HYPER PARAMETER TUNING OF PRE-TRAINED DEEP LEARNING MODEL FOR AN EFFICIENT MEDICAL IMAGE CLASSIFICATION USING CNN

Recently, constructing an efficient CNN architecture has been an active research area. The main objective of this work is to implement tensorflow tool with classification techniques for detecting tuberculosis in the chest images with maximum accuracy.The purpose of the work is to execute a pre-trained Neural Network by means oftransfer Learning. It can be done by re-routing the output of the pre-trained model just prior to its novel classification layers and instead usingthe old classifier. Therefore, the pre trained model can be fine-tuned, to diagnose modalities and immodalities in X-Ray images by implementing fine-tuning pre-trained VGG16 model with estimators and activation functions in hidden layers of CNN architecture. As a result, it is showed that fine tuning pre-trained model can able to obtain an efficient classification with improved performance based on increased accuracy rate, so that it also provides the possibility of automatic diagnosis system. The automatic diagnosis of diseases from the input images are high desired in the area of medical image processing. The proposed work helps to identify the best activation function for classifying the diseases using deep learning.