NAIL: Nail based Artificial Intelligence image Learning for measuring iron status

Low iron levels are detrimental for your health, and possibly even for cognitive and mental well-being. As tests of iron levels are expensive and invasive, requiring a blood draw, large and longitudinal data on how iron may impact well-being is lacking. A non-invasive method to measure iron levels would therefore benefit research and clinical practice.  Luckily, there is some evidence to suggest that the color of your fingernails could be indicative of your iron status. This offers the potential for using images of fingernails to detect iron levels in a non-invasive and low-priced way.

Combining expertise in machine learning, image analysis and app development with a collaboration with Sanquin blood bank, offers a unique opportunity to develop and validate a smartphone app able to measure hemoglobin levels from a picture of your fingernails.

Sanquin takes thousands of hemoglobin measurements a month, from healthy donors of all ages. By asking them to supply an image of their fingernails and by asking them to fill out questionnaires on cognitive and emotional well-being, a prototype can be developed of an artificial intelligence driven app able to assess iron levels through photographs of fingernails taken with one’s smartphone. At the same time, the effect on cognitive functioning and depressive symptoms can be assessed from a large, heterogeneous and healthy sample of blood donors.

Project objectives

As the national blood bank organisation Sanquin routinely performs more than 740.000 hemoglobin measurements from more than 380.000 healthy individuals every year, the project team is in the unique position to achieve the following objectives:

1. collect a unique and large dataset containing hemoglobin measurements, nail images, cognitive functioning and symptoms of depression, in a large, heterogeneous and healthy sample of donors;
2. develop an algorithm able to assess hemoglobin based on fingernail images, and to implement it in a working app prototype;
3. assess the association between hemoglobin, cognitive functioning, depression and general well-being;
4. assess and incorporate ethical principles into the development process, allowing that ethical issues are addressed as early as possible, already in the design stage ('ethics by design'), i.e. through stakeholder interviews and consideration of Ethical, Legal and Social Issues (ELSI).

Potential impact

• Scientific impact:
  • This project will result in a truly unique dataset, consisting of health data (hemoglobin) and questionnaire data on wellbeing in terms of (mental) health, from a large sample of healthy people.
  • The study will add to the knowledge about the link between hemoglobin and well-being as well as cognition.
  • The dataset consisting of healthy participants is also very useful for researchers who usually focus on patient groups, but who lack control data from healthy subjects.
  • The app can open up the possibility to assess hemoglobin for other academics who cannot collect blood samples.
• Societal impact
  • For blood banks, the algorithm and app potentially provide a better way to manage their donor's hemoglobin values.
  • Being able to monitor hemoglobin through an app could provide a solution for people whose hemoglobin need to be monitored (i.e., patients, pregnant women), but who live far away from facilities able to do a blood measurement, are scared of needles, or live in countries where there is no money or facilities for blood sample analysis.
  • The project also offers a reflection upon ethico-legal implications to minimize harms and maximize societal benefits.

Duration

The project will run for nine months starting from October 2023 onwards.

Multidisciplinary project team

• Lead applicant Judita Rudokaitè is a PhD candidate at the Department of Cognitive Science and Artificial Intelligence and affiliated with Sanquin bloodbank. Her PhD research aims at developing a serious game to predict needle fear and anxiety using facial image analysis and machine learning methods.
• Dr. Elisabeth Huis in ’t Veld is an Assistant Professor at the Department of Cognitive Science and Artificial Intelligence and has a scientific education in Medical and Clinical Psychology. In collaboration with Sanquin, she is studying the emotional and physical reactions to needles with the NWO VENI funded project FAINT (FAcial INfrared Thermal imaging in the prevention of needle induced fainting).
• Dr. Sharon Ong is an Assistant Professor at the Department of Cognitive Science and Artificial Intelligence. Her research focuses on artificial intelligence for image and video analysis.
• Mart Janssen PhD is Principal investigator Transfusion Technology Assessment at Sanquin bloodbank and has a scientific education in Mechanical Engineering. His research interest lies with applied mathematical/statistical modelling such as modelling Hb.

Judita, Elisabeth, Sharon and Mart are all part of the team that successfully developed an image-based algorithm to detect needle fear from video and is developing an AI driven serious game application (www.ainar.io).

• Dr. Nicola Ballhausen and dr. Alexandra Hering are both Assistant Professor at the Department of Developmental Psychology and experts in cognitive psychology and well-being across the lifespan. Their expertise will help to analyze the links between Hb and well-being, and Hb and cognition in complex repeated measure analyses.
• Tjaša Petročnik is a PhD candidate at the Tilburg Institute for Law, Technology, and Society. She has expertise in health data governance and regulation and is interested in the associated Ethical, Legal and Social issues (ELSI) regarding AI development.

This project is funded by Tilburg University’s Digital Sciences for Society program: