Malaria is a deadly infectious disease, mainly affecting children and pregnant women in Africa, with hundreds of thousands of deaths each year. Rapid diagnosis is crucial, however, traditional methods such as microscopic examination of blood slides require resources and skilled technicians, which are often scarce in rural areas. The main challenge for doctors is decision-making, particularly due to the complexity and variability of symptoms. Therefore, computer-aided diagnosis systems assist to prioritize high-risk cases and reduce diagnostic errors. To overcome these situations, a computer-aided diagnosis system is developed to automatically identify trophozoite stages of P. falciparum Malaria as early identification species, white blood cell (WBC), and negative cells. A CNN was used as the backbone network for training the artificial intelligence algorithm model architecture to classify whether a cell is infected or not. To achieve high accuracy and recall, YoloV11m and DDQ-Detr (Deformable Dynamic Query DETR) were both combined through NMS ensemble learning techniques to detect and classify trophozoite and WBC. These algorithms were integrated into an end-to-end mobile application, specifically designed to meet diagnostic needs in low-resource settings in Africa. The results showed that the proposed model is capable of detecting the trophozoite stage of Malaria with an accuracy of 0.927 and 0.99 to classify whether a cell is infected or uninfected. Our method outperforms existing approaches in terms of speed and validation on data from diverse fields. This makes it a promising tool for the diagnosis of malaria in resource-limited environments.