Automatic liver tumor segmentation of CT and MRI volumes using ensemble ResUNet-InceptionV4 model

Abstract

Liver cancer affects both men and women globally. Computed tomography (CT) imaging is a commonly employed modality for diagnosing and monitoring patients with hepatic malignancies. Segmentation of the liver and tumors through CT scans represents a challenging task. The present study proposes a novel approach for segmenting liver and tumors from CT scans, which leverages the ResUNet deep learning neural network model to achieve improved outcomes. The Inception v4 Model incorporates the ResNet and Inception v4 architectures to yield a more robust and optimized model. ResUNet is a hybrid of the well-known deep learning models ResNet and U-Net. The objective of this research is to assess the efficacy of ResUNet and Inception v4 models in comparison to conventional methods for liver segmentation. While standard approaches to liver segmentation have been used in the past, deep learning models may be more accurate. For liver segmentation, the ResUNet and Inception v4 models were intensively investigated in this research. The ResUNet and Inception v4 models outperformed traditional approaches for liver segmentation. The proposed method has been evaluated on the 3D-IRCADb-01 liver tumor segmentation benchmark and has given an accuracy of 99.27%, a mIoU of 97.74%, and a Dice score of 98.86%, showing that it segmented liver and tumors accurately from CT images. The coefficient of determination also improved significantly, indicating that the model is now ready to be applied to detect liver cancers. Based on the findings of this study, the ResUNet and Inception v4 model is a useful tool for properly segmenting liver and tumors from CT scans. The model's great precision enables early detection of liver cancers and improved patient outcomes. This study emphasizes the potential of deep learning models in medical image processing and promotes additional research in this field.