Mobile Phone based ensemble classification of Deep Learned Feature for Medical Image Analysis

Abstract

This research proposes a pre-trained mobile application for medical image diagnosis; it examined the benefit of deep learning approaches for white blood cell and chest radiography analysis. The feature extraction network comprised three convolutional layers using several filters with varying dimensions containing two max-pooling and batch normalization layers. The Relu layer was implemented in all the Convolutional Networks, and the learned feature output is extracted using the fully connected layers based on nodes constructed at each layer. While the Ensemble Classifier consists of a Principal Component Analysis based feature reduction, and five base learners using bagging to classify medical image datasets. The front end was designed using Unity 3D while the backend is programed using MATLAB; a comparative analysis showed the effectiveness of the proposed Convolutional Neural Network ensemble for pathological diagnoses and classification bias caused by handcrafted feature sets. The results proved that deep models could potentially change the design structure of the Computer Aided Design systems while excluding the rigorous task of development and selection of problem-oriented features.