Continuous Diameter Measurement of Vascular Structures in Ultrasound Video Sequences

Abstract Today, computer aided assessment of ultrasound video sequences is considered the gold standard. Ultrasound analysis forms an integral part of the medical research studies performed at the DLR. Currently, two studies are being performed that obtain ultrasound measurements of vascular structures. In this project, a software is developed that supports these and future studies in providing a framework for ultrasound analysis. Furthermore a method for obtaining assessment of diameter measurements from vascular structures in ultrasound video material is provided. Following the problem analysis based on scenarios and tasks the use cases were constructed and used for an system design. A virtual window was derived from the requirements specification—use cases. After consulting the customer with the virtual window, it was later realized by implementing the GUI. The developed software is clearly structured and focuses on extensibility. To be able to achieve this goal, the softwares main components clearly separated by integrating the model view controller pattern into the software’s architecture. The facade pattern is utilized to provide easy access to the frameworks interface models and hides the complexity. In the approach of developing the image processing, the image material was examined in depth and characteristics defined. The approach for determining the edges on the vessel walls for diameter measurement is to differentiate the image with a first order derivate of a Gaussian filter. The edges are then detected by analyzing the differentiated image. Diameter measurement between the edges follows certain assumptions that have been made. The diameter measurement algorithm was later evaluated. The evaluation of the diameter measurement algorithm showed that there is considerable variability between the gold standard of manual measurement and the software results. Furthermore, it was shown that the video quality and the definition of the region of interest have an impact on measurement accuracy. Concluding the software developed here delivers a framework for ultrasound image analysis and enables the further refinement of the measurement algorithm and flexible implementation into future ultrasound applications.Today, computer aided assessment of ultrasound video sequences is considered the gold standard. Ultrasound analysis forms an integral part of the medical research studies performed at the DLR. Currently, two studies are being performed that obtain ultrasound measurements of vascular structures. In this project, a software is developed that supports these and future studies in providing a framework for ultrasound analysis. Furthermore a method for obtaining assessment of diameter measurements from vascular structures in ultrasound video material is provided. Following the problem analysis based on scenarios and tasks the use cases were constructed and used for an system design. A virtual window was derived from the requirements specification—use cases. After consulting the customer with the virtual window, it was later realized by implementing the GUI. The developed software is clearly structured and focuses on extensibility. To be able to achieve this goal, the softwares main components clearly separated by integrating the model view controller pattern into the software’s architecture. The facade pattern is utilized to provide easy access to the frameworks interface models and hides the complexity. In the approach of developing the image processing, the image material was examined in depth and characteristics defined. The approach for determining the edges on the vessel walls for diameter measurement is to differentiate the image with a first order derivate of a Gaussian filter. The edges are then detected by analyzing the differentiated image. Diameter measurement between the edges follows certain assumptions that have been made. The diameter measurement algorithm was later evaluated. The evaluation of the diameter measurement algorithm showed that there is considerable variability between the gold standard of manual measurement and the software results. Furthermore, it was shown that the video quality and the definition of the region of interest have an impact on measurement accuracy. Concluding the software developed here delivers a framework for ultrasound image analysis and enables the further refinement of the measurement algorithm and flexible implementation into future ultrasound applications.