Convolutional Neural Networks-Based Histopathological Image Classification

Automated histopathological image analysis is one of the prime areas of research in the field of computer vision community. Histopathology is the study of tissue images for disease detection and drug development. Although there is a sharp advancement in digital technologies, histopathological image analysis is still done manually due to the complex morphological structures of images. Recently, convolutional neural network-based image classification models are presented by researchers which have a good success in different types of images. However, these models require large image datasets for training. Therefore, this is one of the challenges in the field of histopathological images as it is a tough task to create ground truth images and requires large human efforts. Therefore, this work proposes an efficient convolutional neural network-based strategy for the classification of histopathological images in the limitation of extremely little ground truth information. For the same, various augmentation techniques have been used to increase the ground truth information. The proposed model uses different types of layers, namely convolutional layers, max-pooling layers, drop-out layers, and fully connected layers in a sequential manner. The results are compared with other traditional classifiers in terms of precision, recall, F1-score, and accuracy. From the results, it is observed that the proposed model outperforms the existing methods.