Red eyes ruin your family photographs. They also tell you that all is well

Nobody likes red eyes in photographs. The whole imaging industry spends its time making them disappear. That is a shame, because this flaw is in fact a health bulletin: it means that light entered the eye, found a well-vascularised retina at the back, and came out again. What deserves attention is never the red. It is the eye that, across several full-face photographs, fails to turn red, and returns a white, yellow or pearly reflection instead.

THE FLAW THAT REASSURES

Why the flash turns your children into red-eyed creatures

The mechanism comes down to a single sentence of elementary physics. In a dimly lit room the pupil dilates: in a child it can reach six to eight millimetres across. The flash fires, for a few milliseconds, from a point a few centimetres from the camera lens. Light pours in through this wide opening, crosses the cornea, the crystalline lens and the vitreous, and strikes the back of the eye. There it meets the retina, and above all the layer that lies just behind it: the choroid, one of the most richly vascularised tissues in the entire body.

Blood absorbs blue and green, and sends red back. Part of that red light leaves through the same opening, the pupil, and travels back towards its source, so more or less towards the camera lens, since it sits almost in the same place as the flash. The sensor records two red discs. What you take for a photographic mishap is in fact a picture of the back of your own eye, obtained without meaning to.

Three conditions have to come together for the phenomenon to occur: a dilated pupil, and therefore a dark room; a light source almost in line with the camera lens; and a clear optical axis, that is to say a transparent cornea, crystalline lens and vitreous. It is this third condition that gives the red reflex its medical value. If it appears, strong and symmetrical, then nothing stopped the light on its way.

Two technical developments have quietly made this signal rarer. The first is red-eye reduction mode, which fires a pre-flash designed to constrict the pupil before the picture is taken: it does exactly what is asked of it, and removes the retinal reflection along the way. The second is the geometry of phones, where the flash has moved away from the camera lens and where highly sensitive sensors often manage without a flash at all. In other words, today’s family photographs carry less information about the back of the eye than those taken twenty years ago.

THE WORD

Leukocoria: the name for the reflection that is no longer red

The term comes from the Greek leukos, white, and korê, the pupil, literally “white pupil”. It does not name a disease. It names an appearance: the pupil sends back a white, yellowish or pearly glow instead of the expected red. Faced with this sign, the ophthalmologist does not make a diagnosis, he opens a list. And that list is longer than most people imagine.

It contains congenital cataract, in which the clouded crystalline lens reflects light before it reaches the back of the eye; Coats disease, an abnormality of the retinal vessels that produces yellowish exudates; persistent fetal vasculature, when embryonic structures have failed to regress; certain retinal detachments; colobomas, those closure defects that let the white of the sclera show through; the after-effects of infections such as toxocariasis; patches of myelinated nerve fibres, entirely benign; and high refractive error, in particular marked myopia, which on its own is enough to change the colour of the reflection.

And there is retinoblastoma, the malignant tumour of the retina in children. It is the cause everyone has in mind, and it is the reason why leukocoria is never simply left under observation. But it is far from being the commonest.

THE FALSE ALARM

When the white reflection means nothing at all

An observation published in Pediatrics in 2025 deserves to be known by everyone who scrutinises photographs of their children. Two children, a girl of thirteen months and a boy of three, were referred to paediatric ophthalmology for a leukocoria spotted on smartphone pictures. Full examination, with the pupil dilated, was normal in both cases. Going back over the photographs, the authors worked out what had been photographed: the detail of the optic nerve head, which is naturally paler than the rest of the back of the eye.

They propose the term physiological photoleukocoria for this phenomenon. When the child looks slightly to one side at the moment the shutter fires, the axis of the flash falls not on the macula but on the optic disc, and the sensor records a pale glow that is perfectly normal. The same mechanism explains the pale, grey or golden reflections seen when the head is turned, when the light is glancing, or when the pupil has not had time to dilate.

That is no reason to ignore a white reflection. It is a reason not to spend a sleepless night before the appointment. The useful criterion, in practice, is reproducibility: an abnormal reflection that comes back on several full-face photographs, taken on different dates, and always on the same side, weighs far more heavily than a single image in which the child was looking elsewhere.

THE TUMOUR

Retinoblastoma, a cancer revealed by a visible detail

Retinoblastoma is the commonest malignant intraocular tumour in children. Its incidence is stable and low: roughly one birth in fifteen to twenty thousand, which in France represents, as an order of magnitude, a few dozen new cases a year. The reference review published in 2006 by the Institut Curie team in Orphanet Journal of Rare Diseases recalls that the two commonest presenting symptoms are precisely leukocoria and strabismus.

The disease falls into two forms. About 60% of cases are unilateral and discovered at a median age of around two years; about 40% are bilateral and discovered earlier, at around one year. This difference in age is no coincidence: bilateral forms are always linked to a constitutional (germline) mutation of the RB1 gene, most often arising de novo, with no family history; the mutation is present from the very first cell, which speeds up the appearance of tumours. In a child carrying such a mutation, the risk of developing a retinoblastoma exceeds 90%.

A French review published in Archives de Pédiatrie clarifies a point that is often misunderstood: among the unilateral forms, which appear isolated, 10 to 15% nevertheless carry a constitutional mutation. That is what justifies a systematic genetics consultation, even when only one eye is affected: not for the child alone, but for their brothers and sisters and their descendants.

THE GAP

What decides the outcome is not the tumour, it is the delay

The clearest demonstration comes from a worldwide prospective study published in The Lancet Global Health in 2022. It followed 4,064 children treated in the same year across 149 countries. Median age at diagnosis was 23.2 months, with an interquartile range of 11.0 to 36.5 months. But the striking figure lies elsewhere: three-year survival reaches 99.5% in high-income countries, against 57.3% in low-income countries.

The same study gives the key to that gap. Spread of the tumour beyond the eyeball at the time of diagnosis was present in 0.8% of children in high-income countries and in 42.9% of those in low-income countries. It is not the biology of the tumour that differs from one continent to another. It is the moment at which it is seen. An earlier analysis in JAMA Oncology had already reached the same conclusion by cross-referencing clinical presentation with national income level.

It would nonetheless be misleading to conclude that in Europe everything is settled. A prospective analysis of 483 patients from 40 European countries, published in Eye in 2026, reports 12 deaths, or 2.5%, but also enucleation in 58% of the children. In other words: in a health system that works, what is at stake in early diagnosis is no longer so much the child’s survival. It is saving the eye, and the sight in that eye.

SIXTEEN MONTHS AHEAD

The family photographs knew before the doctors did

In 2019, an American team bringing together Baylor University and Harvard published in Science Advances a piece of work whose protocol alone is worth the detour. The researchers asked the families of twenty children with an eye disease, and those of twenty control children, to hand over their entire personal photo library. In all, 52,982 photographs were sifted by an algorithm designed to spot abnormal pupillary reflections.

In 80% of the affected children the sign was already present on pictures predating the diagnosis, on average 1.3 years earlier, with a 95% confidence interval running from 0.4 to 2.3 years. Sixteen months, on average, during which the information lay asleep in a phone or an album. The cohort was not made up only of retinoblastomas: it also included Coats disease, cataracts, amblyopia and high hyperopia, which shows that the pupillary reflection tells us about far more than cancer alone.

This result calls for a cautious reading. The numbers are small, and above all the analysis is retrospective: the researchers knew what they were looking for, and in which children to look for it. So it does not prove that software would have detected these cases blind in the general population. What it does establish, on the other hand, is solid and sufficient: the information really was there in ordinary pictures, long before anyone became concerned.

THE APPS

Why a phone will not replace an examination

Since that publication, several apps have offered to scan the photographs on a phone automatically. A diagnostic meta-analysis in the European Journal of Ophthalmology pooled twelve studies evaluating these machine-learning tools. Pooled sensitivity there reaches 0.972, a spectacular figure, one that the authors themselves qualify with severe warnings: more limited specificity, strong heterogeneity between studies, and substantial methodological bias, starting with image sets that were already selected and already labelled.

The nuance is not academic. Performance measured on selected images says nothing about performance in the general population, where the overwhelming majority of pale reflections will be artefacts. False positives are numerous there, and each one is paid for in parental anxiety and unnecessary consultations. False negatives are more formidable still: an app that answers “nothing to report” risks delaying the very consultation that the initial doubt would have prompted.

The right way to use these tools is therefore the one that reverses the burden: they can help draw the eye to a photograph, never conclude that there is nothing there. A reflection that causes concern is shown to a doctor, whatever the algorithm’s answer.

THE EXAMINATION

What a doctor checks in a few seconds

The examination is called the red reflex test, or Brückner test. All it requires is a direct ophthalmoscope, a darkened room and a little patience with the child. In practice, the doctor stands about thirty centimetres away, lights both eyes at the same time, and compares. He looks at three things: the colour of the reflection, its brightness, and its symmetry. A difference between the two eyes counts for more than any single appearance, because every child has their own background pigmentation.

The American Academy of Pediatrics set this practice out formally in a recommendation published in Pediatrics in 2008: the red reflex should be examined in every newborn, every infant and every child, and any abnormality should lead to an ophthalmological opinion. In France these checks are among the mandatory health examinations recorded in the child health record book, in particular those of the first days of life, the ninth month and the twenty-fourth month.

The same examination also tells us about the alignment of the eyes. The small bright dot that the flash leaves on the cornea should sit in the same position in both eyes. If it is centred on one side and shifted towards the edge on the other, that is the sign of a strabismus: this is the Hirschberg test, and it can be read on an ordinary photograph just as well as in the consulting room. In everyday practice, a strabismus that persists beyond four months is never trivial, and is never a matter of appearance: it is the possible starting point of an amblyopia, and sometimes the first sign of a lesion at the back of the eye.

If doubt persists, what follows is simple: an examination of the back of the eye with the pupil dilated, if necessary under general anaesthesia in the very young, completed where needed by an ocular ultrasound scan and magnetic resonance imaging. These examinations are well tolerated and allow the question to be settled. Recourse to general anaesthesia, when it is necessary in a very young child, is discussed case by case with the parents.

WHEN TO SEEK ADVICE

The situations that warrant a prompt opinion

• A white, yellow or pearly reflection found on several full-face photographs, taken on different dates, and always on the same side.
• A strabismus that persists after the age of four months, or that appears when it was not there before.
• An eye that does not follow faces or objects, or a child who repeatedly bumps into things on one side.
• A difference in the size of the pupils, or a change in the colour of the iris on one side only.
• A red, painful eye that does not settle, or an eyeball that looks larger than the other.
• Any loss of vision reported by the child, whatever their age.

In all these situations, the aim is not to reach a diagnosis at home. It is to obtain an examination of the back of the eye, which is the only step capable of answering the question. If a picture raises doubt, it is worth bringing the phone to the appointment: the photograph is part of the examination.

KEY POINTS

Key points

Red eyes in photographs are not a flaw to be corrected, but a signal to be read. They show that light made the round trip to the retina without meeting an obstacle. Red-eye reduction mode, by constricting the pupil before the shutter fires, removes that signal: for a checking photograph, it is better to switch it off, stand in dim light and photograph the child face-on, from about a metre away.

A white reflection is not synonymous with cancer, and many pale reflections have no significance at all. But none of them is left under observation, because the only way to tell the difference is to look at the back of the eye. What the literature shows is that delay is the decisive variable, and that this delay often plays out in a photo album, between the moment the sign appears and the moment someone decides to show it.

FREQUENTLY ASKED QUESTIONS

Frequently asked questions

Scientific sources

1. Autonomous early detection of eye disease in childhood photographs. Sci Adv. 2019. PMID: 31616792
2. Smartphone-Detected Physiologic Photoleukocoria With Visible Optic Nerve Details. Pediatrics. 2025. PMID: 40074014
3. The Global Retinoblastoma Outcome Study: a prospective, cluster-based analysis of 4064 patients from 149 countries. Lancet Glob Health. 2022. PMID: 35839812
4. Retinoblastoma outcomes in Europe: a prospective analysis of 483 patients from 40 countries. Eye (Lond). 2026. PMID: 41408393
5. Retinoblastoma. Orphanet J Rare Dis. 2006. PMID: 16934146
6. Retinoblastoma update. Arch Pediatr. 2016. PMID: 26679524
7. Incidence, mortality, and survival trends of retinoblastoma in children: a systematic review and meta-analysis of population-based studies. Ital J Pediatr. 2026. PMID: 42374466
8. Can machines see cancer? A systematic review and diagnostic meta-analysis of machine learning in retinoblastoma and leukocoria detection. Eur J Ophthalmol. 2026. PMID: 40910534
9. Red reflex examination in neonates, infants, and children. Pediatrics. 2008. PMID: 19047263
10. EyeScreen: Development and Potential of a Novel Machine Learning Application to Detect Leukocoria. Ophthalmol Sci. 2022. PMID: 36245758
11. Retinoblastoma. Lancet. 2012. PMID: 22414599
12. Global Retinoblastoma Presentation and Analysis by National Income Level. JAMA Oncol. 2020. PMID: 32105305

Further reading

• The retina: anatomy, examinations and disorders
• Cataract: symptoms, examinations and surgery
• Retinal detachment: recognising the emergency
• Dr Moïse Tourabaly — Ophthalmologist

Cachan practice · Tel. +33 1 45 47 08 11

Medical disclaimer

This article is for information only. A personalised ophthalmological opinion remains essential for any treatment decision.

The interactive module is a graphic reconstruction produced from descriptions published in the medical literature. It does not reproduce the eye of any individual child, is neither a test nor a screening tool, and no conclusion about a child’s vision can be drawn from it. The photographs illustrating this article are royalty-free illustrative images, unrelated to any disease. The figures quoted come from the publications referenced above and depend on the populations studied, which explains why they vary from one study to another. Any abnormality of the pupillary reflex, any persistent strabismus and any loss of vision in a child warrant an eye examination, which alone can determine the cause.