Enhancing Medical Image Segmentation with Attention-Based Recurrent Neural Networks

Abstract

In recent years, deep gaining knowledge has emerged as an effective device for medical photo segmentation. This paper proposes a unique model that mixes convolutional neural networks and recurrent neural networks with an attention mechanism to improve the accuracy of segments for medical pictures, including magnetic resonance images. The eye mechanism is used to weigh each pixel, focusing the model's interest on regions of a photo that might be more applicable to classifying the item being segmented. The version is examined on medical imaging datasets - the clinical Segmentation Decathlon and the medical Segmentation Benchmark. The effects demonstrate that using the attention-based recurrent neural networks model considerably outperforms convolutional neural networks and recurrent neural networks on my own, with a median increase in dice score of up to ten%. Those effects suggest that the proposed technique can improve the accuracy of medical photo segmentation and help further facilitate the improvement of deep gaining knowledge of-based medical photograph analysis applications