Data augmentation-based three-dimensional reconstruction of meningiomas: An improved K-Modified Marching Cubes algorithm

Three-dimensional reconstruction of medical images plays a crucial role in aiding diagnosis by providing comprehensive and accurate anatomical information. However, traditional Marching Cubes algorithms face challenges in generating precise models due to issues such as noise, blurring, and incomplete data in medical images. In this paper, we propose a novel method that combines Generative Adversarial Networks with Mixup data augmentation to address the problem of thick segmentation layers in three-dimensional reconstruction. Additionally, an improved Marching Cubes algorithm, referred to as K-Modified Marching Cubes, is introduced to enhance modeling capabilities for multimodal images. Experimental results demonstrate that the proposed method outperforms traditional three-dimensional reconstruction models based on original data and reduces dependency on the number of segmentation layers, offering a more accurate solution for medical image reconstruction.