<![CDATA[Malaria continues to pose a major global health threat, especially in areas where timely and accurate diagnosis is essential for effective treatment. Conventional diagnostic techniques, such as manually examining Giemsa stained blood smears, are often time-intensive, laborious, and susceptible to human error. To overcome these challenges, this study presents red blood cell frame network (RBC_Frame_Net), a novel deep-learning framework that combines convolutional neural networks (CNNs) with transformer based architectures, augmented by attention mechanisms, for the automated identification of RBCs in malaria smear images. The framework leverages the convolutional block attention modules (CBAM)-UNet model for segmentation, enhancing both spatial and channel features through CBAM and integrates the detection transformer (DETR) to accurately detect and classify RBCs within the diagnostic images. The model achieved outstanding performance with a segmentation intersection over union (IoU) of 0.97, a Dice coefficient of 0.98, and near-perfect detection results (precision: 0.999, recall: 0.998, and mean average precision (mAP): 0.995). When compared to leading models such as YOLOv8, faster region-based convolutional neural network (Faster R-CNN), and EfficientDet-D3, and RBC_Frame_Net demonstrated superior accuracy and robustness. The inclusion of attention mechanisms and a hybrid architecture enhance its adaptability, making it well-suited for deployment in real-world, resource limited environments and positioning it as a valuable asset in automated malaria diagnostics.]]>
Copyrights © 2026
