Real-time liver tumor localization via a single x-ray projection using deep graph neural network-assisted biomechanical modeling

Objective. Real-time imaging is highly desirable in image-guided radiotherapy, as it provides instantaneous knowledge of patients’ anatomy and motion during treatments and enables online treatment adaptation to achieve the highest tumor targeting accuracy. Due to extremely limited acquisition time, only one or few x-ray projections can be acquired for real-time imaging, which poses a substantial challenge to localize the tumor from the scarce projections. For liver radiotherapy, such a challenge is further exacerbated by the diminished contrast between the tumor and the surrounding normal liver tissues. Here, we propose a framework combining graph neural network-based deep learning and biomechanical modeling to track liver tumor in real-time from a single onboard x-ray projection. Approach. Liver tumor tracking is achieved in two steps. First, a deep learning network is developed to predict the liver surface deformation using image features learned from the x-ray projection. Second, the intra-liver deformation is estimated through biomechanical modeling, using the liver surface deformation as the boundary condition to solve tumor motion by finite element analysis. The accuracy of the proposed framework was evaluated using a dataset of 10 patients with liver cancer. Main results. The results show accurate liver surface registration from the graph neural network-based deep learning model, which translates into accurate, fiducial-less liver tumor localization after biomechanical modeling (<1.2 (±1.2) mm average localization error). Significance. The method demonstrates its potentiality towards intra-treatment and real-time 3D liver tumor monitoring and localization. It could be applied to facilitate 4D dose accumulation, multi-leaf collimator tracking and real-time plan adaptation. The method can be adapted to other anatomical sites as well.

Fast computation of soft tissue thermal response under deformation based on fast explicit dynamics finite element algorithm for surgical simulation

BACKGROUND AND OBJECTIVES

During thermal heating surgical procedures such as electrosurgery, thermal ablative treatment and hyperthermia, soft tissue deformation due to surgical tool-tissue interaction and patient movement can affect the distribution of thermal energy induced. Soft tissue temperature must be obtained from the deformed tissue for precise delivery of thermal energy. However, the classical Pennes bio-heat transfer model can handle only the static non-moving state of tissue. In addition, in order to enable a surgeon to visualise the simulated results immediately, the solution procedure must be suitable for real-time thermal applications.

METHODS

This paper presents a formulation of bio-heat transfer under the effect of soft tissue deformation for fast or near real-time tissue temperature prediction, based on fast explicit dynamics finite element algorithm (FED-FEM) for transient heat transfer. The proposed thermal analysis under deformation is achieved by transformation of the unknown deformed tissue state to the known initial static state via a mapping function. The appropriateness and effectiveness of the proposed formulation are evaluated on a realistic virtual human liver model with blood vessels to demonstrate a clinically relevant scenario of thermal ablation of hepatic cancer.

RESULTS

For numerical accuracy, the proposed formulation can achieve a typical 10^-3 level of normalised relative error at nodes and between 10^-4 and 10^-5 level of total errors for the simulation, by comparing solutions against the commercial finite element analysis package. For computation time, the proposed formulation under tissue deformation with anisotropic temperature-dependent properties consumes 2.518 × 10^-4 ms for one element thermal loads computation, compared to 2.237 × 10^-4 ms for the formulation without deformation which is 0.89 times of the former. Comparisons with three other formulations for isotropic and temperature-independent properties are also presented.

CONCLUSIONS

Compared to conventional methods focusing on numerical accuracy, convergence and stability, the proposed formulation focuses on computational performance for fast tissue thermal analysis. Compared to the classical Pennes model that handles only the static state of tissue, the proposed formulation can achieve fast thermal analysis on deformed states of tissue and can be applied in addition to tissue deformable models for non-linear heating analysis at even large deformation of soft tissue, leading to great translational potential in dynamic tissue temperature analysis and thermal dosimetry computation for computer-integrated medical education and personalised treatment.