Study Of Liver And Tumor Segmentation In CT Images Based On Deep Learning

The liver is an extremely important organ in the human body’s tissue structure,primarily responsible for functions such as digestion and metabolism.Liver tumors refer to abnormal tumor lesions on the liver organ,which pose a great harm to the human body.Computed Tomography(CT)technology can accurately identify the liver and its related diseases,providing a highly efficient detection method for clinical use,thereby providing patients with more accurate medical information and maximizing its clinical value.However,due to the complex imaging of the liver and liver tumors,it is very challenging to accurately segment the liver and liver tumors in CT images in clinical practice due to the characteristics of varying positions,large individual differences,fuzzy boundaries,and low contrast.In this paper,a liver and liver tumor CT image segmentation method is proposed based on deep learning to address the aforementioned challenges.The main contributions of this paper are as follows:First,a liver segmentation model Res Trans Unet based on CNN and Transformer fusion is proposed.The model is based on the encoding-decoding structure of U-Net,using a dual-path to extract effective features simultaneously in the encoding structure.The special design of the feature enhancement unit continuously feeds back the global features extracted by the Transformer to enhance the features extracted by CNN in each layer of the dual-path.Through ablation experiments and comparative experiments,the experimental results show that the proposed model can accurately segment the liver,and can better solve the problem of under-segmentation of boundaries compared to other models.Second,a liver tumor segmentation model RMAU-Net based on multi-scale feature fusion and attention mechanism is proposed.The model is an improvement and innovation based on U-Net,designed with a Res-SE-Block module based on residual and channel attention mechanisms and an MAB module based on multi-scale features and spatial channel attention mechanisms.Experimental results demonstrate that the RMAU-Net model has a certain accuracy improvement compared to other advanced models and can to some extent solve the difficulty of small tumor segmentation.Third,a liver and liver tumor automatic segmentation system based on Py Qt5 is designed and developed.The system’s main programming language is Python,and it is built on the Py Qt5 framework.The system embeds the image processing algorithms and two segmentation models designed in this paper,which can achieve liver segmentation and liver tumor segmentation according to needs.The system’s functions include CT image loading and display,image processing and display,liver segmentation,and liver tumor segmentation,etc.It has good operational status and stability,simple operation,and high efficiency,and can assist doctors in diagnosis and treatment in clinical practice.