Abstract

Segmentation of injured or unusual anatomic structures in medical imagery is a problem that has continued to elude fully automated solutions. In this paper, the goal of easy-to-use and consistent interactive segmentation is transformed into a control synthesis problem. A nominal level set PDE is assumed to be given; this open-loop system achieves correct segmentation under ideal conditions, but does not agree with a human expert's ideal boundary for real image data. Perturbing the state and dynamics of a level set PDE via the accumulated user input and an observer-like system leads to desirable closed-loop behavior. The input structure is designed such that a user can stabilize the boundary in some desired state without needing to understand any mathematical parameters. Effectiveness of the technique is illustrated with applications to the challenging segmentations of a patellar tendon in MR and a shattered femur in CT.

Highlights

  • Microwave-frequency Magnetic-Resonance-Imaging (MRI) [1] and X-Ray Computed Tomography (CT) [2] yield three-dimensional volumetric images which are viewed by a medical professional for diagnosis, treatment planning, or population studies [3], [4]

  • We present a framework for interactive segmentation using feedback augmentation of a level set partial differential equation (PDE) system; the results and theory are a substantial extension of the preliminary version [26]

  • This paper has presented a modeling approach that enables control-theoretic analysis and design for interactive medical image segmentation

Read more

Summary

Introduction

Microwave-frequency Magnetic-Resonance-Imaging (MRI) [1] and X-Ray Computed Tomography (CT) [2] yield three-dimensional volumetric images which are viewed by a medical professional for diagnosis, treatment planning, or population studies [3], [4]. An alternative approach is to formulate segmentation as a time-independent problem; in [21]– [23], user input acts as a constraint in finite-dimensional nonlinear optimization problems It is not known a-priori whether the user's ideal region boundary is a feasible solution for some collection of constraints, while a changing number of user input constraints can affect the computational complexity. This paper is motivated by the following observation: when human users influence the level-set evolution, they have in mind a desired reference state and are trying to apply control to an image-dependent PDE system. To the best of our knowledge, this is the first approach to interactive level set segmentation with input from the user used in feedback to guarantee stabilization about a reference boundary. Key components of the final closed-loop system and corresponding paper sections are visualized in Fig-2

Level Sets and Automated Segmentation
Review of Level Set Methods
Segmentation as an Open-Loop System
Feedback Augmentation of a Narrowband Levelset PDE
Auxiliary System Design
User Input Processing
Application to MRI and CT Images
CT Segmentation with Mean-Alignment
MRI Segmentation with Localized Statistics
Overview
Quantitative Comparison with GrabCut
Findings
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call