Content Based Image Retrieval For Real-Time Registration In Image-Guided Interventions

超声成像提供了重要的实时信息，以指导外科手术。术中腹腔镜超声（LUS）已广泛应用于肝脏手术，并可改善肝脏肿瘤的检测、表征和定位。然而，使用LUS可能很困难，因为外科医生在执行其职责时通常需要在一个显示器上查看腹腔镜视频，然后在另一个显示器上查看UltraSound（US）图像。此外，难以将实时LUS图像与术前图像（诸如磁共振（MR）或计算机断层扫描（CT）扫描）相关联，因为超声是二维的，而MR/CT扫描大多是三维的，并且两者对于患者的解剖结构看起来非常不同。因此，包括我们自己在内的多个作者已经开发了图像引导的腹腔镜手术系统，以提供集成的计算机系统来引导外科医生，其目的是以方便和易于使用的方式提供所有可用成像数据的直观显示。然而，尽管在过去的20年里，学术界已经开发了许多引导技术的进步，但只有很小一部分已经过渡到临床使用。面临的挑战包括缺乏空间，需要无菌，要求尽可能少的用户互动和时间限制所造成的成本剧院的时间。许多研究原型根本不符合这些标准，无法转化为临床有用的产品。通常，当前的数据配准（对齐）方法通常太慢或太笨拙，无法实时使用。我提出了一个可行性研究，以建立一个新的框架，实时注册，可应用于不同的应用，如腹腔镜手术，内窥镜胎儿手术，机器人辅助手术和图像引导消融tumors.The拟议的方法将利用图像模拟手术前。计算机将模拟并存储来自各种不同位置和方向的大量超声图像。然后在手术期间，实时超声馈送将与术前图像、模拟数据库相匹配。这将实现术前数据的实时对齐，从而提供更易于使用的系统。

Ultrasound imaging provides important real-time information to guide surgical interventions. Intra-operative Laparoscopic UltraSound (LUS) has been widely used in hepatic surgery, and may improve the detection, characterisation and localisation of hepatic tumours. However, it can be difficult to use LUS as the surgeon is normally required to look at both the laparoscopic video on one monitor and then the UltraSound (US) images on a different monitor, while performing their duties. Furthermore it is difficult to relate live LUS images to pre-operative images such as Magnetic Resonance (MR) or Computed Tomography (CT) scans, as the ultrasound is 2 dimensional, and the MR/CT scans are mostly 3 dimensional, and both look very different to the patients anatomy.So, various authors including ourselves have developed image-guided laparoscopic surgery systems to provide an integrated computer system to guide the surgeon, where the aim is to provide an intuitive display of all the available imaging data, in a convenient and easy to use fashion. However, while many advancements in guidance technology have been developed by the academic community over the last 20 years, only a very small proportion has transitioned to clinical use. Challenges include the lack of space, the need for sterility, the requirement for as little user-interaction as possible and the time constraints imposed by the cost of theatre time. Many research prototypes simply do not meet these criteria and fail to translate to a clinically useful product.Typically, current methods of registering (aligning) data are often too slow or too awkward to be used in real-time. I propose a feasibility study to build a new framework for real-time registration that can be applied to applications as diverse as laparoscopic surgery, endoscopic fetal surgery, robot-assisted surgery and image-guided ablation of tumours.The proposed method will make use of image simulation prior to surgery. The computer will simulate and store a large quantity of ultrasound images, from a variety of different positions and orientations. Then during surgery, the live ultrasound feed will be matched to the images pre-operative, simulated, database. This will enable real-time alignment of the pre-operative data, thereby providing a much easier to use system.