Epithelial Tissue Classification using Pre-Trained Deep Convolutional Neural Networks

Machine learning models can analyze medical images in a more objective and complete manner since they are algorithmic and automated. Many research efforts have explored machine learning’s potential in the detection, classification, segmentation, and other difficulties associated with different illness states. To deal with the deficiency of available data for deep learning model training, a popular solution has been to use transfer learning (TL). In particular, the ImageNet pre-trained models have been one of the most prominent applications of TL. In this study, we use transfer learning to categorize several Epithelial tissues (e.g., nervous tissues, blood, etc.), each of which is made up of cells that have a common ancestry, a comparable cellular structure, and a shared function. As our pre-trained model in Transfer Learning, we have mostly concentrated on the Xception network, a widely used CNN architecture. Along with the suggested model, some more Transfer Learning architectures like VGG16, InceptionV3, and ResNet50 were also tried. The proposed Xception model achieved an accuracy of 88%, making it the most reliable option.