MERON
MERON is a smartphone app that leverages deep learning technology to map facial features from photos and quickly detect malnutrition in children under five.
Type of Digital Public Good
- Open content
- Open data
- ✅ Open software
- Open standard
- Open AI model
1. Is it relevant to one of the Sustainable Development Goals?
- 2. Zero Hunger
Link to Evidence: https://kimetrica.com/our-projects/?country=&service=&search=Methods+for+Extremely+Rapid+Observation+of+Nutritional+Status#start
- 3. Good Health and Well-being
Evidence: In 2020, 45 million children under five were wasted (https://data.unicef.org/topic/nutrition/malnutrition/). These children require urgent feeding, treatment and care to survive. Early detection of nutritional status is essential, but current methods are either invasive, time-intensive, error-prone, and too dangerous to use during the COVID-19 pandemic. As such, there is a need for a safer, faster and more affordable method for assessing nutritional status in children especially when clinical capacity is stretched. MERON helps to quickly identify, locate and target the children most in need of nutritional assistance. With MERON’s more timely, cheaper, yet accurate, information, we can advance the prevention, diagnosis, and treatment of childhood undernutrition. For severe acute malnutrition (SAM) children alone, this means reducing deaths by 55%, thereby preventing over 150,000 deaths across low- and middle-income countries.
- 9. Industry, Innovation and Infrastructure
Evidence: In 2020, 45 million children under five were wasted (https://data.unicef.org/topic/nutrition/malnutrition/). These children require urgent feeding, treatment and care to survive. Early detection of nutritional status is essential, but current methods are either invasive, time-intensive, error-prone, and too dangerous to use during the COVID-19 pandemic. As such, there is a need for a safer, faster and more affordable method for assessing nutritional status in children especially when clinical capacity is stretched. MERON helps to quickly identify, locate and target the children most in need of nutritional assistance. With MERON’s more timely, cheaper, yet accurate, information, we can advance the prevention, diagnosis, and treatment of childhood undernutrition. For severe acute malnutrition (SAM) children alone, this means reducing deaths by 55%, thereby preventing over 150,000 deaths across low- and middle-income countries.
- 17. Partnerships to achieve the Goal
Evidence: In 2020, 45 million children under five were wasted (https://data.unicef.org/topic/nutrition/malnutrition/). These children require urgent feeding, treatment and care to survive. Early detection of nutritional status is essential, but current methods are either invasive, time-intensive, error-prone, and too dangerous to use during the COVID-19 pandemic. As such, there is a need for a safer, faster and more affordable method for assessing nutritional status in children especially when clinical capacity is stretched. MERON helps to quickly identify, locate and target the children most in need of nutritional assistance. With MERON’s more timely, cheaper, yet accurate, information, we can advance the prevention, diagnosis, and treatment of childhood undernutrition. For severe acute malnutrition (SAM) children alone, this means reducing deaths by 55%, thereby preventing over 150,000 deaths across low- and middle-income countries.
2. Does it use an appropriate open license?
Yes, this project is licensed under the following license(s):
3. Is ownership clearly defined?
Is the ownership of the project and everything that the project produces clearly defined and documented?
Yes
If yes - please link to the relevant copyright, trademarks, or ownership documentation for the project.
4. Does the license of libraries/dependencies undermine the openess of the project?
Does this open project have mandatory dependencies (i.e. libraries, hardware) that create more restrictions than the original license?
No
If yes - are the open source components able to demonstrate independence from the closed component(s) and/or are there functional, open alternatives?
Not Applicable
If yes - please describe how the open source components are independent and/or list the open alternatives for the closed component:
Not Applicable
5. Is there documentation?
Does some documentation exist of the source code, use cases, and/or functional requirements. For software projects, this should be present as technical documentation that would allow a technical person unfamiliar with the project to launch and run the software. For datasets and data projects, this should be present as documentation that describes all the fields in the set, and provides context on how the data was collected and how it should be interpreted. For content collections, this should indicate any relevant compatible apps, software, hardware required to access the content and any instructions about how to use it.
Yes
If yes - please link to the relevant documentation:
6. Is non PII data and/or content accessible?
Does this project collect or use non-personally identifiable information (non-PII) data and/or content?
Yes
If yes - is there a mechanism for extracting or importing non-personally identifiable information (non-PII) from the system in a non-proprietary format?
Yes
If yes - describe the mechanism for extracting or importing non-personally identifiable information from the system in a non-proprietary format:
REST API
7. Does the project adhere to privacy and other applicable international and domestic laws?
Has this project taken steps to ensure adherence with relevant privacy, domestic, and international laws? For example, the General Data Protection Regulation (GDPR) in the European Union or the Supplementary Act A/SA.1/01/10 on Personal Data Protection for the Economic Community of West African States (ECOWAS) (yes/no)
Yes
If yes, please list some of relevant laws that the project complies with:
- General Data Protection Regulation (GDPR)
If yes, please describe the steps this project has taken to ensure adherence (include links to terms of service, privacy policy, or other relevant documentation):
- Data is controlled in isolated storage and used only for offline training with separation between the data scientist and the data. Individual data points containing PII are only used for random QA sampling and never leaves controlled storage.
8. Does the project adhere to standards and best practices?
Does this project support standards? (i.e. Web Content Accessibility Guidelines (WCAG) 2.1 or other standards such as those listed on W3C)
Yes
Which standards does this project support (please list)
- REST API
- JSON
- JPEG
Can you point to evidence of your support? (i.e. please link to your validator, open test suite, etc.)
Was this project built and developed according to or in adherence with any design, technical and/or sector best practices or principles? i.e. the Principles for Digital Development?
Yes
Which principles and best practices does this project support (please list)
- MERON is designed with all of the Principles for Digital Development in mind:
- Design with the user: MERON was designed after conversations with parents of young children who expressed concern with hands-on health surveys as well as with survey teams who were seeking additional ways to screen for malnutrition when on other missions.
- Understand the existing ecosystem: MERON was designed with the existing ecosystem in mind. Given an ecosystem that includes periodic health surveys as well as numerous other non-health interactions, MERON was designed to take the valuable data collected in health surveys and make non-health interactions useful as a malnutrition screening opportunity. Additionally, care was taken to protect dignity and be culturally aware through less invasive techniques than are currently standard.
- Design for scale: MERON is designed to improve as adoption increases, with near-instantaneous results at scale for field users.
- Build for Sustainability: Once trained, the MERON application runs with very minimal overhead, and with the proper partnership the project can sustainably run and improve itself with low cost.
- Be data driven: MERON is meant to generate the critical insights that are currently difficult to gather through health monitoring surveys alone. MERON will drive down the cost of data, making it practicable to use in estimating malnutrition impacts and informing interventions.
- Use open standards, open data, open source, and open innovation: The MERON code is open source, leverages open projects and standards. The only protected information is individual health data, which must be secured to ensure privacy and dignity.
- Reuse and Improve: MERON is designed to reuse the valuable truth data collected through health surveys and improve it such that a quick smartphone photo can screen for malnutrition in non-health related contexts.
- Address privacy and security: MERON secures individual’s private information and only uses this information for model training purposes. Use of the MERON application results only in individual private results that are not shared or exposed in a way that puts any individuals at risk.
- Be collaborative: Kimetrica’s vision for MERON is as a capability that is operated as a common good for all. We envision a mutually beneficial partnership through the sharing of resources from health surveys and broader MERON use within a population. This way, all participants including funding agency, implementing partner, and potentially malnourished people involved all benefit.
9. Does the project do no harm by design?
Has this project taken steps to anticipate, prevent and do no harm by design?
On the whole, does this project take steps to ensure that it anticipates, prevents and does no harm by design?
Yes
Is there any additional information you would like to share about the mechanisms, processes or policies that this project uses to avoid doing harm by design?
MERON takes an open approach to technology, however due to the privacy risk of personal health data, we have designed the system to operate without direct access to any data at the individual level. All training is performed in a secure system and only anonymized model parameters are available through a public interface. This, coupled with the limitations put on the system to prevent abuse in the field strongly coerces a user into only beneficial use patterns and limits risk to individuals and partnering organizations.
9.a. Data Privacy & Security
Does this project collect or store personally identifiable information (PII) data and/or content?
Yes
If yes - please list the types of data and/or content collected and/or stored by the project:
- Facial photographs
- Height
- Weight
- Age
- Gender
- Location
If yes - does this project share this data and/or content with third parties?
No
Please describe the circumstances with which this project shares data and/or content with third parties. Please add links as relevant.
Not Applicable
If yes - does the project ensure the privacy, security and integrity of this data and/or content collection and has it taken steps to prevent adverse impacts resulting from its collection, storage and distribution.
Yes
If yes - please describe the steps, and include a link to the privacy policy and/or terms of service:
Data is controlled in isolated storage and used only for offline training with separation between the data scientist and the data. Individual data points containing PII are only used for random QA sampling and never leaves controlled storage.
9.b. Inappropriate & Illegal Content
Does this project collect, store or distribute content?
No
If yes - what kinds of content does this project, collect, store or distribute? (i.e. childrens books)
Not Applicable
If yes - does this project have policies that describe what is considered innappropriate content? (i.e. child sexual abuse materials)
Not Applicable
If yes - please link to the relevant policy/guidelines/documentation.
Not Applicable
If yes - does this project have policies and processes for detecting and moderating innappropriate/illegal content?
Not Applicable
If yes - please describe the policies and processes for detecting, reporting and removing innapropriate/illegal content (Please include the average response time for assessment and/or action. Link to any policies or descriptions of how inappropriate content is handled):
Not Applicable
9.c. Protection from harassment
Does this project facilitate interactions with or between users or contributors?
No
If yes - does the project take steps to address the safety and security of underage users?
Not Applicable
If yes - please describe the steps this project takes to address risk or prevent access by underage users:
Not Applicable
If yes - does the project help users and contributors protect themselves against grief, abuse, and harassment?
No
If yes - please describe the steps taken to help users protect themselves.
Not Applicable
Development & deployment countries
List of countries this project was developed in.
- Kenya
- United States of America
List of countries this project is actively deployed in.
- Kenya