Identifying Microscopic Augmented Images using Pre-Trained Deep Convolutional Neural Networks

Because of the success of deep learning in multiple sectors, it is gaining unquestionable acceptance in the therapeutic field also. The key challenge is determining how to construct a deep CNN model with an inadequate amount of training data. One key question is whether transfer learning and fine-tuning can be employed in biomedical image analysis to lessen the load of manual data labelling while still producing a complete deep representation for the task. In this paper, we compare the performance of transfer learning and machine learning for nuclei categorization to answer this question statistically. The used machine learning approaches are Decision Tree, Support Vector Machine, Quadratic Discriminant Analysis, K Neighbors, Ada-Boost, Gaussian Naïve Bayes, Logistic Regression, Extra Trees, Random Forest, Histogram Gradient Boosting. This paper shows how to recognize nuclei from microscope pictures using a deep learning model and an image processing-based processing flow. Convolutional neural networks and Inception Resnet V2 deep networks have been used to generate better results. The data augmentation is also done to address the paucity of data.