Effective Identification And Diagnosis Of Malaria Parasite In Blood Cell Images Through Deep Learning Approach

Malaria is a disease touching more than million’s of lifes in tropical regions all over the world. While benign in most cases, certain variants of malaria are known to be fatal if not treated at the right time. As such, this is a serious threat and has received attention from researchers far and wide. Due to the long, delicate, and labor-intensive process of pathological diagnosis, malaria is often erroneously diagnosed or not diagnosed early enough to prevent the ill effects. Therefore, this study based on trying to devise a method of computer aided diagnosis of malaria with high accuracy rates, using machine learning and deep learning methods. With the enormous research already dedicated to the subject, we have found several previously created models available for image based diagnosis of the same. Here the objective is to increase the accuracy of the tests while decreasing the time taken for the individual testing. In the process, also seek to verify, through replication of process, the possibility of data leak and overfitting of the models. Further methodology of quantization and sparsity checks on Deep Learning models, and transfer learning will be applied for betterment of accuracy rates to reduce the size of the model and make these models ready to apply on edge devices. Here the used methods provides more than 90 % accuracy for classifying malaria disease properly.