Random Walks for Image Segmentation
IEEE Transactions on Pattern Analysis and Machine Intelligence
| Hi-index | 0.00 |
We present the first system for measurement of proximal isovelocity surface area (PISA) on a 3D ultrasound acquisition using modified ultrasound hardware, volumetric image segmentation and a simple efficient workflow. Accurate measurement of the PISA in 3D flow through a valve is an emerging method for quantitatively assessing cardiac valve regurgitation and function. Current state of the art protocols for assessing regurgitant flow require laborious and time consuming user interaction with the data, where a precise execution is crucial for an accurate diagnosis. We propose a new improved 3D PISA workflow that is initialized interactively with two points, followed by fully automatic segmentation of the valve annulus and isovelocity surface area computation. Our system is first validated against several in vitro phantoms to verify the calculations of surface area, orifice area and regurgitant flow. Finally, we use our system to compare orifice area calculations obtained from in vivo patient imaging measurements to an independent measurement and then use our system to successfully classify patients into mild-moderate regurgitation and moderate-severe regurgitation categories.