Bibliographic citations
Navarro, J., (2024). Modelo de visión artificial empleando imágenes de las uñas para la detección no invasiva de anemia ferropénica en estudiantes universitarios [Universidad Nacional de San Martín. Fondo Editorial]. http://hdl.handle.net/11458/6106
Navarro, J., Modelo de visión artificial empleando imágenes de las uñas para la detección no invasiva de anemia ferropénica en estudiantes universitarios []. PE: Universidad Nacional de San Martín. Fondo Editorial; 2024. http://hdl.handle.net/11458/6106
@mastersthesis{renati/1054049,
title = "Modelo de visión artificial empleando imágenes de las uñas para la detección no invasiva de anemia ferropénica en estudiantes universitarios",
author = "Navarro Cabrera,Jorge Raul",
publisher = "Universidad Nacional de San Martín. Fondo Editorial",
year = "2024"
}
In this study, the use of a machine vision model based on DenseNet is proposed for the non-invasive detection of iron deficiency anemia using nail images captured with smartphones, responding to the need for accessible and efficient diagnostic methods in educational contexts. The main objective was to validate the results of the Rad-67 hemoglobin meter using this machine vision model. A detailed protocol was designed for the collection of fingernail images from students, and several models were built and evaluated, with DenseNet standing out for its performance. In the validation set, DenseNet showed an accuracy of 0.6983, a sensitivity of 0.6477, an F1-Score of 0.6525 and an AUC ROC of 0.7409, indicating a positive correlation with the Rad-67 meter results and demonstrating its ability to classify anemic states. Despite not achieving extremely high accuracy, the progress achieved is significant, and additional adjustments to the model's preprocessing and configuration may further improve its performance. This approach promises to facilitate early and accurate diagnosis of anemia, promoting the use of mobile technologies and machine learning for more accessible and efficient diagnostics at a global level.
This item is licensed under a Creative Commons License