An Analysis of Deep Learning Architectures for Cancer Diagnosis

It was analyzed the reference information on Deep Learning applications in the areas of diagnosis and prediction of different types of cancer. The problem is to perform the analysis and obtain the criteria to select a Deep Learning architecture for cancer diagnosis. The objective is to carry out an analysis of Deep Learning architectures and select a model to apply training and tests that assist in the diagnosis of cancer. It was used as a method the exploratory research and deduction to analyze the reference information on Deep Learning theories and architectures applied in cancer diagnosis; it also describes the reasons for selecting a model, scope, proposal, configuration parameters and structure for a CNN network. It resulted in the Impact of Deep Learning in cancer diagnosis, Training the CNN network, and Testing the CNN network. It was concluded that the 9-layer CNN Simple model used for training and testing on a data set of 8801 breast cancer images, has good properties and generates quantitative results for image classification; in the adopted model, a precision rate was obtained on the data set that reached 85.67% in training and 85.87% in tests; the quality of the model in classification tasks is 86%; this indicates the good stability and efficiency of the model.