Non-invasive multimodal dataset for the detection of iron deficiency anemia in young adults: fingertip videos, palm videos, and nail photographs

Iron deficiency anemia affects a significant proportion of the young population in both rural and urban areas of Peru. In response to the need for non-invasive, accessible, and reproducible methods for its detection, we developed this dataset as part of a research project funded by the Universidad Nacional de San Martín, which applies computer vision techniques to automatically classify patients as anemic or non-anemic. The aim is to provide a standardized base of videos and images that supports the development and validation of classification and regression models to estimate hemoglobin levels without the need for blood extraction. This data paper presents a multimodal dataset composed of non-invasive visual records collected to facilitate the detection of iron deficiency anemia in young adults through machine learning models. The dataset includes 909 fingertip videos, 909 palm videos (with controlled hand opening), and 909 nail photographs, all linked to individual clinical data such as age, sex, hemoglobin level, and symptomatology.