Transformation of 1-D data to 2-D image using stochastic mapping method for secured skin lesion detection

Abstract Mostly accurate medical diagnosis is performed from image data. Some of the complex diseases and their data are available from different gene structures. For detection purposes, the image is the accurate and secured form for analysis. In this work, the gene expressions are converted to 2-D images for further processing. The images are detected using a convolutional neural network. To convert the gene data into image data, encoding mapping through t-Distributed Stochastic Neighbor, Multi-dimensional scaling, and Locally linearly embedding is used. Further, the application of the Convex Hull algorithm forms the fixed boundary. The data set is collected from the Gene expression Omnibus of the NCBI platform. The CNN model provided 96.32% and 92.13% training and testing accuracies that show the effectiveness of data conversion in the field of skin cancer detection.