利用人工智能与语音接口自动化超声医师的超声指令
Automating Sonologists USG Commands with AI and Voice Interface
AI总结：
本文提出一种AI驱动的超声成像系统，结合Mask R-CNN实时分割、器官追踪与语音控制，实现免提操作并提升诊断效率，肝脏纤维化检测准确率达98.6%。
摘要：
本研究提出了一种先进的人工智能驱动的超声成像系统，该系统融合了实时图像处理、器官追踪和语音指令，以提高临床实践中诊断的效率和准确性。传统超声诊断通常需要大量时间，并且由于用户交互而引入一定程度的主观性。这一创新解决方案的目标是利用人工智能、计算机视觉和语音技术，为超声医师提供一种更可预测、更高效的成像流程。该系统的功能采用计算机视觉和深度学习算法，具体采用Detectron2中的Mask R-CNN模型对器官和关键标志进行语义分割。这种自动化通过以最少的人工输入提取有价值的信息，提高了诊断准确性。此外，该系统还包含语音识别功能，支持免提操作，使用户能够通过诸如“冻结”或“肝脏”等指令控制系统，同时始终保持对患者的关注。该架构包括视频处理和实时分割模块，使系统能够执行基本的成像功能，例如冻结和放大帧。针对检测纤维化而优化的肝脏组织病理学模块取得了令人瞩目的98.6%的准确率。此外，器官分割模块输出的置信水平在50%到95%之间，证明了其在器官检测方面的有效性。
英文摘要：
This research presents an advanced AI-powered ultrasound imaging system that incorporates real-time image processing, organ tracking, and voice commands to enhance the efficiency and accuracy of diagnoses in clinical practice. Traditional ultrasound diagnostics often require significant time and introduce a degree of subjectivity due to user interaction. The goal of this innovative solution is to provide Sonologists with a more predictable and productive imaging procedure utilizing artificial intelligence, computer vision, and voice technology. The functionality of the system employs computer vision and deep learning algorithms, specifically adopting the Mask R-CNN model from Detectron2 for semantic segmentation of organs and key landmarks. This automation improves diagnostic accuracy by enabling the extraction of valuable information with minimal human input. Additionally, it includes a voice recognition feature that allows for hands-free operation, enabling users to control the system with commands such as freeze or liver, all while maintaining their focus on the patient. The architecture comprises video processing and real-time segmentation modules that prepare the system to perform essential imaging functions, such as freezing and zooming in on frames. The liver histopathology module, optimized for detecting fibrosis, achieved an impressive accuracy of 98.6%. Furthermore, the organ segmentation module produces output confidence levels between 50% and 95%, demonstrating its efficacy in organ detection.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
