Article Text

Download PDFPDF

Use of digital camera imaging of eye fundus for telemedicine in children suspected of abusive head injury
  1. M Saleh1,
  2. S Schoenlaub2,
  3. P Desprez2,
  4. T Bourcier1,
  5. D Gaucher1,
  6. D Astruc2,
  7. C Speeg-Schatz1
  1. 1
    Department of Ophthalmology, University Hospital of Strasbourg, France
  2. 2
    Department of Pediatrics, Intensive Care Unit, University Hospital of Strasbourg, France
  1. Dr M Saleh, Service d’ophtalmologie, 1 place de l’hôpital, 67000, Strasbourg, France; maher.saleh{at}


Aim: Pilot study of the role of RetCam imaging for telemedicine in lieu of availability of ophthalmologist examination for cases of suspected abusive head injury.

Design: Cross-sectional observational study.

Participants: 21 children admitted in the paediatric units of the University Hospital of Strasbourg (France) with suspicion of abusive head trauma were included.

Methods: Children were examined by standard ophthalmoscopy. Photographs were taken using the RetCam-120 Digital Retinal Camera. Eye fundus images were stored and remotely read by an ophthalmologist. Patients also had radiographic skeletal series to look for bone fractures, and CT scan and/or MRI of the head to look for intracranial haemorrhages.

Main outcome measures: The absence or presence of retinal haemorrhages was assessed by both methods. Feasability, sensitivity and specificity of the digital camera procedure were determined.

Results: 85.7% of the children presented cerebral bleeding, and 14 out of the 21 (66.7%) had retinal haemorrhages on ophthalmoscopy. The digital camera detected the retinal abnormalities in all cases. One false-positive case was also reported. The sensitivity of the digital camera detection method was 100% with a specificity of 85.7%. 14 patients were eventually diagnosed as suffering from abusive trauma. RetCam helped establishing the diagnosis of abuse in 92.8% of these cases.

Conclusions: Digital photography compared with ophthalmoscopy has a good sensitivity and specificity in detecting retinal haemorrhages. Remote reading of RetCam-120 photographs could be a promising strategy in detecting children with abusive head trauma.

View Full Text

Statistics from

The incidence of abusive head trauma in infants, known as the shaken baby syndrome (SBS), is estimated to be between 171 and 29.7/100 000 person-years.2 SBS is more often seen in male children under 12 months of age1 from low socio-economic backgrounds. Up to 30% of these children die, and 30% keep severe neurological deficits.3 4 The prognosis of SBS is thereby severe, and physical abuse is the leading cause of traumatic death during infancy.5 Despite the awareness of SBS in the medical field, one-third of cases are misdiagnosed on their first presentation in an emergency room.2 This underlines the need to introduce innovative screening tools to detect SBS.

The RetCam (Clarity Medical Systems, Pleasanton, California) is a digital colour fundus camera with a contact fibre-optic providing a 120° field imaging that allows rapid capture of eye fundus images.6 This camera, which was mainly developed for preterm-retinopathy screening, was also tested to estimate various fundus abnormalities, including changes of optic nerve head cupping in paediatric glaucoma.7 A pilot study has suggested the usefulness of this device in screening children suspected of abusive head trauma.8 We compared the RetCam findings with ophthalmoscopy, in a group of children highly suspected of having suffered head abuse.


All children admitted during a 36-month period to the paediatric units at our tertiary care centre for suspected abusive head trauma were included in this study.

Patients underwent an extensive medical examination, including an ophthalmological examination with both direct and indirect funduscopy (gold standard), skeletal survey, a computed tomography (CT) scan and/or MRI of the head, and eye fundus imaging provided by the RetCam 120 (RetCam II, Clarity Medical Systems) with a wide angle lens of 120° (the same model as that used in retinopathy of prematurity (ROP) screening).

Suspicion of abusive head trauma encompassed children under the age of 3 years with falls reported by a caretaker without any other witness, intracranial haemorrhages, coma from unknown origin and the siblings of patients who experienced abusive head trauma.

The ICU team was already acquainted with RetCam examination in infants, an experience acquired during routine preterm-retinopathy screening. The technical support department of the digital camera manufacturer provided training for the senior paediatrician, who, in turn, supervised the procedure. Two hours of practice for each team member were required before starting the recordings.

The paediatric team took digital photographs of the fundus of both eyes of the children, and then a senior ophthalmologist examined all eyes by ophthalmoscopy.

RetCam imaging was performed after dilation with tropicamide and phenylephrine 2.5%. Surface anaesthesia was obtained by the instillation of a drop of oxybuprocaine before the insertion of an eyelid speculum. An interface between cornea and camera was provided by hydroxyprolyl methylcellulose. The head position of the infant was gently maintained by the assistant while the use of midazolam or oral sucrose with a dummy kept the child safe and calm during the procedure. The duration of the RetCam procedure for both eyes was approximatively 30 min, with an average of four images acquired per eye (range from three to 10 pictures depending on the child’s cooperation). Digital images were saved and transmitted for remote reading by mail, via the internal network of the institution to another ophthalmologist located in the Department of Ophthalmology, 6 km away. The latter, masked to the examining ophthalmologist’s results, compared RetCam conclusions with the findings of ophthalmoscopy, to ensure that the digital photos did not miss any eye with haemorrhage and to eliminate any possible haemorrhage caused by the RetCam procedure itself. Retinal drawings and personal notes were used to compare the findings of both examiners.

Major fundus abnormalities were expected to be retinal haemorrhages, classified as subtle (less than 10 RH), moderate (between 11 and 30 RH) or severe (more than 30 RH or extensive retinal haemorrhage of more than twice the optic disc surface).


Clinical findings

Twenty-one children have been included during a 36-month period (15 males, six females). The mean age of the studied individuals was 4.65 (SD 5.59) months with a range extending between 3 days and 24 months (median age of 3 months). Trauma was reported in six cases (fall from table or slaps on the face). Fifteen children were admitted to the paediatric intensive care unit with conscience disorder of unknown origin. Ten children presented obvious signs of physical abuse and neglect, including cutaneous haematomas of various ages, dehydration and growth retardation.

Table 1 summarises the age, sex, neuroimaging, RetCam and ophthalmoscopic findings, outcome and final diagnosis of overall patients.

Table 1 Age, sex, neuroimaging, RetCam and ophthalmoscopic findings, outcome and final diagnosis of overall patients

The group of patients included in this study was representative of the classic population suspected of inflicted head injury that would benefit from a screening for early detection of SBS. The incidence of cerebral bleeding, with minor or no history of trauma reported to the physician, reached 85.7%, compared with 80% in other case series of suspected abusive head trauma,9 and external signs of abuse or neglect were also absent in up to half of cases.10

Neuroimaging results

Computed axial tomographic scans or MRI of the head were performed in all patients and revealed intracranial haemorrhages in 18 patients (85.7%). Subdural haematomas (n = 11), extradural haematomas (n = 2), subarachnoid haemorrhages (n = 3) and intracerebral haemorrhages or oedema (n = 4) were found isolated or in association. The high incidence of cerebral bleeding explained the frequency of severe neurological deficits at the admission.

Retinal findings

RetCam provided readable images in all the assessed eyes and detected abnormal eye fundus in 15 patients (71.4%). Most abnormalities were retinal haemorrhages, subtle (fig 1), moderate (fig 2) or severe (fig 3). Moderate and severe retinal haemorrhages were multilayered (pre-, intra- and subretinal), even if it was difficult to ascertain this in a 2D plane image. Most of the RH were numerous, bilateral (84.6%), and tended to extend to the ora serrata. One child also presented moderate vitreal haemorrhage.

Figure 1

Small and round macular haemorrhages (left eye).

Figure 2

Numerous peripapillar haemorrhages (right eye).

Figure 3

Numerous intra- and preretinal confluent haemorrhages (right eye).

Therefore, the digital images show the classic pattern of SBS-related retinal haemorrhages in eight patients with abusive head trauma and unilateral haemorrhages in three (14%) others, a rate frequently reported in SBS (16%).11 12 Neither traumatic retinoschisis nor retinal detachment was noticed, even if some authors have assumed that retinoschisis is diagnostic of SBS in this context.13 Moreover, one case of macular fold was detected at the RetCam examination and confirmed by ophthalmoscopy, accordingly with the incidence of macular folds reported in SBS.14

After complete medical and social assessments, diagnosis of SBS was eventually confirmed in 14 children. Eleven (78.6%) of the 14 patients with RH on eye funduscopy at the initial presentation suffered from head trauma from abusive origin, and seven children suspected of being abused were finally diagnosed as sustaining head injuries from other causes (one case of birth-related trauma, two cases of accidental head traumas, three cases of metabolic diseases including urea-cycle dysfunction and fatty-oxidation disorders responsible for violent seizures, and one case eventually remaining unexplained). Even if RH is known to be a common complication of vacuum delivery extraction15 and can be associated with accidental head traumas,16 the rate of RH in these situations is lower than in SBS, where it can reach 50 to 100%.16

Ophthalmoscopy helped to differentiate haemorrhages seen on the RetCam from artefacts in one set of images. Compared with ophthalmoscopy, which was the gold standard for detecting retinal haemorrhages, RetCam Imaging showed a sensitivity of 100% (calculated as the number of true positives/(number of true positives+number of false negatives)), and a specificity of 85.7% (number of true negatives/(number of true negatives+number of false positives)).

The positive predictive value of the RetCam in detecting these haemorrhages (calculated as number of true positive/(number of true positives+number of false positives)) reached 93.3% (table 2).

Table 2 Retinal findings with RetCam digital camera and direct/indirect ophthalmoscopy (n = 21)

Four patients eventually died, and seven held on severe neurological deficits. All of the patients holding severe prognosis, but two, also presented retinal haemorrhages. This apparent association between presence of retinal haemorrhages and bad prognosis was not statistically significant (Fisher test, p = 0.182. GraphPad Instat V3.06).


The presence of RH in a context of head trauma in children is highly predictive of inflected SBS9 where a violent shaking is presumed to cause acceleration–deceleration movements responsible for sudden tears of bridging veins causing subdural haemorrhage. In the same way, the shaking can produce retinal haemorrhages after tractions between the vitreous, blood vessels, retina and optic nerve fibres.9 The importance of these RH is correlated with the severity of parenchymal brain injuries.11 Therefore, RH are strongly associated with cerebral lesions in a context of shaken baby syndrome.

In this series, retinal haemorrhages were detected at both ophthalmoscopic and RetCam examinations, in 14 of the 18 (78.6%) patients suffering from cerebral bleeding, when Kivlin et al12 reported such an association in 83% of cases.

If we consider retinal haemorrhages in a context of trauma as a marker of abuse, detecting them could be a means to screen children suspected of suffering from shaken baby syndrome. This approach is controversial because it might be suitable for detecting acute abusive head lesion but not occult head injury. Some authors assume that ophthalmological examination is a poor screening method in this context and advise proceeding directly to CT or MRI.17

In comparision, RetCam imaging has already demonstrated its usefulness in telemedicine screening for ROP.1821 Intraphysician agreement in this context, between ophthalmoscopic examination and telemedical interpretation, appeared to be high.20 RetCam succeeded in identifying most of the cases of referral-warranted disease, with a sensitivity reaching 100%18 21 and a specificity ranging between 95%19 and 97.5%.21 Training was easily implemented in these reports, and there was no adverse outcome to the use of RetCam. However, some authors were concerned by the loss of 3D perspective provided by ophthalmoscopy and reported high rates of missed cases of ROP by the RetCam either in zone 3 or in the outer part of zone 2.22 Difficulties in photographing the paediatric retina till the ora because of the restricted mobility of the camera head were incriminated. They concluded that hardware and software improvements are needed before the RetCam 120 can be used routinely in ROP screening. We did not encounter such an issue while imaging the fundus periphery of children suspected of suffering from head abuse. Our concern was more about the false-positive case. A child suspected of suffering from head abuse presented a blurred round dark image on the eye fundus imaging that could be consistent with a single haemorrhage, but this finding was contradicted by ophthalmoscopy. The implications of this, 5% of patients studied, is significant in terms of the effect it may have on diagnosis and the family if used inappropriately. The positive predictive value of this tool could be shadowed by its negative predictive value as has already been shown in ROP screening.21 For this reason, we suggest that all positive RetCam findings be subjected to an ophthalmological exam.

In addition, the use of the RetCam 120 in telemedicine strategies to assess optic head cupping in paediatric glaucoma was also tested and revealed to be disappointing.7 The lack of reliability of the technique was due to variations in contrast, luminance and image pixelation resulting in poor agreement between the observers. In contrast, in abusive head-trauma screening, fundus abnormalities were easier to document and were not masked by fluorescein on cornea.

In this report, RetCam finally helped in the diagnosis of abusive head injuries in 92.8% of patients. It provided real-time, high-resolution images of the eye fundus, completing routine neuroimaging evaluation. It allowed subtle retinal changes to be detected, even in uncooperative and very young children. The sensitivity (100%-CI95%: 77.8–100%) and specificity (85.7%-CI95%: 42.1–99.6%) of this technique were both satisfactory, even if these numbers should be taken with caution because of the limited size of the sample.

The results of this pilot study of the role of RetCam imaging for telemedicine in lieu of the availability of ophthalmologist examination for cases of suspected abusive head injury are encouraging and suggest that the practical usefulness of the RetCam in the ICU setting should be evaluated more extensively. In particular, it should be weighed up against the cost and maintenance requirements of the device, and with the need to train paediatricians and nurses how to use the camera. The legal veracity of the digital images should also be defined.


RH are common findings in children who experienced abusive head injury. RetCam provides a valuable screening tool to detect them. The RetCam exam can be performed at the patient’s bedside with minimal training and rapid acquisition. It can be used in paediatric intensive care units, where management of SBS requires decisions to be taken rapidly. Digital data can be transmitted for specialised remote assessment and stored for further comparisons of eye fundus images that could be useful for follow-up. Thus, RetCam could play a role in future telemedicine strategies for abusive head-trauma screening in children.


View Abstract


  • Competing interests: None.

  • Ethics approval: Ethics approval was provided by the Ethical Committee in Pediatrics of Hautepierre Center.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • At a glance
    Harminder S Dua Arun D Singh