A Deep Learning Architecture to Enhance Cancer Diagnosis Using Convolutional Neural Network and Discrete Cosine Transform

Deep learning architectures are formed by the composition of multiple linear and non-linear transformations of the data, with the goal of yielding more abstract and ultimately more useful representations. The concept of deep learning is inbuilt with many algorithms which help to train the input data set into different layer of perceptron. The convolutional neural networks, which is one of the specialized deep learning algorithms, is used efficiently for analyzing medical images. This paper presents an architecture for efficient and improved clinical diagnosis in scanned image without losing any information produced by the original image using Discrete Cosine Transform (DCT) and classify the features using Neural Network. This transformation has an advantage of compressing the data sets and expresses the data set in a finite sequence of arithmetic cosine functions. A novel characteristic of this approach is that it extends the deep learning architecture to include an interpretable layer that highlights the visual patterns thatdiscriminate between cancerous and normal tissues patterns, working similar to a digital staining which spotlights image regions important for diagnostic decisions. The deep convolutional network, which is one of the deep learning techniques, is used for the purpose of medical image analysis. Segmentation, abnormality detection, disease classification, computer aided diagnosis and retrieval are some of the application areas of convolutional neural networks. In this paper, a comprehensive review of the current state-of-the-art in medical image analysis using deep convolutional networks is also presented. The challenges and potential of these techniques are also highlighted. When compared to the previous cancer detection approaches, the main advantage of the proposed architecture is the possibility of applying data from various types of cancer to automatically form the characteristics which facilitate to enhance the detection