Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
The Road Ahead - Reinventing the Approach
to Diabetic Retinopathy in the Era of AI
Eye – Primary Human Cognition
Visual perception of our surroundings gives us the sense of reality around us. Interpretation
of light from surrounding medium gives us visual perception. Eyes are our source to process
light from our surroundings, to create visual perception. The Eye is a major sensory organ
and is an important medium in the human cognition system. Up to 80% of human cognition
data is consumed through visual medium. Vision, being our dominant sense, combined with
other sensory organs provides us with a sense of being and reality.
Diseases of the eye not only affect the vision but also
coordination of our cognitive abilities. Blindness
reduces people’s ability to perform daily tasks and
move about unaided, affecting their quality of life and
ability to interact with the surrounding world. It is the
most severe form of visual impairment. Most of the
diseases and conditions causing blindness can be
prevented or treated with known and cost-effective
interventions, if detected early.
Blindness – worrying statistics
Blindness and vision impairment are a major health problem and 80% of all blindness and
vision impairment is preventable or treatable. Some related facts are illustrated in table
below:
Blindness Facts
232.5 Million4 people in the world are visually impaired
• People aged 50+ year make up 81% of global population who are blind or have
moderate to severe vision impairment
• Among children an estimated 19 million are vision impaired.
• Worldwide, 425 million adults1 (1 in 11) have diabetes; half the cases are
undiagnosed. This number is expected to reach 629 million by 2045.
• Of Indian citizens, 72.9 million2, or nearly one in thirteen, have diabetes and by
2025 the it would be astonishing 134.3 million.
• Of all visual impairment 4.19% is caused by DR
Approximately 80%3 of all vision impairment globally is considered
avoidable.
Leading causes of blindness
Cataract, refractive error and diabetic retinopathy are the most common causes of blindness.
They all fall under the category of vision impairment that are preventable and treatable.
Intervention by ophthalmology methods to improve eye health have been found to be the
most cost-effective in tackling them.
CataractRefractive Error
DiabeticRetinopathy
Blindness - leading causes
Ophthalmology & Inverted Healthcare Pyramid
Ophthalmology has the potential to provide solutions for the cause of blindness. But this
area suffers from a few shortcomings. In India there are only about 35K+ ophthalmologists
for its 1 billion-plus population, resulting in a ratio of 1 ophthalmologist for every 70,000
people, while an adequate ratio is ideally 1 ophthalmologist for every 15,800 people.
On top of it Urban: Rural divide has increased this ratio further, hence it has led to a steady
increase in the numbers of patients with blindness in India i.e 1: 99000 ophthalmologist
This is further accentuated by the inverted healthcare pyramid.
The inverted healthcare pyramid, three layered government health centre model is tuned for
graded care and brings health and medical care closer to patient’s home but hinders timely
detection in case of eye diseases affecting blindness.
Among the causes that lead to blindness Diabetic Retinopathy is the leading cause for the
most productive age, 21 to 60, in the working population of India.
The shortcomings of the inverted healthcare pyramid are illustrated in the diagram below
with diabetic retinopathy representing the predicted growth in cases in India on the right side
and operational factors affecting the health care model on the left.
Visual interaction with digital devices, in the growing digital economy, will play the dominant
role amongst all the cognitive abilities. And to have a large population in the working age
affected by this disease will hamper their ability to benefit from digital interactions in the
newer economy. The health care model specifically and industry in general is not fully
geared to tackle the growth in diabetic retinopathy cases.
Operationally more than 60% patients with early symptoms would need only initial screening.
These large number of population can be screened through AI – Deep Learning Technology
which is revolution in Healthcare industry. The inverted health care model even though
beneficial in other disease handling, lacks resources trained to diagnose diabetic
retinopathy, at the health centers.
In cases where trained resources are available, they are spread thin geographically and
patients in turn suffer from delay in follow-ups. Eventually patients lose traceable and graded
treatment for diabetic retinopathy.
The problem is compounded by the lack of coordination between diabetic care and eye care.
Both are also yet to exploit advances made in technology driven diagnosis.
India to have 134.3 million people with Diabetes by 2025
Diabetic Retinopathy is leading cause of blindness in working age (21 - 60)
Expected to affect 27 miiion people by 2025
Early detection & timely treatment can prevent vision loss due to Diabetic
Retinopathy by 95%
> 60% patients at tertiary centers need only intial screening
Lack of trained resources
Delay in follow-ups across health centres
Lack of coordination between eye care and diabetic care
Inverted Healthcare Pyramid
Op
era
tion
al Is
su
es
in H
an
dlin
g D
R
Gro
wth
in D
R c
as
es
Sub-Centers
Community Health Centers
Primary Health Centers
District Hospitals
Diabetic Retinopathy (DR)
What causes diabetic retinopathy
Chronically high blood sugar from diabetes is associated with damage to the blood vessels
in the retina, leading to diabetic retinopathy. The retina detects light and converts it to signals
and send through the optic nerve to the brain. Diabetic retinopathy can cause blood vessels
in the retina to leak fluid or hemorrhage (bleed) leading to decrease in vision. In its most
advanced stage, new abnormal blood vessels proliferate (increase in number) along with
fibrous proliferations on the surface of the retina, which can lead to tractional retinal
detachment with or without vitreous hemorrhage.
Diabetic retinopathy is the most common cause of irreversible blindness in working-age
people. Elevated sugar levels from diabetes can damage the small blood vessels that
nourish the retina leading to leakage of fluid, exudates or blood from them and at times may
even block them completely. Diabetic retinopathy occurs in at least 25 % of the people who
develop diabetes(5).
The tragedy is even greater when one considers that this
blindness is largely avoidable, with regular screening.
Diabetic Retinopathy (DR) Stages
Diabetic retinopathy may progress through four stages:
1. No DR /Normal – A healthy retina
2. Mild Non-proliferative DR (NPDR) - The earliest stage of the disease consists of
microaneurysms caused by localized swelling (aneurysmal outpouching of the
capillaries) in the blood vessels of the retina.
3. Moderate NPDR - Blood vessel wall may be damaged as the disease progresses
and causes it to leak its contents (leads to hemorrhage, exudation, edema). At times
they may also lose their ability to transport blood (leading to axonal infarcts – cotton
wool spots). Both conditions cause characteristic changes to the appearance of the
retina and may contribute to DME – Diabetic Macular Edema.
4. Severe NPDR – The process of the disease seen in moderate stage is exaggerated
leading to more hemorrhages in the all the quadrants of the retina or venous beading
or the presence of communication between arterioles and venules are noted in this
stage. Cells in areas of retinal ischaemia (inadequate blood supply) secrete growth
factors that signal the retina to grow new blood vessels.
5. Proliferative DR (PDR) - At this advanced stage, Growth factors secreted by the
retina trigger the proliferation of new blood vessels, which grow along the inside
surface of the retina and into the vitreous gel that fills the eye. The new blood
vessels that start developing are likely to leak blood into the vitreous cavity. If
associated with development of fibrous proliferations may cause traction on the retina
causing permanent vision loss due to retinal detachment - the pulling away of the
retina from underlying tissue, like wallpaper peeling away from a wall. Retinal
detachment can lead to permanent vision loss and even after surgery vision recovery
is partial.
Diagnosis & Examination
Doctors have long used an ophthalmoscope to
look at the back of your eye for screening for
Diabetic Retinopathy. However, even if an
adequate number of ophthalmologists are
available, using ophthalmologists to screen
every person usually is not feasible and is likely
to be an inefficient use of resources. A
screening examination theoretically could
include a complete ophthalmic examination with
best-corrected visual acuity after refraction,
pupil dilation, and latest retinal imaging, such
as with wide-field retinal photography and optical coherence tomography (OCT).
Retinal imaging takes a digital picture of the
back of your eye. It shows the retina, the optic
disk, and blood vessels. This helps
ophthalmologists find certain diseases and
check on the health of your eyes. Retinal
imaging allows doctors to get a much wider
digital view of the retina. It doesn’t replace a
regular eye exam but adds another layer of
precision to it.
The retinal examination may be accomplished in the
following ways: (1) direct or indirect ophthalmoscopy
(2) retinal (fundus) photography. This could be done
with or without accompanying OCT. Low-cost
cameras are now widely available. The retinal
examination could also include telemedicine
approaches. Using retinal imaging, several
abnormalities can be identified – microaneurysms,
hemorrhages, macular oedema, exudates, and cotton-wool spots. The earliest
manifestations of diabetic retinopathy are focal, saccular dilations of the retinal capillary bed,
classically termed microaneurysms. These have the appearance of small red dots.
Artificial Intelligence (AI) – The augmentation between DR growth and healthcare
pyramid
AI-assisted medical screening and diagnosis based on retinal images can be used for
auxiliary screening in detection of DR and diagnostic enhancement technique for severe
cases.
The automatic and augmented identification of DR allows the health care pyramid discussed
earlier to be scaled at all levels.
Existing health care workforce can be trained to take retinal images with minimal training at
health care centers. Initial screening can be performed at community or primary health care
centers to segregate cases of DR vs Non-DR. DR cases can be referred to the next levels
with the AI augmented retinal images helping the ophthalmologists to prepare diagnosis
plans at scale.
The augmentation from AI also helps to bridge the coordination gap that exists across
diabetic and eye care verticals in healthcare. AI driven technology devices deployed across
diabetic and eye care centers allow them to perform DR screening at either of the vertical
centers and cross refer the patients with relevant medical history with traceability being
passed on.
In order to exploit AI technology at scale and be able to be used by last mile health care
worker there exists a need to develop an AI platform that can use industry wide standard
retinal images and augment the health care pyramid with traceable and explainable
diagnose for eye care by simplifying the use of complex systems needed to use AI.
AISeon Healthcare Technologies is focused on the developments of the solution for
ophthalmology platform named Dhi Ajna [i.e Intelligent Eye], with vision to become part of
leading force to fight against preventable blindness across the world.
It aims to reach out to indigent individuals suffering from curable eye
ailments and uses Assisted Augmented Intelligence to enable
Ophthalmologist to take informed decision with power of cutting-edge
technologies.
Its solution is developed on the bedrock of technologies using Artificial Intelligence, Machine
Learning and Deep Learning and is geared for “AI for Good” initiative. It has adopted an
AI
Deep Learnig
Machine Learning
Ethical approach to using Artificial Intelligence. Its solutions are geared towards achieving
automation of retinal imaging and analysis and delivering this at scale.
Dhi Ajna Platform
Dhi Ajna Platform is an automated AI [Artificial / Assisted Intelligent platform] which will help
ophthalmologist to diagnose and treat preventable blindness. This platform can be used to
sense not only retinal problems but also can be used to detect Diabetic Retinopathy,
Cataract, Glaucoma, Macular Degeneration as we scale our platform.
The platform uses an innovative approach of cross feeding dual cycle pattern to automate
the diagnosis of DR and cross feeds this diagnosis with industry standards and policy for
adaptive decision making based on the diagnosis. This dual cycle is a unique approach to
allow the use of AI technologies in a responsible and ethical manner. The diagnosis and
decisions provided by this AI driven approach allow for the process of diagnosis to be
transparent and empowers the human by keeping them in the loop adhering to policies and
regulations.
The dual AI cycles in the platform are the Autonomic Augmentation cycle and Adaptive
Decision cycle.
The Autonomic Augmentation cycle uses retinal images to produce a knowledge graph. This
knowledge graphs represents salient features in retinal images and establishes relations of
features in respect to retinal images.
Knowledge
Graph
AI Diagnose
Graph
Ground Truth
Retina ImagesDe-biasing
Feature Collation
Retinal Image
Capture
Image Decision
Map
Industry Standard
Mapping
Augumented
Diagnose Report
AI Diagnose
Module
Autonomic Augmentation Cycle Adaptive Decision Cycle
This knowledge graph helps the system designers to have “Human in The Loop”
functionality. Due to its “Human in The Loop” functionality, the knowledge graph can be
configured and tuned for different demographic profiles. It can also help the ophthalmologist
to get a better understanding on how the Dhi Ajna platform uses retinal data to arrive at a
diagnosis
The Autonomic Augmentation cycle has an in-built mechanism to de-bias the data it uses to
build its knowledge graph. This cycle builds a weighted graph from knowledge graph so that
it can be used in the next cycle.
The Adaptive Decision cycle has the ability to embed the ever-evolving AI Diagnosis Graph
into its AI Diagnosis module and apply that to retinal image captured through portable or
mobile devices.
This cycle produces decision map based on captured retinal image input and this is mapped
to the current industry standard, policies and regulations adopted by practitioners to produce
an augmented diagnose report.
The augmented diagnosis report helps the ophthalmologist to make informed and guided
decisions. The report helps to establish a digital traceability of decision cycle and quickens
the referrals in severe cases.
Dhi Ajna can parallelize processing pipelines of the dual cycle and thus the platform is
geared to auto scale for mass screening. The dual cycles act as feeders to each other to
provides continuous improvisations mechanism of knowledge base as well as diagnosis
module.
Dhi Ajna – Backbone to Octave methodology
AISeon has built proven OCTAVE methodology for seamless end to end integration from the
stage of capturing the data and retina images up to integration of Dhi Ajna produced and
validated reports being pushed to an EMR system.
Dhi Ajna platform will form the backbone of the workflow for this methodology to provide AI
driven diagnose, analysis, integrations and scale for comprehensive eye screening and care.
Features and Benefits
What Next – Way Ahead
Approximately 80% of vision impairment globally is considered avoidable including DR.
Emerging healthcare technologies focus to minimize unnecessary visit to medical
practitioners, minimizing the cost of treatment making the solution portable.
This is step in right direction to benefit mankind and enables us to participate in government
initiatives like Ayushman Bharat, Camp outreach program through government, NGOs, clinic
and hospitals. The solution is poised to help mankind by augmenting doctor’s decision
making and not replacing them…
We aim to Ignite the Vision for Better Tomorrow!!!!
Co
mp
reh
en
sive
E2
E So
luti
on • Camp Management
• Retina Grader
• DR Prediction & Diagnostic Report
• Knowledge Management Fe
atu
res • Human @ the Center
• Trust and Transperacny
• Enhanced Judgement –Augment Medical Professional e.g. Diabetologist , Health Worker
• Scalable & Interfaceable–Cloud and API Economy
• Feedback Loop
• Solution is extensible (e.g., for detection of glaucoma, retinopathy of prematurity, etc.)
Be
nef
its • Early Detection of DR
• Enables early treatment of preventable blindness
• Economy of scale
• Camp Outreach - City , Village , Remote Locations
• Scalable – Support multiple camp, clinic ,hospital (Multitenant),Quick Response Time
References
1. N.H.Cho, J.E. Shaw, S. Karuranga et al. IDF Diabetes Atlas: Global estimates for the
prevalence of diabetes for 2017 and projections for 2045. Diabetes Research and Clinical
Practice. 2018 Apr; 138: 271-281.
2. Report of International Diabetes Federation.
3. Bourne RRA, Flaxman SR, Braithwaite T, Cicinelli MV, Das A, Jonas JB et al. Vision Loss
Expert Group. Magnitude, temporal trends and projections of the global prevalence of
blindness and distance and near vision impairment: a systemic review and meta-analysis.
Lancet Global Health. 2017 Sep; 5(9): e888-97.
4. World Health Organization, the Global Burden of Disease Study 2010,2012..
5. Salil S. Gadkari, Quresh B Maskati, Barun Kumar Nayak, Prevalence of diabetic
retinopathy in India: The All India Ophthalmological Society Diabetic Retinopathy Eye
Screening Study 2014. Indian Journal of Ophthalmology. 2016; 64(1):38-44.
ABOUT THE AUTHORS
Bhavin Mehta is Managing Director @ Accenture Technology Solutions in India. Bhavin
in his role as Global Data Solution Architect Lead and AAPAC Lead for Accenture’s IBM
Practice drives the thought leadership on driving platform adoption by providing all
required supports to client teams across the globe in selling and delivering solution based
on Data In the New and IBM Technologies. He is certified Master technology Architect
(MTA), through his engagement with clients and being a part of their journey to stay
relevant in extremely competitive world. He is Digital and AI Transformation Expert along
with his key focus on Hybrid Cloud Architecture as Evangelist. Unique blend of
Executive and Technology Skills to Drive Transformation & Digital Business
Solutions at scale. Experienced Enterprise Architect with overall 20 years of IT
experience includes 18 years with Accenture.
His passion to drive the innovation and work with Gen Y, he is playing key role as
Technology Mentor for AISeon Healthcare Technologies, to Reinvent the
differentiated approach to cater to Diabetic Retinopathy in the Era of AI.
He can be reached at [email protected] |Linkedin:
https://in.linkedin.com/in/bhavinpmeta
Vishal Chahal is the Chief Architect for Data and AI technologies at Global System Integrators lab @ IBM. He is also the Technical Lead for IBM Machine Learning Hub at IBM Software Labs @ Bangaluru. Vishal is a Thought Leader on Cognitive products & platforms and in above roles drives the adoption of IBM Data and AI technologies in partner ecosystem by advising, mentoring and supporting partners across the globe. He is Level 3 Certified IT specialist from The Open Group. He specializes in Watson Cognitive Products, Advanced Analytics, Advanced Visualization, DataWarehous and Data Integration technologies. Through his client engagements, he has architected Cognitive and Analytics solutions for multiple customers across Telco, Banking, Insurance, Aviation and Healthcare Industries. He has a rich background of product architecture and development experience on portfolio across Watson, SPSS, Cognos, DB2 and Websphere. Vishal is passionate about driving innovation through developer advocacy and adoption of AI solutions. He constantly contributes to the developer community through his Tech Talks on AI & Robotics and machine learning code published on Github. He is playing a key role as AI mentor to Aiseon Healthcare Technologies on Watson platform. He can be reached at [email protected] | LinkedIn: https://in.linkedin.com/in/vishalchahal
Dr Ashish Ahuja is Ophthalmology, Vitreo-Retina Specialist, Innovator. He completed his
MBBS, DNB, FICO, FAICO(Retina), F.A.I.CO.(UVEA) F.V.R.S.(Aravind eye hospital, Madurai). He has more than 5 years of experience and is currently working as a Consultant at Sadhu Kamal Eye Hospital, Mumbai
He is a passionate innovator with several publications on low cost devices in Eye care with a focus on technology integration with ophthalmology and is an Advisor to AISeon Healthcare technologies. He has also won several Awards for his innovative work at
national and international conferences.
He has also filed a patent for post vitrectomy recovery system to help patient’s recovery after retinal surgeries. His area of interest is to develop low cost devices to impact rural healthcare delivery and to improve access of healthcare to all. He can be reached at [email protected]
Dr. Kushal Agrawal is a Vitreo- Retina Surgeon, he is working as a domain expert in
AISeon Healthcare Technologies and he is consultant in Jupiter Hospital, Thane.
He has Completed his MBBS from 2006 – 2012 from V.S Hospital, Ahmedabad. Post that he has completed his M.S. in Ophthalmology from 2012 – 2015 from Civil Hospital, Ahmedabad.
He has done his fellowship in Retina from Prestigious LV Prasad Eye Institute,
Hyderabad. He has published scientific papers in American Journal of Ophthalmology,
British journal of Ophthalmology and various Indian Journals.
His team has been awarded best Surgical Video award by American Academy of
Ophthalmology in 2017. He has presented in various conferences in India. He is keen to
do research and involved in academic activities.
He is well versed with all kind of Retinal Surgeries and having special interest in Diabetic
Retinopathy.
He can be reached at [email protected]