An efficient model of residual based convolutional neural network with Bayesian optimization for the classification of malarial cell images.

BACKGROUND

Malaria is a disease caused by the Plasmodium parasite, which results in millions of deaths in the human population worldwide each year. It is therefore considered a major global health issue with a massive disease burden. Accurate and rapid diagnosis of malaria is important for treatment. Rapid diagnosis of this disease will be very valuable for patients, as traditional methods require tedious work for its detection. The aim of the study is to classify malaria cell images using machine learning and deep learning methods.

MATERIAL & METHODS: The National Institutes of Health (NIH) database was used for malaria cell images classification as infected and uninfected, with a total of 27,558 malaria cell images used in the experimental study. Additionally, the training option parameters (initial learning rate, L2 regularization, and momentum values) of the Residual Convolutional Neural Network (CNN) were optimized using the Bayesian method. In the study, Residual CNN, k-Nearest Neighbors (k-NN), and Support Vector Machine (SVM) classifiers were used to classify the malaria cell images. Neighborhood Components Analysis (NCA) were observed to increase the performance of classifiers used in the classification of malaria cell images.

RESULTS AND CONCLUSION

The Accuracy (Acc), Sensitivity (Se), Specificity (Spe), and F-score were used as the performance metrics for the classifier performances. The best classification results were achieved with SVM (Acc 99.90%, Se 99.98%, Spe 87.50%, and F-Score 99.90%). As a result, a high level of classification performance was achieved from creating a hybrid model with Bayesian optimization and Deep Residual CNN features.