In order to solve the training difficulty of small sample data in deep learning and increase the training efficiency of DCGAN (Deep Convolutional Generative Adversarial Network), an improved DCGAN algorithm was proposed to perform the augmentation of small sample data. In the method, Wasserstein distance was used to replace the loss model in the original model at first. Then, spectral normalization was added in the generation network, and discrimination network to acquire a stable network structure. Finally, the optimal noise input dimension of sample was obtained by the maximum likelihood estimation and experimental estimation, so that the generated samples became more diversified. Experimental result on three datasets MNIST, CelebA and Cartoon indicated that the improved DCGAN could generate samples with higher definition and recognition rate compared to that before improvement. In particular, the average recognition rate on these datasets were improved by 8.1%, 16.4% and 16.7% respectively, and several definition evaluation indices on the datasets were increased with different degrees, suggesting that the method can realize the small sample data augmentation effectively.

When using AlexNet to implement image classification of gastric tumor cells, there are problems of small dataset, slow model convergence and low recognition rate. Aiming at the above problems, a Data Augmentation (DA) method based on Radial Transformation (RT) and improved AlexNet was proposed. The original dataset was divided into test set and training set. In the test set, cropping was used to increase the data. In the training set, cropping, rotation, flipping and brightness conversion were employed to obtain the enhanced image set, and then some of them were selected for RT processing to achieve the enhanced effect. In addition, the replacement activation of functions and normalization layers was used to speed up the convergence and improve the generalization performance of AlexNet. Experimental results show that the proposed method can implement the recognition of gastric tumor cell images with faster convergence and higher recognition accuracy. On the test set, the highest accuracy is 99.50% and the average accuracy is 96.69%, and the F1 scores of categories:canceration, normal and hyperplasia are 0.980, 0.954 and 0.958 respectively, indicating that the proposed method can implement the recognition of gastric tumor cell images well.

Aiming at the characteristics of high dimension and complexity of gastric mucosal tumor cell images, a new method based on Fisher Discrimination Dictionary Learning and Regularized Robust Coding (FDDL-RRC) was proposed for the recognition of tumor cell images, so as to improve the robustness of sparse representation for image recognition. Firstly, all the original stained tumor cell images were transformed into gray images, and then the Fisher discrimination dictionary learning method was used to learn the global features of training samples and obtain the structured dictionary with class labels; lastly, the new discriminative dictionary was used to classify the test samples by the model of RRC. The model of RRC was based on Maximum A Posterior (MAP) estimation, and the sparse fidelity was expressed by the MAP function of residuals, so the problem of identification was converted to the optimal regularized weighted norm approximation problem. The highest recognition accuracy rate of the proposed method for tumor cell images can reach 92.4%, which indicates that the presented method can effectively and quickly distinguish the tumor cell images.

The method for iterative probability relaxation segmentation used in cell division can overcome the difficult issues on account of complicated cellular structure and phenomenon of serious adhesion, while the general segmentation algorithm cannot make it effectively. In addition, because of tense embedded resources under the environment of Linux system, the iterative relaxation cellular segmentation algorithm has been improved and then added to the embedded cellular segmentation system based on Qt and OpenCV. The experimental results demonstrate that the improved algorithm can effectively solve the difficult problem of cell division efficaciously and the naked eye can clearly distinguish the difference between the nucleus, cytoplasm and glands. The improved algorithm increases the processing speed and can be transplanted to the embedded facilities convenient for carrying, diagnosing and treating.