To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
Predictive Deep Learning Models for Malaria Using Cell Images Dataset |
|
Author |
Ahmed Abba Haruna |
| Citation |
Vol. 22 No. 9 pp. 680-686 |
|
Abstract |
Malaria, which is an endemic disease in most nations around the world, is also one of the most lethal diseases to children in Africa, particularly Nigeria, where twenty-nine countries account for 95% of malaria cases globally, with Nigeria accounting for 27%. Malaria prevention and treatment are also major difficulties in most African countries, and the disease is frequently diagnosed by health workers, particularly microbiologists, using microscopic blood smear samples. As a result, it has put a significant strain on the few medical facilities and health personnel available in most African nations, particularly Nigeria, the Democratic Republic of the Congo, Uganda, Mozambique, and Niger. Hence, artificial intelligence techniques, particularly deep learning, are increasingly widely employed for disease classification, diagnosis, and prediction. Deep learning predictive models for malaria were developed in this study utilizing a dataset of cell images. The convolutional neural network and ResNet-50 algorithms were used to create malaria models with cell image datasets, and the random rotational image augmentation technique was employed to maintain and optimize the models' performance. The predictive models' performance was evaluated, and the results show that ResNet-50 predictive models outperformed convolutional neural network predictive models in terms of being able to classify and predict infected and uninfected malaria cells with 95% accuracy, as well as correctly predicting negative cases of malaria with 95% accuracy. However, in terms of successfully predicting positive cases of malaria, the convolutional neural network predictive model surpassed the ResNet-50 predictive model with 95% accuracy. |
|
Keywords |
Malaria; Deep Learning; Convolutional Neural Network; Residual Network; ResNet-50; |
|
URL |
