Lung motion modeling with deformable registration: non-linearity and hysteresis estimation and analysis

Abstract : Detailed internal information of lung morphology is made possible by the advent of four-dimensional computed tomography (4D-CT) and fast magnetic resonance imaging (MRI). Motion information is needed, and may be vital, in radiotherapy for lung cancer treatment, where having a very accurate modeling of the organ motion and deformation due to breathing is a challenging task. Deformable registration methods can be used to estimate deformations between 3D-CT images corresponding to different anatomical states. In this work our goal is to propose a motion model that follows lung and tumor displacements and deformations in order to extract and analyze physiological parameters like nonlinearity and hysteresis of motion during free breathing. We used a lagrangian approach tu build the motion model from vector fields computed between the reference end-of-exhale phase and all others. The motion model was evaluated for accuracy and consistency. Non-linearity and hysteresis of the lung motion were than estimated. To evaluate motion hysteresis, we introduce a measure inspired by the Frechet distance from computational geometry. Results demonstrate that lung motion during free-breathing is prone to nonlinearity and hysteresis, and that these physiological effects vary across different regions within the lung. However, for 3D-CTs with a 2.5 mm inter-slice distance, motion between end-exhale and end-inhale is well approximated with a straight-line trajectory for approximately 95% of both lungs for patients 1 and 3, and approximately 80% of both lungs for patient 2. Physiological information like nonlinearity and hysteresis of lung motion could be integrated to create a general lung atlas, with multiple clinical applications like inter-patient breathing studies, detection of specific (abnormal or not) breathing situations, and precise dosimetry study in radiotherapy.
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01588437
Contributor : Équipe Gestionnaire Des Publications Si Liris <>
Submitted on : Friday, September 15, 2017 - 4:23:39 PM
Last modification on : Thursday, November 1, 2018 - 1:21:01 AM

Identifiers

  • HAL Id : hal-01588437, version 1

Citation

Vlad Boldea, Gregory C. Sharp, Steve B. Jiang, David Sarrut. Lung motion modeling with deformable registration: non-linearity and hysteresis estimation and analysis. XVth International Conference on the Use of Computers in Radiation Therapy (ICCR), Jun 2007, Toronto, Canada. ⟨hal-01588437⟩

Share

Metrics

Record views

112