Nuclei cells detection and segmentation with deep neural network

Deep learning allows computational models consisting of multiple processing layers to learn representations of data with multiple levels of abstraction. These methods have significantly advanced cutting-edge technology in speech recognition, visual object recognition, object detection and many other fields such as drug discovery and genomics. One of the most popular deep neural networks is convolutional neural networks. In recent years, semantic segmentation based on convolutional neural networks has shown significant advantages over traditional image segmentation and has been successfully applied to many visual tasks. Semantic segmentation is the process of assigning a label to each pixel in the image. This is the opposite of classification, where a single tag is assigned to the entire image. Semantic segmentation treats multiple objects of the same class as a single entity. In this thesis, semantic segmentation method is used for core image segmentation and detection. Semantic segmentation based on convolutional neural networks has shown outstanding performance for kernel image segmentation and detection. Publicly available breast cancer images from the PSB 2015 crowdsourced core dataset and the Kaggle 2018 Data Science Bowl dataset were used to test the method. The complexity matrix, which calculates true negative, true positive, false positive and false negative, was used for accuracy assessment. Also; Sensitivity, Specificity, Jaccard Index, Dice, Precision, Recall, and F1 Score performance scales were also used. As a result of the experiments; the average values of Sensitivity, Specificity, Accuracy, Jaccard, Dice, Precision and F1 Score were obtained as follows, respectively: 0.9269; 0.7160; 0.851; 0.4855; 0.6434; 0.844 and 0.6434. The output results for the 2018 Data Science Bowl dataset were compared with the exact reference images, and the average values of Sensitivity, Specificity, Accuracy, Jaccard, Dice, Precision, and F1 Score, respectively, were obtained as follows: 0.8719; 0.9881; 0.9768; 0.7715; 0.8632; 0.8843 and 0.8632 values have been obtained. As a result, it has been shown to have higher performance than other methods suggested in the literature.