Detecting Cancerous Cells Using Data Augmentation In Deep Cascaded Networks

In this article, an approach has been introduced for detecting cancerous cells. Image processing techniques have been used, based on cancer cell area using CNNs. A very intriguing aspect of this experiment was that from a very small image dataset, a large number of images were generated using information augmentation which was then taken as the training set data. The suggested scheme detects cancer behaviors through a convolutional neural network in images of celled samples. Previously, the same attempts failed to stay away from the database dependencies, which were somewhat proportional to the number of images in datasets, so we used a method called data augmentation on smaller sets of images. The scheme preprocesses the input image by grayscale, binarization, inversion, median filtering, and flood-fill procedures. Depending on the sort of feature to be identified, the preprocessed image is then cancerous cell detected. This methodology was used for several sets of pictures, and the systemwas optimized with the feedback from those tests. For independent cancer cell detection with narrower datasets, the suggested technique can be efficiently used, which will greatly accelerate the study of cancer and open greater dimensions.