Investigating the Use of Hierarchical Recurrent Neural Networks for Medical Image Segmentation

Investigating the Use of Hierarchical Recurrent Neural Networks for Medical Image Segmentation

연구 분야: Artificial Intelligence

학회: International Conference on Data Science, Machine Learning and Applications

This paper investigates the usage of hierarchical recurrent neural networks (HRNNs) for clinical photo segmentation. HRNNs are a neural network structure combining more than one layer of recurrent cells and layers of convolution cells. The intention of this paper changed to evaluate the effectiveness of HRNNs for medical picture segmentation and how well it can be applied for this reason. The paper evaluates a present HRNN implementation for medical picture segmentation and shows its accuracy primarily based on the dice coefficient (DC). The authors additionally proposed a changed HRNN architecture to improve the segmentation performance: the architecture integrated residual connections, multi-scale inputs, multi-resolution outputs, and an utterly convolution structure. The proposed DT-SL loss was potent in improving the segmentation accuracy. Typically, the outcomes reveal that HRNNs are promising devices for clinical photograph segmentation tasks and can effectively facilitate faster and extra correct segmentation.