NiADA (Non-invasive Anemia Detection App)

Anaemia is a leading contributor to global burden of disease that particularly affects young children, menstruating adolescent girls and women, and pregnant and postpartum women. WHO estimates that 269 million or 40% of children of 6–59 months of age, 37% of pregnant women, and 571 million or 30% of women 15–49 years of age worldwide are anemic.

Iron deficiency Anemia (IDA) counts for more than half of all Anemia cases around the world. Anemia is a silent killer as the symptoms of this disease are non-specific including fatigue, dizziness, and lack of concentration.

Failure to detect and control Anemia affects physical and cognitive development and obesity in generations of children, causes repeat emergency visits and premature death, increases risk of premature delivery, low birth weight of the babies and perinatal and maternal mortality, leaves millions of women with poor health and quality of life and causes economic harm. One study estimates the economic cost of IDA is 4.05% of global GDP.

In India, ranking 170 among 180 countries, 187 million women of reproductive age, 7.6 million of pregnant women and 62 million of children below 5 years are anemic. Anemia Mukt Bharat is a govt program which was established in 2018 to actively address Anemia prevalence.  Indian National Family Health Survey, 2021, reports that Anemia prevalence has increased significantly since then.

In United States, a report published in June 2022, US Department of Health and Human services/Centers for Disease Control and Prevention(CDC) shows that prevalence of anemia increased 13% between 2008 and 2018 among participants of The Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) across USA. Anemia, among Black pregnant women, was classified as a moderate public health problem (20.0%–39.9%).

And yet, routine surveillance of Anemia is staggeringly limited. All these initiatives face and report the challenge of measurement and missing data to effectively monitor Anemia presence and distributing nutritional supplements accordingly among a target group.   

Currently, diagnosis of Anemia is mostly done using CBC (Complete Blood Count) tests, which:

• Is Invasive, blood needs to be drawn.  
• Requires a special lab setup and skilled technicians which are not available in suburban or remote areas anywhere in the world.
• Is challenging to conduct frequently on children and infants.
• Is time-consuming and often expensive for lower-income communities and daily-wagers.

In recent years there has been very limited use of Hemoglobinometers to help detect Anemia as a point-of-care solution, but it remains an invasive procedure and most importantly its accuracy is still inconclusive.

We, the members of Future Data team, embarked into the journey of building NiADA (Non-invasive Anemia Detection App) because of recent personal experiences with chronic undiagnosed Anemia among our female friends. As we started exploring the options for combatting Anemia personally, it was clear to us how common and yet under-informed, precarious and yet under-financed the problem is. This problem requires a novel non-invasive solution that is easy-to-use, accessible and scalable for a large target group . 

We present NiADA, a non-invasive, easy-to-use, real-time and scalable solution to detect and monitor Anemia efficiently and accurately. 

The solution is built upon the successful marriage between two main components.  The user facing app, NiADA and the data platform, the backbone, Andromeda.

The user facing solution, NiADA, is a smart phone app, built using react-native.  The app employs a three steps process to detect Anemia.

1. User takes a photo or a selfie of a person’s inner eyelid.
2. The system analyzes the image with Deep Convolutional Neural Network (CNN, a type of Artificial Intelligence) model.
3. The system predicts hemoglobin level within seconds using our pre-built model, indicates the severity level as per WHO 
4. The system stores the record in the history log for future trend analysis based on a time period, location , age groups , sex , pregnancy status 

NiADA can scale infinitely with the number of users and stores historical data.  History is currently used by the user or service providers and will be used for automated recommendation later, to manage nutritional supplement requirement and intake. NiADA can be used by any hospital urgent care, primary care health service providers, non-profit organizations who are part of Anemia eradication program to distribute nutrition supplement, at school health camps and by an individual at their home.

NiADA is supported by a solid data foundation, Andromeda, as the backbone and the core of the solution. This data platform is fed with regular eyelid images and mapping hemoglobin level from CBC test, from our partner hospitals daily. The steps for data collection, validation and model training are as follows.

1. Our Data Collectors (humans) uses the data collection app, Andromeda, to collect images from consenting patients at hospital outdoor lab/indoor facilities.
2. Upon collection, our MD, Pathologist cofounder, Dr. Jhuma Nandi, validates the images and mapping hemoglobin level to ensure data integrity. 
3. Validated and approved new images are ingested into deep learning, AI model and a new model is deployed for NiADA to use, if testing for accuracy goes well.

Deep learning algorithms are already proven to be useful in diagnosis from medical images. These algorithms have a single necessary requirement for high quality and high amount of data, in this case inner eye images. Previous research and effort to produce a scalable application fall short in collecting and using enough data to be a viable solution in the market. 

Our team addressed this challenge since the beginning. 

• As a first step, we onboarded high volume hospitals to create our selective data sourcer group.  
• Currently, we are building a platform with a daily flow of about 200 inner eye lid image data from six hospitals in West Bengal, India 
• Continuously validating our model in place, against the lab blood test result. 

Daily input from a vast population ensures that the model is trained on enough data that is well distributed in terms of patient’s age, gender and hemoglobin level. This data platform is the backbone and the heart of our solution, NiADA. 

Iron Deficiency Anemia affects reproductive age women disproportionately. We plan to reach the affected women population first .

Due to our deeper connection in India and India ranking 170 out of 180 countries with Anemia problem , we are targeting Indian population first . 

Target population 1

We are working with schools in a few Indian cities and nearby areas to use NiADA on girls aged 15 to 19 to regularly monitor Anemia . 

During our primary market research (PMR) with school board members , much interest for buying NiADA was expressed . In last few years the condition has worsen among adolescent girls from lower and middle income families in both rural and urban areas. 

Anemia in the adolescence causes reduced physical and mental capacity and diminished concentration in work and educational performance, and also poses a major threat to future safe motherhood in girls- NIH states.

NiADA would help measure and monitor presence of Anemia onsite , at school health camp with instant result and help distribute proper nutritional supplement as needed , instead of blind distribution.

Target population 2

Rural and suburban reproductive age women 

We have done PMR with the National direct and head of two NGOs who are member of Anemia Mukt Bharat forum and provide training among women to grow iron , folic acid , B12 reach food. They were very interested in NiADA as it is non-invasive , NGO workers will be able to use it onsite and detect presence of Anemia on the spot to motivate the women to plan for better nutrition. 

Currently monitoring Anemia is a hassle as travelling to lab for blood test means losing a day of work and NiADA can save that.

Anemia leaves millions of women with poor health and quality of life and nations to lost economic productivity and development and NiADA can help mitigate that by early detection.

Target population 3

Pregnant women has increased need for iron and folic acid supplement. 

In pregnant women Anemia is associated with increased risk of premature delivery and low birth weight of the babies, perinatal and maternal mortality. 

Without easy surveillance mechanism for Anemia detection and monitoring , govt programs ( like WIC by CDC in USA and in India through Asha program ) complain that it is hard to manage the disease. NiADA can provide an easy and real time surveillance mechanism.

Target population 4

NIADA can be used for Children under 5 are served through primary care centers at rural areas . NiADA being non-invasive will be very useful for testing Anemia in little children.

Failure of control Anemia affects physical and cognitive development and obesity in generations of children.

Target population 5 

Senior citizens and chronic disease patients can use NiADA to test and monitor Anemia easily at home or at senior care centers , which can reduce repeat visits to ER and primary care centers and also untimely death , Anemia being the primary diagnosis.

The journey started with a few personal incidents for me, Mou, one of the cofounders at Future Data.

Two of my closest childhood friends, in India, were diagnosed with severe and life-threatening Anemia. They were living with it for months and attributed their tiredness and low energy to everyday workload. It only took one to fall asleep on the car wheel and a fever that made her completely unable to move and then admitted to hospital special care unit for plasma injection and the other to preoperative testing for a tumor only to postpone the operation and learn Anemia was killing them silently.

These incidents took me by surprise and hopelessness as it hit close to home. More I talked to my other friends; it shows almost all my female friends either are living with Anemia with occasional spike in severity or had Anemia during their pregnancy. In addition to talking to our friends , we started a google survey among our extended friend circles to gather feedback from women of reproductive age in both India and USA.

Anecdotally and statistically, two out of three women are living with severe Anemia and there was nothing available for us to measure and monitor it until we visit the doctor’s office. So, we stall it as long as we can.

I started reading anything I could find on Anemia to understand if something can be done to make Anemia detection more commonplace and easier. There are plenty of resources online and research in medical journals focusing on studies on Anemia prevalence including WHO website. Interestingly, some articles explained that inner eyelid paleness is the most reliable indicator for doctors to suspect Anemia, before conducting a CBC blood test to confirm. This sounded to me like a classic case for using deep learning for reading and analyzing images and predicting hemoglobin level.

Next step was to run a proof of concept with my two cofounders . We bought 218 inner eye lid images that were available at IEEE  to run the POC. The POC revealed that we must build a platform to source eyelid images regularly, with lab tested CBC generated hemoglobin level to train the model. Previous research works were never productized at scale due to this operational difficulty of collecting enough and diversified data for generalization of the model.

With our ties to medical community in India and our ties into tech world as the alumni of prestigious Jadavpur University, we were able to build a software platform and a data collector team within two months to support a daily flow of 200 eye images from six hospitals. We continue to expand our network of data sourcer hospitals to build a diverse source pool and working on partnering up with local hospital chains like Intermountain Health in Utah.

We connected with a few NGOs and schools’ boards in India through our friends who are actively working on Anemia eradication program and eagerly waiting for NIADA to pilot at their organizations.

Our solution has graduated from a concept to prototype in the last two months . The timeline below charts our journey for last 4 and half months.

1. Conception  - Dec 16th, 2022

We started brainstorming on the concept on December 16th, 2022, as I received the news of severe Anemia postponing a required tumor operation for my friend. 

2. Start of POC - January, 2023

We started working on the proof-of-concept in early January 2023 when we got hold of about 218 inner eyelid images available on IEEE data site for the paper1. At this point, we can prove that deep learning algorithm trained on inner eye lid images can detect the blood hemoglobin level with more than 90% accuracy. 

3. Primary Market research -February, 2023

We have done our PMR among the following potential customer categories. 

• School Board members in India
• NGO National Directors in India 
• Hospital administration heads in India and Utah
• Owner of Day care centers in NY
• Doctors at pediatric center in Utah
• Adult women with smart phone globally

At this point we were ready to build our prototype and needed more data to scale and generalize the model. This required putting together a regular operation department and signing & onboarding of hospitals to let us collect data from the hospital lab and outdoor. We currently have six hospitals that have joined our Data Sourcer group.

4. Starting Data Collection ops – Andromeda(Data Collection Platform)  goes live - 3rd week of March, 2023

We launched this Andromeda, the data platform app, in the field in the third week of March 2023 and hired nine data collectors to use the App every day .

The data collection app is used by the data collectors and three administrators to create a regulars supply of valid eyelid images from the hospital lab . 

We now run a streamlined operation for 

• collecting eye image data first day 
• collecting and uploading the CBC result report next day
• mapping the matching tested hemoglobin level the next day for each patient 
• validating the data quality before ingesting into model required the data platform

Andromeda is designed and built to work smoothly, with and without internet and with an option to upload images when internet is available. 

5. Building Model V1 - End of April, 2023

As the new data is now coming into our data platform, the model is being tested on the new data that has not yet received lab result. The model is used in the backend API to predict the hemoglobin value and stored it in the data platform . The next day when the lab results are in , model accuracy is calculated and the model is updated as needed. 

An automated data pipeline runs from data collection to model building and publishing that model to be used by NiADA . Our v1 model already shows promising result for hemoglobin level prediction.

In the month of May 2023, we are all set to test the model on 3000 to 4000 patients and continuously improve it.

6. NiADA – V1 launched - May 7th, 2023

We start testing on on-demand eye lid images by friend and family using Model V1. 

The first release of NiADA uses the v1 model to predict the hemoglobin value in real time as the user takes a picture of their patient and we use this feedback to further improve the model.

NiADA v1 is also being tested for the optimal design of the camera inside the app to minimize the noise introduced by common users.

The product , NiADA is in prototype stage and yet to serve any real user.

While our data platform solution is in production for last two months and collecting data using the app, the prediction model is in testing as we collect data from patients at the hospitals.

We plan to go through NiADA testing phase for another month before it pilots in a few schools and private primary care centers.

In the month of May, 2023 , 

• NiADA will be tested on new 3000 to 4000 patients ( projected based on daily data collection volume ) 

We are hoping to get access to MIT's network to

1. be part of a diverse and vibrant network for any help building effective GTM strategy for global market.
2. be part of an innovative peer group where we can participate in meaningful exchange for improving the solution .
3. be able to attract exceptional and passionate talent to work with us.
4. get us some regulatory help for the process of registering our App as SaMD(Software As Medical Device) in FDA classified list, this classification is quite new, so that it could be comparatively easier to introduce in North American market.
5. connect ourselves to potential funding opportunities which are dedicated to solve women's health issue problem as Anemia is one.
6. get funding for the pilot phase to get us to revenue generating stage

Our solution approaches the Anemia surveillance mechanism in a novel way. Our solution encompasses all of the following properties  

• easy to use - a smart phone app that takes a photo of inner eyelid to show the result.
• non-invasive , does not require blood test for Anemia screening.
• accessible - usable in both internet and no-internet area using mobile friendly model.
• convenient as no travel to lab is required just for an Anemia screening.
• inbuilt observability - shows history and trend at patient level and also at location or program level and at the demographic group(age, sex , pregnant, menopausal ) level.
• scalable - grows with use as backend is based on a separate data platform
• a technology architecture backed by and differentiated with a data platform that can incorporate incoming data from various global locations from diverse population and support continuous update the AI model based on feedback . 

The above attributes makes our solution first of its kind to be available at scale and most desirable by the doctors, insurance , schools , NGOs and agencies implementing govt programs to use regularly. Given that Iron Deficiency Anemia costs countries 4.05 % of their GDP, our solution would introduce a new way for preventive and early detection of this commonplace and yet silent killer disease.

NiADA being a smart phone app backed by a scalable data platform that track history , it is easier for institutional users ( govt and non-gov officials and/or hospital administration) to observe the trend with frequent tests and identify special interest groups for focused and intensified measures to manage and reduce prevalence of Anemia and distribution of nutritional supplement equitably.

Current year plan : 

Target Market 

This year , we would like to empower public health administrators , NGOs and School administrators to test , treat and prevent anemia using NiADA and thus participating in Indian government programs for "Anemia Free India" to achieve the goal to reduce prevalence of anemia by 3 percentage points per year among children, adolescents and women in the reproductive age group (15–49 years) as stated here.

Intended impact of using NiADA

• Reliable and accurate anemia screening on the spot 
• Monitoring progress of Anemia prevalence for the wide area population and also down to the individual beneficiary over a time period
• Effective and equitable distribution of nutritional supplement at village/block level using the trend in history data 

Next year plan : 

Target Market 

Our next year's plan is to classify NiADA as SaMD - Software as Medical Device under FDA and work with CDC for their WIC (women and Infant Children) program to help ease the Anemia surveillance problem as stated in their report. In last decade , Anemia has progressed among WIC participant groups by 13% across USA. Anemia is particularly prevalent among black women in their third semester. 

Intended impact of using NiADA

• Reduced repeated visits to primary care centers
• Reduced number of emergency room visits 
• flexibility of test for anemia presence at home 

Next four years plan : 

We would like to expand NiADA's reach to  Africa , Europe and APAC countries next.

Intended impact of using NiADA

Similar to India target population in Africa . And for Europe and APAC , the impact is similar to north America.

Anemia causes both health and economic harm and affects women disproportionately. As part of UN SDG 2 target , it aims to halving the prevalence of Anemia among reproductive age women by 2030. The indicator 2.2.3 is designed for UN SDG 2 to measure the change in prevalence of anaemia in women aged 15 to 49 years, by pregnancy status (percentage).

This aligns with our solution as women of reproductive age as our target users through schools , NGOs and govt programs for surveillance of anemia. 

NiADA helps detect and monitor Anemia just using a smart phone on the spot. While doing the screening it also captures data for age, sex and pregnancy status and stores the history data . 

Following reports/dashboards for public health administrators can help in anemia management for the population in an area over time.

• NiADA can be used more frequently than the classic invasive lab test to detect and tracks pregnancy status vs presence of anemia for the target population more closely to show the trend over time and administering appropriate, informed measures . 
• In addition to pregnancy status, NiADA can also create reports based on age, sex, location/area, and menopausal status to measure the progress in a target population over a time period.

A change is when a product/process/solution/system is built to enhance accessibility and equitability for an existing problem-context at scale. 

To make a change happen or to have an impact on the problem, it is important to address the root cause of the issue. Here, the problem in hand is the worldwide prevalence of Anemia, however, one of the root causes of that problem is insufficient capability of real-time, on-the-spot detection that enables necessary intervention. Anemia exists with non-specific symptoms for women to ignore them long enough. Late detection often leads to loss in productivity and even life-threatening conditions. The absence of an effective detection and tracking mechanism impedes the journey of reducing anemia prevalence. Therefore, to provide a real-time, non-invasive, easy-to-use and accessible solution to screen anemia is one of the key prerequisite for having an impact on the problem.

NiADA addresses this key problem by providing an easily accessible ,smart phone app to provide a non-invasive and point-of-care solution for detecting anemia. It is going to empower the the beneficiaries(users) and the administrators alike. To be informed about one's state of health , not only detection of anemia but also effective follow-up with history tracking , as frequently as they want, without the hassle of going to lab and waiting for the result for hours, are the game changer capabilities of NiADA. NiADA ,being a non-invasive solution is particularly very helpful for one of the target population groups, children under 5 year old. Use of NiADA will significantly reduce the issues originating from the late detection for all target groups, which will expedite the process of management and cure of Anemia. NiADA is designed to help public health administration with an effective and accurate way for conducting Anemia surveillance among a wider population with relevant reports and dashboards so that Anemia prevalence can be measured regularly and easily. 

Here in Healthworld, an Economic Times publication, Dr. P Siva Kumar, a leading Indian Doctor, commented, “With the ease of the use and operation of PoC testing of hemoglobin, the volume of hemoglobin tests is therefore expected to grow and will be a key contributor to reducing the global burden of anemia worldwide”. Along with other recent reports , this advocates the importance of a "Point of Care (PoC) testing" as a key tool in winning the war against anaemia. The point-of-Care solution, mentioned in the above article, is being used in the ‘Anemia Free India’ program. The solution is still an invasive one and the feedback from ground says that the accuracy is not satisfactory. We can confidently say that the availability of an accurate, non-invasive solution like NiADA will surely accelerate the journey towards anemia free world.

Our solution uses react-native to build the user interface , the mobile app. 

We are using multiple AI models based on Convolutional Neural Network ( Deep Learning )  for noise-reduction in user taken photos and building the Anemia detection algorithm. Model architectures are customized for our training images.

Our user interface app is backed by a scalable data platform that stores data anonymously and encrypt them at rest and hosted in cloud. This architecture for a separate data platform makes our solution easily extendable for a future problem solving .

The data platform enables easy and scalable historical information storage inside a search engine which can produce reports and dashboards on demand. Use of search engine enables easy recommendation system for possible treatment of anemia and analysis of correlation between anemia and other medical conditions by integrating digital health records. 

Our leadership team(cofounders) includes two women of color and a person of color from India. Our data collectors include two women and we strive to hire more women . One of our part-time developer is a women of color , originally from Iran. Throughout our career in technology and medicine we have watched DEI program being deployed sometimes successfully and sometimes as a token position. We have experienced exclusion firsthand to understand the need for inclusion. However , this is a continuous process and we commit to keep this not only a priority but a must have criteria for our growing workforce to stay innovative and exciting place to work .

During our primary market research, we have identified multiple channels and hybrid model to take NiADA to the market.

Targeting women of reproductive age, NiADA aligns well with goals and objectives of multiple NGOs, middle and high schools, OBGYN centers, Rural health centers in India.

Key customer-1:

School boards for middle and high schools

Beneficiary 

All adolescent girls (10 -19)

Objective 

School wants to ensure the best possible physical and mental capacity for better work and educational performance.

Customer challenge 

Anemia causes diminished cognitive ability. School has a need for replacing existing invasive and time-taking anemia testing process with NiADA like app.

Value-prop 

NiADA can empower the school physical ed. department to conduct weekly tests and monitor the presence of anemia easily, at the school site just using a mobile device. They love that the test is non-invasive.

Revenue opportunity 

At the time of PMR, school board members stated that schools have operating budget for health camps, and they are eager to use NiADA when ready.

Key customer-2:

NGOs and private agencies working for Govt and who are part of Anemia free India forum to reduce prevalence of anemia among the target group by three percentage point.

Objective 

• providing effective training for growing and buying food for the household that can reduce anemia prevalence

Customer challenge 

As an accurate point-of-care solution for anemia detection not easily available, the training becomes ineffective. Without any visible proof of anemia presence, many women do not feel the urge to follow a nutritional routine. Hemoglobino-meter was introduced to run onsite tests, but it mostly inaccurate and an invasive procedure.

Key Customer-3:

• Indian govt agencies like ASHA,POSHAN, RBSK , AWK working on anemia free India get grants from govt and in USA as part of CDC – WIC program  

Objective

• frequent and accurate surveillance of anemia for better management and nutritional supplement distribution

Beneficiary for Customer-2 and 3

 Women of reproductive age (15 – 49) and their household in suburban and rural areas

Value-prop for Customer-2 and 3

• NiADA can empower the NGO/govt workers to perform on-the-spot testing using their mobile phone. The NGO administration believes, NiADA’s non-invasive test and real-time result with history-tracking can and will motivate many more women to follow and stick to a better nutritional routine to reduce anemia among this population.
• By tracking the history and trend for a local area population, NiADA will facilitate nutritional supplement distribution.

Revenue opportunity for Customer-2 and 3

• NGOs/Govt have budget for Anemia tests, and they are willing to buy in bulk.
• A subscription model for the B2G strategy will be the preferred monetizing policy for govt orgs.

Key customer-4:

• Middle class and upper middle-class women in India
• Women, and senior citizens in North America

Value-prop -These users are familiar with mobile app use and like the flexibility of anemia testing at home without blood draw for themselves and their children.

Revenue opportunity - A subscription model for the B2C strategy will be the preferred monetizing policy for this group.

Short term plan - 

Bootstrapping - We are funding ourselves through the prototype phase of NiADA.

Friend's and family fund - We have a few friends to collect some fund help us through the  phase piloting 

Applying for funding - We are applying for funding at NSF , NIH and here at MIT Solve to get us through the piloting phase. 

Raising investment capital - We have started exploring a few investors who are passion driven and invest in startups that work on women related problems. 

Long term plan -

We plan to be self-sustainable as soon as possible. 

We start hiring sales personnel by the end of q3 this year and start selling to Schools and NGOs first as they can use their operating budget for piloting NiADA . 

Our calculation shows , if we can onboard 100 schools by Q1 year 2 and sale to one NGO which works in 500 units across 10 states in India, less than .03 % of rural and suburban reproductive age women population through one Indian govt agency  and 0.01% of Indian's urban upper middleclass reproductive age women we can break-even by the Q1 year2. Please see the j-curve as of now. 

We have not started approaching outside investors for funding yet. 

We just started the venture four months ago we plan to invest 100k ourselves and building the product ourselves until we successfully graduate from prototype stage to piloting stage. We wanted to make sure the prototype works well before asking for fund from outside. 

During this time, we are applying for research funding grants from MIT Solve,  NSF and NIH as our solution qualify for those and Utah governor office (SBIR) is helping us through the process of grant applications. 

We are also short-listing investors who want to invest in women (public) health issues and asking for investment from friends who are passionate about the cause and will walk the mile with us.

The above j-curve forecasting shows us that with the above business model and GTM strategy in place we will reach a break-even point by the end of qtr. 4 this year.