Implementation Of Deep Learning Approach For Anemia Detection Using Palpebral Conjunctival Images

Anemia is a prevalent health condition normally diagnosed using invasive blood tests. This workpresents a deep learning-based, non-invasive approach to anemia detection using conjunctivalimages. A convolutional neural network (CNN) learns deep features after preprocessing methodssuch as normalization and contrast stretching are applied. The model is trained and tested on labeledconjunctival images with related haemoglobin (Hb) values. Performance is assessed by means ofmetrics such as mean absolute error (MAE), root mean square error (RMSE), and Pearsoncorrelation coefficient (R²). High accuracy in Hb estimation is revealed through results, showinghigh correlation with laboratory findings. The method provides a quick, low-cost option forstandard blood testing, especially in resource-constrained environments. Early anemia diagnosis isfacilitated without the use of specialized laboratory facilities. Future studies will concentrate onincreasing the dataset, enhancing model generalization, and incorporating the system into a mobilebased diagnostic application to make it more accessible.