Fusion of Inception Net and Efficient Net for Enhanced Malaria Parasite Detection

Abstract

Accurately identifying malaria remains one of the most challenging medical endeavors globally. We present a hybrid neural network approach that integrates Inception V3 with EfficientNetB0 to address this challenge. Our hybrid architecture effectively distinguishes between parasitized and uninfected blood smear images, achieving a commendable accuracy of 94.27% through these two cutting-edge models' advanced feature extraction capabilities. This strategy yields a lightweight and highly accurate diagnostic tool by leveraging Inception's multi-scale feature extraction and EfficientNet's optimized performance. The innovative architecture of our model illustrates the potential for integrating complementary neural networks to enhance the accuracy of medical imaging diagnoses through the amalgamation of pooled data from both networks. This fusion-based technique may enhance malaria screening, which is particularly beneficial for healthcare practitioners in resource-limited regions requiring rapid and precise diagnoses.

Keywords: Multi-Model Ensemble; Healthcare AI; Convolutional Neural Network; Parasite Classification; Malaria Detection;