Detecting multi-class kidney abnormalities using Deep learning

Deep learning now days is powerful tool to perform task like image classification and in the medical domain there is lack of data it's an issue to get balanced amount of quality dataset, for deep learning model it requires good amount of data for algorithms to train and to enhance their performance. Issues related to Kidney like stone, cyst and tumor are common. In this research we are focused on deep learning models that will classify kidney CT-scan images. Using state of the art classification models like CNN and Data Augmentation, normalization and Transfer learning are techniques that are used to enhance the models performance and improve the accuracy of model. The model of CNN that is being used is VGG-16 and after that custom model is developed to train it further. To balance the distribution of dataset techniques like data augmentation is being used, finally we came up with the accuracy of 98 and F1 score around 99.