Biofidelic image registration for head and neck region utilizing an in-silico articulated skeleton as a transformation model

OBJECTIVE

We propose an integration scheme for a biomechanical motion model into a deformable image registration. We demonstrate its accuracy and reproducibility for adaptive radiation therapy in the head and neck region.
Approach: The novel registration scheme for the bony structures in the head and neck regions is based on a previously developed articulated kinematic skeleton model. The realized iterative single-bone optimization process directly triggers posture changes of the articulated skeleton, exchanging the transformation model within the deformable image registration process.
Accuracy in terms of target registration errors in the bones is evaluated for 18 vector fields of three patients between each planning CT and six fraction CT scans distributed along the treatment course.
Main results: The median of target registration error distribution of the landmark pairs is 1.4 ± 0.3 mm. This is sufficient accuracy for adaptive radiation therapy. The registration performs equally well for all three patients and no degradation of the registration accuracy can be observed throughout the treatment.
Significance: Deformable image registration, despite its known residual uncertainties, is until now the tool of choice towards online re-planning automation. By introducing a biofidelic motion model into the optimization, we provide a viable way towards an in-build quality assurance.&#xD.

Tissue-specific transformation model for CT-images

During radiotherapy, posture changes and volume changing deformations like growing or shrinking tissue result in anatomical deformations. The basis for investigating the impact of such deformations on dose uncertainties, are model-based tools for deformation analysis. In this context, we propose a transformation model based on the information of CT-images, which allows an on-the-fly calculation of voxel volumes. Our model is based on the concept of the chainmail algorithm and describes deformation on voxel-level. With an exemplary input of a set of landmark pairs, generated by a kinematic head-and-neck skeleton model, CT-images (512x512x126 voxel) can be deformed with an on-the-fly volume calculation in less than 70s. The volume calculation delivers insight into model-characteristic volume changes and is a prerequisite for implementing tissue growth and shrinkage.

Handling images of patient postures in arms up and arms down position using a biomechanical skeleton model

Deformable image registration is gradually becoming the tool of choice for motion extraction during adaptive radiotherapy. Achieving a motion vector field that accurately represents the anatomical changes requires a tissue specific transformation model. Therefore, widely used spline based models most likely fail in appropriately reproducing large anatomical changes such as the arms of the patient being positioned up and down. We present the application of a tissue specific biomechanical model with the goal to mimic patient motion even in presence of large motion. Based on the planning CT, delineated bones are used to represent the rigid anatomy of the patient. We implement ball-and-socket joints between corresponding bones in order to achieve mobility of the skeleton. An inverse kinematics approach enables the propagation of motion between individual bones across their joints, leading to an articulated skeleton that can be controlled by feature points on one or more bones. The transformation of each bone initializes a chainmail based soft tissue model to also propagate the motion into the surrounding heterogeneous soft tissue. Representation of different postures like arms up and down can be achieved within less than 1 s for the skeleton and ∼10 s for the soft tissue. Especially for large anatomical changes, the kinematics approach benefits from the direct articulation at specific joints, considerably lowering the degrees of freedom for motion description. Being the input for the chainmail based soft tissue model, the transformed bones guarantee for its meaningful initialization. The proposed biomechanical skeleton model is promising to facilitate the registration of patients’ anatomy, being positioned with arms up and arms down. The results encourage further refinement of the joints and the soft tissue model.

Analyzing human decisions in IGRT of head-and-neck cancer patients to teach image registration algorithms what experts know

Background

In IGRT of deformable head-and-neck anatomy, patient setup corrections are derived by rigid registration methods. In practice, experienced radiation therapists often correct the resulting vectors, thus indicating a different prioritization of alignment of local structures. Purpose of this study is to transfer the knowledge experts apply when correcting the automatically generated result (pre-match) to automated registration.

Methods

Datasets of 25 head-and-neck-cancer patients with daily CBCTs and corresponding approved setup correction vectors were analyzed. Local similarity measures were evaluated to identify the criteria for human corrections with regard to alignment quality, analogous to the radiomics approach. Clustering of similarity improvement patterns is applied to reveal priorities in the alignment quality.

Results

The radiation therapists prioritized to align the spinal cord closest to the high-dose area. Both target volumes followed with second and third highest priority. The bony pre-match influenced the human correction along the crania-caudal axis. Based on the extracted priorities, a new rigid registration procedure is constructed which is capable of reproducing the corrections of experts.

Conclusions

The proposed approach extracts knowledge of experts performing IGRT corrections to enable new rigid registration methods that are capable of mimicking human decisions. In the future, the deduction of knowledge-based corrections for different cohorts can be established automating such supervised learning approaches.

Construction of a biomechanical head and neck motion model as a guide to evaluation of deformable image registration

The use of deformable image registration methods in the context of adaptive radiotherapy leads to uncertainties in the simulation of the administered dose distributions during the treatment course. Evaluation of these methods is a prerequisite to decide if a plan adaptation will improve the individual treatment. Current approaches using manual references limit the validity of evaluation, especially for low-contrast regions. In particular, for the head and neck region, the highly flexible anatomy and low soft tissue contrast in control images pose a challenge to image registration and its evaluation. Biomechanical models promise to overcome this issue by providing anthropomorphic motion modelling of the patient. We introduce a novel biomechanical motion model for the generation and sampling of different postures of the head and neck anatomy. Motion propagation behaviour of the individual bones is defined by an underlying kinematic model. This model interconnects the bones by joints and thus is capable of providing a wide range of motion. Triggered by the motion of the individual bones, soft tissue deformation is described by an extended heterogeneous tissue model based on the chainmail approach. This extension, for the first time, allows the propagation of decaying rotations within soft tissue without the necessity for explicit tissue segmentation. Overall motion simulation and sampling of deformed CT scans including a basic noise model is achieved within 30 s. The proposed biomechanical motion model for the head and neck site generates displacement vector fields on a voxel basis, approximating arbitrary anthropomorphic postures of the patient. It was developed with the intention of providing input data for the evaluation of deformable image registration.

Real-time markerless lung tumor tracking in fluoroscopic video: Handling overlapping of projected structures

PURPOSE

Fluoroscopic imaging is a well-suited technique for online visualization of tumor motion in the thoracic region. Template-based approaches for tumor tracking in such images are commonly used. However, overlapping of different structures, mainly bones, can lead to limited visibility of the projected tumor shape, which in turn can negatively affect the performance of the tracking method. In this study, a method based on multiple-template matching was developed, providing fast and robust detection of tumor motion even under the influence of occurring tumor overlaps.

METHODS

A cohort of 14 patients with varying tumor sizes and locations was investigated. Image data from eight of these patients were used for evaluation. Based on the requirement of tumor visibility, the remaining datasets did not qualify for tracking. Generation of multiple templates was improved by implementation of an algorithm for automated selection of reference images containing the most characteristic tumor appearances. Various measures were taken to ensure real-time capability of the algorithm. A prematching step was introduced in order to reduce dispensable comparison operations by selecting the most appropriate template. Subsequent matching was further optimized by using prior knowledge about likely tumor motion to effectively limit necessary matching tasks.

RESULTS

Tracking accuracy of the developed multiple-template method was compared with that of single-template. Mean errors of the multiple-template approach were 0.6 ± 0.6 mm in left-right and 0.9 ± 0.9 mm in superior-inferior direction in the isocenter plane. The single-template approach achieved mean errors of 0.7 ± 0.7 mm in left-right and 1.5 ± 1.3 mm in superior-inferior direction. These results derive from evaluation against manual tumor tracking performed by four expert observers. Computational times needed for tumor detection in a single fluoroscopic frame ranged between 1 and 29 ms depending on the tumor size and motion amplitude.

CONCLUSIONS

This study shows that in case of tumor overlapping with dense structures, multiple-template tracking provides more accurate results than a single-template approach. The developed algorithm shows promising results in terms of suitability for real-time application and robustness against frequently changing overlapping.