ENHANCING MALARIA DETECTION WITH AI: AN ATTENTION-DCNN APPROACH

Abstract

Malaria detection remains a vital area of focus in global health, with early and accurate diagnosis being crucial for effective treatment and management. Conventional microscopy methods, while standard, require considerable expertise and time, leading to potential delays in diagnosis. Traditionally, Deep Convolutional Neural Networks (DCNNs) have been employed to automate this task; however, their performance is often limited by the models' inability to focus on subtle but critical features within blood smears. This study introduces an advanced Attention-DCNN model, designed to overcome these limitations by implementing an attention mechanism that highlights informative features, enhancing model sensitivity and accuracy. The dataset comprises microscopic images from a Public Health Database, consisting of 5,000 training, 1,000 validation, and 1,500 test images, each preprocessed for normalization and resizing to ensure uniformity. Experimental results indicate a marked improvement, with the Attention-DCNN approach achieving 95% accuracy on the test set, outperforming conventional methods by a significant margin. In conclusion, the proposed Attention-DCNN framework demonstrates a promising advance in medical AI, offering a robust tool for improving malaria detection and potentially augmenting clinical workflows.