Deep Learning based Medical Image Classification

With the advancement of the Computer Technology classification of Medical Images has become more viable. Use of the traditional features has made the system to lose the ability to represent the higher-level domain problem. Deep Learning models had paved a way for the development of a generalization ability even for the poor models. Medical Images have high resolution and availability of the dataset are also small, making these Deep Learning models deteriorate from various limitations and huge computational costs. In this paper a model is proposed a profound learning model that incorporates Convolutional Neural Network (CNN), Naïve Bayes, Support Vector Machine (SVM) and Multilayer Perceptron (MLP), which consolidates high level features that are separated from a CNN model and some chosen conventional features. The development of the proposed model incorporates the accompanying advances. To start with, a CNN is trained in a supervised manner and the outcome is that it can program the raw pels of Images into include vectors that address undeniable level ideas for characterization. In the next step, a bunch of chosen customary features dependent on foundation information on medical images are extricated. At last, a proficient model that depends on Neural Networks to intertwine the diverse element bunches acquired in the previous steps is proposed. The datasets used for the evaluation of the model are HIS2828 and ISIC2017. A general classification accuracy of 97%and 96%, separately, which are higher than the current techniques are accomplished.