Between February and April 2018, a convenience sample of 62 children aged between birth and 59 months of age was recruited in Korle Bu Teaching Hospital, Ghana. Recruitment was carried out in accordance with University College London Ethics application 4050/003, and proposals submitted to the Ethical and Protocol Review Committee of the College of Health Sciences, University of Ghana as CHS-Et/M.5-P1.11/2017–2018. Written consent was given by the parents of the participants. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the (S1 Checklist ).
Participants were only recruited when they were clinically stable, where a blood test was requested by their attending doctor as part of their standard clinical care, and when urgent care would not be delayed by the acquisition of the images for this study. Participants were excluded if they were receiving oxygen, had received a blood transfusion in the last 24 hours, or were otherwise deemed to be very ill or unstable. Participants were excluded when conditions such as uveitis or gingivitis that affected the colouration of the imaged mucosa were recorded. Before participation, the research nurse/doctor explained the purpose of the study, and answered any questions which the parent or child raised. After all questions were answered, parents were provided with written consent forms prior to image acquisition.
For each participant, blood haemoglobin concentration was measured with a HemoCue Hb 301 point-of-care anaemia screening device (HemoCue AB, Ängelholm, Sweden) in accordance with the manufacturer’s guidelines, after which they were immediately imaged by a trained clinician. The imaging was carried out using the back-facing camera on a Samsung Galaxy S8 smartphone with a custom mobile application which captured and stored raw, lossless (Adobe Digital Negative) camera images. Using raw images means that there was minimal post-sensor processing. There were 3 pairs of images taken: (1) an image of the sclera and surrounding skin, (2) an image of the folded-over lower eyelid, and (3) an image of the lower lip. Each pair of images consisted of an image taken with the smartphone camera flash turned on, and another taken automatically afterwards with the smartphone camera flash turned off. The flash was diffused using a polymer diffuser in order to ensure the flash brightness fell within safe limits.
Participants were only recruited when they were clinically stable, where a blood test was requested by their attending doctor as part of their standard clinical care, and when urgent care would not be delayed by the acquisition of the images for this study. Participants were excluded if they were receiving oxygen, had received a blood transfusion in the last 24 hours, or were otherwise deemed to be very ill or unstable. Participants were excluded when conditions such as uveitis or gingivitis that affected the colouration of the imaged mucosa were recorded. Before participation, the research nurse/doctor explained the purpose of the study, and answered any questions which the parent or child raised. After all questions were answered, parents were provided with written consent forms prior to image acquisition.
For each participant, blood haemoglobin concentration was measured with a HemoCue Hb 301 point-of-care anaemia screening device (HemoCue AB, Ängelholm, Sweden) in accordance with the manufacturer’s guidelines, after which they were immediately imaged by a trained clinician. The imaging was carried out using the back-facing camera on a Samsung Galaxy S8 smartphone with a custom mobile application which captured and stored raw, lossless (Adobe Digital Negative) camera images. Using raw images means that there was minimal post-sensor processing. There were 3 pairs of images taken: (1) an image of the sclera and surrounding skin, (2) an image of the folded-over lower eyelid, and (3) an image of the lower lip. Each pair of images consisted of an image taken with the smartphone camera flash turned on, and another taken automatically afterwards with the smartphone camera flash turned off. The flash was diffused using a polymer diffuser in order to ensure the flash brightness fell within safe limits.
Free full text: Click here
