Advanced Learning-Based Segmentation of Liver and Tumor 3D Images for Early Disease Diagnosis

Accurate liver and tumor image segmentation is crucial for medical applications including liver functional assessment, disease diagnosis, treatment planning, and computer-aided diagnosis for liver cancer. This study introduces a novel approach for transforming 3D CT scans to segment liver tumors. We propose a novel deep learning technique utilizing convolutional neural networks (CNN) and an Unet-based architecture for liver tumor segmentation. Our suggested model achieves state-of-the-art performance on the publicly available Liver Tumor Segmen-tation (LiTS) challenge dataset, demonstrating its effectiveness in precisely segmenting liver tumors. We evaluate different models and report their performance using dice coefficient scores. The results show that the proposed model Unet-3D has the highest accuracy, with dice scores for liver and liver tumors being 0.8684 and 0.4821, respectively.