A Super-Resolution Technique With An Ensemble Deep Learning Based For Cervical Cancer Detection System

A novel way for classifying cervical cells using Convolutional Neural Networks (CNN) is presented in this work. Features are learned automatically in this technique. Classification accuracy and model diversity can be enhanced using ensemble of CNN. This is a multi-step process and it requires the training of various CNN models and they must selected accordingly for ensemble. So, for resolving these issues and for creating expert medical decision making systems, Convolutional Neural Network and Random forest with Boosting called CNNRFB is used in current research work. For cervical cancer Prediction, an optimal prediction model can be constructed by medical practitioners using this solution. One dataset with all relevant information about various cervical cancer is created initially. In an image, without information loss, image pre-processing is done using super resolution technique. Self-similarity based Non-Local-Mean (NLM) metric is used as a base for this Super-Resolution (SR) algorithm. The NLM weight is used as a self-similarity metric under noisy environment for computing best self-example accurately. Deep ensemble learning methods are used for performing image classification and cancer prediction. System's accuracy are decided by dataset values and deep learning training technique. Python code is used for executing and simulating all the techniques. The experimental results shows the accuracy, recall, f-measure and precision produced by proposed technique shows better results when compared with the traditional approches.