CP-AnemiC (A Conjunctival Pallor) Dataset from Ghana

Licensed biomedical scientists were given a proficiency training at the Centre for Research in Applied Biology, the University of Energy and Natural Resources, Ghana on how to capture conjunctival images from children aged 6-59 months presenting to assigned healthcare facilities within the stated period using electronic instruments (Kobo Collect v2021.2.4, Massachusetts, USA) deployed on mobile tablets (Samsung Galaxy Tab 7A, Samsung, Vietnam). The system consisted of a form to capture biodata of patients (Hb levels, age, gender), and a remark based on the Hb Value obtained during laboratory assessment, and thereafter took a picture of the image of the conjunctiva of the eyes to upload to the database which allows for easy availability and accessibility. To capture the conjunctiva, the lower eyelid was gently pushed back with the thumb with the aid of the index finger. All images were taken by laboratory personnel in ambient natural light with a standard camera of 12MP image resolution mounted on an electronic tablet. Furthermore, the cameras' spotlights were turned off when capturing the photographs to prevent excessive shine effects created by the picture quality, that is to eradicate ambient light, which dramatically influences detection or classification by the models. This method is a great technique to remove the effect of ambient light on photographs in datasets. The ROI of the conjunctiva of the eyes was extracted after the application of the triangle thresholding algorithm, in combination with the entropy grayscale image algorithm. The images (dataset) were taken from 10 hospitals situated in Ghana.