Knowledge integration in ultrasound-guided medical interventions

这项研究的目标是通过提供图像分析工具来改进计算机辅助医疗干预，以便从术中图像中提取有用的指导信息。超声(US)具有灵活、便携、安全和低成本的特点，成为一种很有前途的术中手段，随着自由手3D US允许使用临床上广泛使用的传统2D探头获取3D数据。利用这些数据涉及的挑战涉及同时跟踪超声换能器运动和组织变形、组织特征、分割和配准到术前患者图像。新工具将通过将来自图像本身的信息与位置跟踪设备的测量以及关于美国数据采集协议、患者解剖和组织生物力学特性的先验知识相结合来解决这些问题。一个关键的假设是，通过明确建模不同的知识来源，包括特定于应用的知识来源，并量化它们各自的可靠性，可以进行最佳的数据融合，从而产生更准确的术中指导以及更安全和更短的干预。将开发新的算法来在3D中跟踪超声换能器的位置和方向，同时补偿由于探头压力变化而引起的组织变形。通过与Sainte-Justine医院肌肉骨骼变形研究小组的合作，这些新算法将应用于微创脊柱椎间盘切除术，根据术前MRI或双平面X射线数据得出的关节3D脊柱模型必须进行更新和显示，以反映局部手术操作引起的脊柱形状的全局变化。在这里，Free Hand 3D US为里程碑式的数字化提供了一种非侵入性和实用的替代方案。由此获得的模型更新将帮助外科医生调整其策略，提高患者治疗的质量和医疗系统的成本。此外，在该研究计划内开发的方法将适用于其他类型的干预，如图像引导的针吸活组织检查和放射治疗。

The objective of this research program is to improve computer assisted medical interventions by providing image analysis tools to extract useful guidance information for the clinician from intra-operative imagery. Ultrasound (US) being flexible, portable, safe and low cost, emerges as a promising intra-operative modality, with freehand 3D US allowing the acquisition of 3D data using a conventional 2D probe widely available in the clinic. The challenges involved in exploiting these data pertain to simultaneous tracking of US transducer motion and tissue deformations, tissue characterisation, segmentation and registration to pre-operative patient imagery. The new tools will solve these problems by integrating information from the images themselves with measurements from position tracking devices as well as prior knowledge regarding the US data acquisition protocol, patient anatomy and tissue biomechanical properties. A key hypothesis is that by explicitly modelling the different sources of knowledge, including application-specific ones, and quantifying their respective reliability, optimal data fusion can be performed, leading to more accurate intra-operative guidance as well as safer and shorter interventions. New algorithms will be developed to track the position and orientation of an ultrasound transducer in 3D while compensating for tissue deformations arising from varying probe pressure. Through a collaboration with the Musculo-Skeletal Deformation Research Group at Sainte-Justine Hospital, these new algorithms will be applied to the context of minimally invasive spine discectomy, where an articulated 3D spine model derived from pre-operative MRI or bi-planar X-ray data must be updated and displayed to reflect global changes in spine shape caused by localised surgical manoeuvres. Here, freehand 3D US offers a non-invasive and practical alternative to landmark digitisation. The model updates thus obtained will assist the surgeon in adapting his strategy, improving the quality of patient treatment and cost to the healthcare system. Additionally, the methods developed within this research program will be adaptable to other types of interventions like image-guided needle biopsy and radiotherapy.