Anemia is a common worldwide health concern affecting almost 25% of people, including 37% of pregnant women, 40% of children between the age of 6 and 59 months, and 30% of women between the age of 15 and 49 years. Anemia is primarily diagnosed through blood tests that measure hemoglobin levels. However, these conventional diagnostic methods rely on invasive blood sampling, which is costly, time-consuming and inaccessible in resource-limited areas. To address these challenges, this study proposes a non-invasive method for anemia detection using Haar Wavelet Transform for feature extraction from conjunctiva images of patients. The essential statistical features are extracted from the conjunctiva images using Haar Wavelet Transform that captures the crucial information indicating anemia. These extracted Haar features are used for training various machine learning methods like k-Nearest Neighbour, Decision Tree, XGBoost, Gradient Boosting, Random Subspace, and Random Forest. The accuracy achieved through k-Nearest Neighbour, Decision Tree, XGBoost, Gradient Boosting, Random Subspace and Random Forest are 99.18%, 97.66%, 95.08%, 79.25%, 95.55%, and 98.24% respectively. The k-NN method outperformed the others with an accuracy of 99.18%. This remarkable performance suggests that non-invasive machine learning techniques based on conjunctiva image analysis could serve as a promising alternative for anemia detection.
