Colorimetric and Longitudinal Analysis of Leukocoria in Recreational Photographs of Children with Retinoblastoma

Abstract

Retinoblastoma is the most common primary intraocular tumor in children. The first sign that is often reported by parents is the appearance of recurrent leukocoria (i.e., “white eye”) in recreational photographs. A quantitative definition or scale of leukocoria – as it appears during recreational photography – has not been established, and the amount of clinical information contained in a leukocoric image (collected by a parent) remains unknown. Moreover, the hypothesis that photographic leukocoria can be a sign of early stage retinoblastoma has not been tested for even a single patient. This study used commercially available software (Adobe Photoshop®) and standard color space conversion algorithms (operable in Microsoft Excel®) to quantify leukocoria in actual “baby pictures” of 9 children with retinoblastoma (that were collected by parents during recreational activities i.e., in nonclinical settings). One particular patient with bilateral retinoblastoma (“Patient Zero”) was photographed >7, 000 times by his parents (who are authors of this study) over three years: from birth, through diagnosis, treatment, and remission. This large set of photographs allowed us to determine the longitudinal and lateral frequency of leukocoria throughout the patient's life. This study establishes: (i) that leukocoria can emerge at a low frequency in early-stage retinoblastoma and increase in frequency during disease progression, but decrease upon disease regression, (ii) that Hue, Saturation and Value (i.e., HSV color space) are suitable metrics for quantifying the intensity of retinoblastoma-linked leukocoria; (iii) that different sets of intraocular retinoblastoma tumors can produce distinct leukocoric reflections; and (iv) the Saturation-Value plane of HSV color space represents a convenient scale for quantifying and classifying pupillary reflections as they appear during recreational photography.

Retinoblastoma

“…102 Other pioneering efforts target software to enable cameras to detect photoleukocoria. 103 The global imaging industry could play a role in early diagnosis of retinoblastoma.…”

Crowd-sourced Ontology for Photoleukocoria: Identifying Common Internet Search Terms for a Potentially Important Pediatric Ophthalmic Sign

“… 41 Thus, we cannot know whether the participants would have noted the abnormal findings in the photographs if they were not prompted to do so. Indeed, Abdolvahabi's 19 retrospective review of a vast bank of photographs of nine children with a subsequent diagnosis of retinoblastoma demonstrated the presence of photoleukocoria well before the parent observed the sign and was prompted to seek attention.…”

“… 11 , 18 Leukocoria, however, can also result as an artefact of lighting or photographic angle—where light bounces off the optic nerve head or due to the type of camera used. 19 In developed countries, more than 50% of retinoblastoma cases will present with leukocoria. 16 , 20 , 21 In resource-poor settings, this statistic can reach as high as 80% 22 or be as low as 30% when children are more likely to present with advanced, extraocular disease as the initial leukocoria has been missed or ignored.…”

A Convolutional Neural Network approach for classifying leukocoria

“…The set of leukocoric eye images and part of the set of normal eye images overlaps the dataset used in [2]. We assumed that the set images from Flickr do not demonstrate true leukocoria, as true leukocoria is rare while bright white-eye reflection in images taken by an iPhone using the flash is quite common.…”

“…This symptom is called leukocoria (white pupillary reflection), and is present in 60% of the reported cases in the United States [1]. Researchers at Baylor University characterised leukocoria in recreational photography, and they concluded that the intensity of the symptom is an indicator of the cancer's stage [2]. An example of leukocoria compared to a "normal eye" can be seen in Figures 1a and 1b.…”