基于自动乳腺超声的肿瘤定位与三维重建系统

Tumor localization and 3D reconstruction system based on automatic breast ultrasound

目的自动乳腺超声通常无法一次性捕获完整乳腺图像,导致单张图像上的肿瘤定位困难且重复扫描的肿瘤难以确定为同一实体,针对自动乳腺超声图像带来的问题,设计一种基于自动乳腺超声的肿瘤定位与三维重建系统。方法系统主要包括3大模块:显示交互模块、三维重建模块和肿瘤映射模块。显示交互模块负责界面显示和人机交互处理;三维重建模块负责肿瘤的三维重建,综合利用图像分割技术和三维重建技术实现肿瘤模型的重建,并提出一种优化的肿瘤重建技术,依据肿瘤特性,优化重建的肿瘤模型;肿瘤映射模块提出了一种肿瘤位置映射技术,负责计算图像上发现的肿瘤在乳房中的位置,并映射到乳房模型中显示。最后利用体模和患者数据对系统功能进行检验。结果本文提出的优化肿瘤重建技术在患者数据和体模测试中均提高了肿瘤模型与标准模型的相似度。通过对优化前后的肿瘤模型的Dice相似系数(Dice similarity coefficient,DSC)进行评估,结果显示,在患者数据测试中,优化后的肿瘤模型的平均DSC提高了13.1%;在体模数据测试中,优化后的肿瘤模型的平均DSC提高了31.3%。肿瘤位置映射技术能准确映射肿瘤在乳房中的位置,并能将不同图像上的同一肿瘤映射到相近的位置,肿瘤中心的位置差异的平均值为1.22 mm,标准差为0.83 mm。结论优化的肿瘤重建技术可以提高肿瘤的三维重建效果,肿瘤位置映射技术能准确地计算肿瘤位置,融合两项技术的系统可以有效辅助医生定位肿瘤。

Objective Automatic breast ultrasound often fails to capture the entire breast image in a single scan,leading to difficulties in tumor localization on individual images and challenges in confirming repeated scans of tumors as the same entity.To address the issues associated with automatic breast ultrasound images,a tumor localization and 3D reconstruction system is designed,and the system is designed on automatic breast ultrasound.Methods The system comprises three main modules:display interaction,three-dimensional reconstruction,and tumor mapping.The display interaction module is responsible for interface display and human-computer interaction processing.The three-dimensional reconstruction module is tasked with the three-dimensional reconstruction of tumor models,utilizing image segmentation and three-dimensional reconstruction technology to rebuild tumor models,and introduces an optimized tumor reconstruction technique based on tumor characteristics to optimize the reconstructed tumor model.The tumor mapping module proposes a tumor position mapping technology,which is responsible for calculating the position of the tumor found on the image in the breast,and mapping it to the breast model for display.Finally,the phantom and patient data are used to test the system function.Results The optimized tumor reconstruction technology proposed in this paper has improved the similarity between the tumor models and the standard models in both patient data and phantom tests.Using patient data,the average dice similarity coefficient(DSC) increases by 13.1%,and using phantom data,the average DSC increases by 31.3%.The tumor location mapping technology accurately maps the position of tumors within the breast and can map the same tumor from different images to similar positions,with an average positional difference of the tumor center at 1.22 mm and a standard deviation of 0.83 mm.Conclusions The optimized tumor reconstruction technology can enhance the three-dimensional reconstruction of tumors.The tumor location mapping technology accurately calculates the position of tumors.A system that integrates both technologies can effectively assist doctors in locating tumors.