CNN-Based Anemia Detection in Blood Smear Images
Maribelle Dequilla Pabiania, Kristine Joyce P. Ortiz, Daniel Ehsan Waje Abdel Bari, Princess Czarina Alcosiba Marilag, Patrick Aren Acosta Nisperos · 2024
Anemia is a global health concern impacting vulnerable populations which necessitates improved diagnostic methods beyond traditional approaches such as complete blood count. This study focuses on utilizing Convolutional Neural Networks (CNN) and image processing to enhance the accuracy and efficiency of anemia detection in blood smear images. The research objectives encompass the development of a CNN model tailored for automated anemia detection and the evaluation of its performance against traditional manual methods. Leveraging datasets from Roboflow Universe and Kaggle, the CNN model is trained and tested through 31 trials, yielding a reliability coefficient of 71%. The findings of the study demonstrate the model's proficiency, with 15 accurate predictions of anemia and seven typical cases, indicating the potential for technologically driven improvements in anemia diagnosis. Recommendations include refining the model, addressing misclassification rates, and exploring additional data sources for further enhancements.