver stress. It will give reminders about
upcoming doctor's appointments. ElliQ works to keep older adults engaged and
active personalized suggestions such as going for walks, or listening to music.
Hus it will make all effort to reduce their loneliness.
·
Intouch Health [46] has had more than 750,000 remote clinical encounters
through so far.
·
TUG Robot [47] able to carry around more than 400 kilograms of medication as it travels the
halls, rides elevators and audibly speaks while performing its tasks.
Hospitals using TUGs are not only serious about improving efficiency but also
it enhances their reputation as being on the leading edge of medical care.
·
Japanese Bear-Shaped robot [40] can lift patients from beds into
wheelchairs or help them to stand up, promising ‘powerful yet gentle care’ for
the elderly (see figure 4).
Figure 4: Robot
which can lift Patient
·
Less than a millimeter sized Microbot delivers drugs through bloodstream
[49]. The battery in any device always runs the risk of exploding. Therefore,
it became necessary to develop a device which is light powered and has no battery.
The ability to self-organize pulsing recurrent motions, such as the repetitive
flipping motion [50] is one of the fundamental characteristics of living
organisms and this can be used in the future to develop bio-inspired molecular
motors and robots that will find applications in wide areas, including
medicine. Researchers of this tech have no plans to commercialize the
technology as yet (see figure 5).
Figure 5: Small, light-powered
robots swimming around your body sounds a lot more fun that tiny, squid-like
battery-powered mechanism
·
Veebot [51] draws blood in less than a minute to make the whole procedure
of venipuncture automated to reduce error and decrease venipuncture times.
Way forward
Robots and
computers will probably never completely replace doctors and nurses but
application of ML/deep learning/AI will change the healthcare industry by
improving diagnostics, predicting outcomes, and providing personalized care and
thus changing the way doctors think about providing care which is generally to
follow the 19th century processes.
A general purpose AI system is what may be more useful
to healthcare environment provided it can always give reasoning to the
actionable info it has come up with. To safeguard all interests, either reasoning
is given for every recommendation or Human in the loop (HITL) approach is used
for by the AI Automation to get acceptance in healthcare domain. Further,
accuracy of outcomes suggested by AI system improves as it has access to more
and more data.
”By 2025, AI systems could be involved in everything
from population health management, to digital avatars capable of answering
specific patient queries.” — Harpreet Singh Buttar, analyst at Frost &
Sullivan.
Cognitive analytics with patient data will
become pre-dominant in healthcare Industry and more than one third of providers
will use it by 2018.
References
[1]
Progress and Perils of Artificial Intelligence (AI)
[2] Invited Chapter 6 - Evolutionary Algorithms and Neural Networks, Pages 111-136, R.G.S. Asthana
in book, Soft Computing and Intelligent Systems (Theory and Applications),
Academic Press Series in Engineering, Edited by:Naresh K. Sinha, Madan M. Gupta
and Lotfi A. Zadeh ISBN: 978-0-12-646490-0
http://www.sciencedirect.com/science/book/9780126464900
[3] Future 2030
by Dr. RGS Asthana, Senior Member IEEE
[4] Machine
Learning (ML) and Artificial Intelligence (AI) – Part 1, by Dr. RGS Asthana, Senior Member IEEE
[5]
Machine Learning (ML) and Artificial Intelligence (AI) – Part Two, by Dr. RGS
Asthana, Senior Member IEEE
[6] Internet of
Things (IoT)
[7] Machine Learning (ML) and Artificial Intelligence (AI): Cognitive
Services and Robotics – Part Three by Dr. RGS Asthana, Senior Member IEEE
[8] Machine
Learning (ML) and Artificial Intelligence (AI): Big Data and 3 D Printing
– Part four by Dr. RGS
Asthana, Senior Member, IEEE.
[9]
Machine Learning (ML) and Artificial Intelligence (AI): Drones and
Self-driving Cars– Part Five by, Dr. RGS Asthana, Senior Member IEEE
[10]
Machine Learning (ML) and Artificial Intelligence (AI): Healthcare– Part Six
by, Dr. RGS Asthana, Senior Member IEEE
http://newblogrgs10.blogspot.com/2017/05/machine-learning-ml-and-artificial_26.html
[11]
Deep mind website
[12]
IBM Watson Website
[13] Artificial
Intelligence Will Redesign Healthcare
[14] How
ML, Big Data and AI are changing Healthcare forever
[16] Hand-picked doctors, supported by cutting edge technology
[17] Sensely website
[18] Changing
the course of genomic medicine
[19] Atomwise
website: Artificial Intelligence for Drug Discovery
[20] How ML, Big
Data and AI are changing Healthcare forever
[21] Berghealth
website
[22] Freenome
website
[23] Enlitic
website
[24] Google's
DeepMind to peek at NHS eye scans for disease analysis
[25] Freenome
[26] Cloud
Pharmaceuticals website
[27] From Virtual Nurses to Drug
Discovery: 106 Artificial Intelligence Startups in Healthcare
[28] Pathway Genomics website
[29] Moorfield’s Eye Hospital NHS
Foundation Trust website
[30]
Lumiata website
[31]
Watson for Oncology
[32]
Ginger.io
[33]
AiCure website
[34]
NextIT website
[35]
Cafewell Website
[36]
Lecture 1: What makes healthcare unique?
[37] EHR Data, Machine Learning
Create Cost-Based Clinical Pathways
[38] Machine Learning (ML) and Artificial Intelligence
(AI): Will AI/ML intelligence surpass humans?
Part Seven by
Dr. RGS Asthana, Senior Member IEEE
http://newblogrgs10.blogspot.in/2017/06/machine-learning-ml-and-artificial.html
[39] AI Clinical Pathways/Decisions
[40] SNOMED CT
[41] Bio Medical Informatics: Discovering knowledge in Big Data
[42] Semantic Web
Representation of LOINC: an Ontological Perspective
[43] Deontics Website
[44] 9 Exciting facts about Medical Robots
[45] The Robot used by more hospitals than any other device for no-touch
room disinfection.
[46] Intouch health Website:
Virtual health anytime anywhere
https://www.intouchhealth.com/
[47] TUG Healthcare
[48] Robear: the bear-shaped nursing robot who'll look after you when you
get old
[49] Microbots: Using Nanotechnology In Medicine
[50] Light-powered microbots could deliver drugs directly to our
bloodstream
[52] Paro Robot Website
[53] Europe Bets on Robots to Help Care for Seniors
[54] Can robots solve Grandma's loneliness?
[55] Persistent: Machine Learning: Data Integration
https://www.persistent.com/machine-learning/
[56] With machine learning and AI in
healthcare, can you speak the language?
[57] Ayasdi
[59] On ML
& AI in Healthcare
[60] Mercy Optimizes and Automates Patient
Care Pathways with Ayasdi
