Diagnosis of Iron Deficiency by Artificial Intelligence Analysis of Eye Photography. (CaFerIA)

University Hospital, Clermont-Ferrand

Status: Not yet enrolling

Conditions: Iron-deficiency

Treatments: Other: photographs of each eye

Study type: Observational

Funder types: Other

Identifiers

NCT05395468

AOI 2021 LOBBES

2021-A03087-34 (Other Identifier)

Details and patient eligibility

About

The objective of our work is to predict the value of ferritin from the eye, thus constituting an original, non-invasive diagnostic method of iron deficiency. To be usable in real life, the algorithm must be comparable to the performance of the reference diagnostic test (determination of ferritin), allowing to obtain a sensitivity of about 90% and a specificity > 95%.

Full description

Currently, the diagnosis of iron deficiency is invasive, as it requires a venous puncture for serum ferritin assay and blood count analysis to diagnose iron deficiency anemia. This dosage is expensive and represents a major brake in the large-scale screening of iron deficiency, especially in developing countries. Most of the clinical signs of iron deficiency (asthenia, cheilitis, glossitis, alopecia, restless legs syndrome) are not very specific and the diagnosis is most often fortuitous or carried out as part of screening in a population at risk.

Iron is essential for many functions of the body, including the synthesis of collagen: in case of deficiency, it is produced with an altered and finer structure. In the eyes, the sclera consists of collagen type IV, whose thinning causes the visualization of the choroidal vessels responsible for a characteristic blue tint. A preliminary work carried out by our team made it possible to measure the increase in the amount of blue color in the sclera of deficient patients, objectifying this clinical sign for the first time. From photographs of patients' eyes, we extracted the percentile of blue contained in the pixels of the digital images of the sclera. This work continued with the automation of the recognition of eye structures, especially the sclera.

In order to improve the diagnostic performance of this original and non-invasive method, we want to apply deep-learning methods, which have already been proven in several areas: related to ophthalmology but also in a very encouraging way in the non-invasive diagnosis of anemia.

The objective of our work is to predict the value of ferritin from the eye, thus constituting an original, non-invasive diagnostic method of iron deficiency. To be usable in real life, the algorithm must be comparable to the performance of the reference diagnostic test (determination of ferritin), allowing to obtain a sensitivity of about 90% and a specificity > 95%.

Enrollment: 200 estimated patients

Sex: Female

Ages: 18+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• Female sex

• Age ≥ 18 years old

• Able to express non-opposition to participation in rese

• Patients affiliated to a social security scheme

• Screenng for iron deficiency within 15 days of inclusion, including

  • Blood count : value of hemoglobin, mean blood volume
  • Serum ferritin

Exclusion criteria

• Personal history of severe trauma or surgery of both eyes (apart from refractive surgery performed more than 3 months ago)
• Personal history of hereditary connective tissue pathology including Marfan's disease, Ehler Danlos syndrome, imperfect osteogenesis.
• Personal history of pathology responsible for chronic hemolysis due to yellow coloration induced by hyperbilirubinemia: sickle cell disease, major thalassemia.
• Prolonged treatment with minocycline (> 1 month).
• Oral or intravenous martial supplementation started more than 15 days prior to taking the sclera photographs.
• Person deprived of liberty by administrative or judicial decision or placed under judicial protection (guardianship or supervision)
• Pregnant or breastfeeding woman
• Expression of opposition to research.