Robust Path Planning via Learning From Demonstrations for Robotic Catheters in Deformable Environments

OBJECTIVE

Navigation through tortuous and deformable vessels using catheters with limited steering capability underscores the need for reliable path planning. State-of-the-art path planners do not fully account for the deformable nature of the environment.

METHODS

This work proposes a robust path planner via a learning from demonstrations method, named Curriculum Generative Adversarial Imitation Learning (C-GAIL). This path planning framework takes into account the interaction between steerable catheters and vessel walls and the deformable property of vessels.

RESULTS

In-silico comparative experiments show that the proposed network achieves a 38% higher success rate in static environments and 17% higher in dynamic environments compared to a state-of-the-art approach based on GAIL. In-vitro validation experiments indicate that the path generated by the proposed C-GAIL path planner achieves a targeting error of 1.26 0.55 mm and a tracking error of 5.18 3.48 mm. These results represent improvements of 41% and 40% over the conventional centerline-following technique for targeting error and tracking error, respectively.

CONCLUSION

The proposed C-GAIL path planner outperforms the state-of-the-art GAIL approach. The in-vitro validation experiments demonstrate that the path generated by the proposed C-GAIL path planner aligns better with the actual steering capability of the pneumatic artificial muscle-driven catheter utilized in this study. Therefore, the proposed approach can provide enhanced support to the user in navigating the catheter towards the target with greater accuracy, effectively meeting clinical accuracy requirements.

SIGNIFICANCE

The proposed path planning framework exhibits superior performance in managing uncertainty associated with vessel deformation, thereby resulting in lower tracking errors.

Toward certifiable optimal motion planning for medical steerable needles

Medical steerable needles can follow 3D curvilinear trajectories to avoid anatomical obstacles and reach clinically significant targets inside the human body. Automating steerable needle procedures can enable physicians and patients to harness the full potential of steerable needles by maximally leveraging their steerability to safely and accurately reach targets for medical procedures such as biopsies. For the automation of medical procedures to be clinically accepted, it is critical from a patient care, safety, and regulatory perspective to certify the correctness and effectiveness of the planning algorithms involved in procedure automation. In this paper, we take an important step toward creating a certifiable optimal planner for steerable needles. We present an efficient, resolution-complete motion planner for steerable needles based on a novel adaptation of multi-resolution planning. This is the first motion planner for steerable needles that guarantees to compute in finite time an obstacle-avoiding plan (or notify the user that no such plan exists), under clinically appropriate assumptions. Based on this planner, we then develop the first resolution-optimal motion planner for steerable needles that further provides theoretical guarantees on the quality of the computed motion plan, that is, global optimality, in finite time. Compared to state-of-the-art steerable needle motion planners, we demonstrate with clinically realistic simulations that our planners not only provide theoretical guarantees but also have higher success rates, have lower computation times, and result in higher quality plans.

A Heuristically Accelerated Reinforcement Learning-Based Neurosurgical Path Planner

The steerable needle becomes appealing in the neurosurgery intervention procedure because of its flexibility to bypass critical regions inside the brain; with proper path planning, it can also minimize the potential damage by setting constraints and optimizing the insertion path. Recently, reinforcement learning (RL)-based path planning algorithm has shown promising results in neurosurgery, but because of the trial and error mechanism, it can be computationally expensive and insecure with low training efficiency. In this paper, we propose a heuristically accelerated deep Q network (DQN) algorithm to safely preoperatively plan a needle insertion path in a neurosurgical environment. Furthermore, a fuzzy inference system is integrated into the framework as a balance of the heuristic policy and the RL algorithm. Simulations are conducted to test the proposed method in comparison to the traditional greedy heuristic searching algorithm and DQN algorithms. Tests showed promising results of our algorithm in saving over 50 training episodes, calculating path lengths of 0.35 after normalization, which is 0.61 and 0.39 for DQN and traditional greedy heuristic searching algorithm, respectively. Moreover, the maximum curvature during planning is reduced to 0.046 from 0.139 mm-1 using the proposed algorithm compared to DQN.

Path Planning of Mobile Robots Based on an Improved Particle Swarm Optimization Algorithm

Aiming at disadvantages of particle swarm optimization in the path planning of mobile robots, such as low convergence accuracy and easy maturity, this paper proposes an improved particle swarm optimization algorithm based on differential evolution. First, the concept of corporate governance is introduced, adding adaptive adjustment weights and acceleration coefficients to improve the traditional particle swarm optimization and increase the algorithm convergence speed. Then, in order to improve the performance of the differential evolution algorithm, the size of the mutation is controlled by adding adaptive parameters. Moreover, a “high-intensity training” mode is developed to use the improved differential evolution algorithm to intensively train the global optimal position of the particle swarm optimization, which can improve the search precision of the algorithm. Finally, the mathematical model for robot path planning is devised as a two-objective optimization with two indices, i.e., the path length and the degree of danger to optimize the path planning. The proposed algorithm is applied to different experiments for path planning simulation tests. The results demonstrate the feasibility and effectiveness of it in solving a mobile robot path-planning problem.

A hybrid inductive learning-based and deductive reasoning-based 3-D path planning method in complex environments

Traditional path planning methods, such as sampling-based and iterative approaches, allow for optimal path’s computation in complex environments. Nonetheless, environment exploration is subject to rules which can be obtained by domain experts and could be used for improving the search. The present work aims at integrating inductive techniques that generate path candidates with deductive techniques that choose the preferred ones. In particular, an inductive learning model is trained with expert demonstrations and with rules translated into a reward function, while logic programming is used to choose the starting point according to some domain expert’s suggestions. We discuss, as use case, 3-D path planning for neurosurgical steerable needles. Results show that the proposed method computes optimal paths in terms of obstacle clearance and kinematic constraints compliance, and is able to outperform state-of-the-art approaches in terms of safety distance-from-obstacles respect, smoothness, and computational time.

Insights into Infusion-Based Targeted Drug Delivery in the Brain: Perspectives, Challenges and Opportunities

Targeted drug delivery in the brain is instrumental in the treatment of lethal brain diseases, such as glioblastoma multiforme, the most aggressive primary central nervous system tumour in adults. Infusion-based drug delivery techniques, which directly administer to the tissue for local treatment, as in convection-enhanced delivery (CED), provide an important opportunity; however, poor understanding of the pressure-driven drug transport mechanisms in the brain has hindered its ultimate success in clinical applications. In this review, we focus on the biomechanical and biochemical aspects of infusion-based targeted drug delivery in the brain and look into the underlying molecular level mechanisms. We discuss recent advances and challenges in the complementary field of medical robotics and its use in targeted drug delivery in the brain. A critical overview of current research in these areas and their clinical implications is provided. This review delivers new ideas and perspectives for further studies of targeted drug delivery in the brain.

Inverse Reinforcement Learning Intra-operative Path Planning for Steerable Needle

OBJECTIVE

This paper presents a safe and effective keyhole neurosurgery intra-operative planning framework for flexible neurosurgical robots. The framework is intended to support neurosurgeons during the intraoperative procedure to react to a dynamic environment.

METHODS

The proposed system integrates inverse reinforcement learning path planning algorithm combined with 1) a pre-operative path planning framework for fast and intuitive user interaction, 2) a realistic, time-bounded simulator based on Position-based Dynamics (PBD) simulation that mocks brain deformations due to catheter insertion and 3) a simulated robotic system.

RESULTS

Simulation results performed on a human brain dataset show that the inverse reinforcement learning intra-operative planning method can guide a steerable needle with bounded curvature to a predefined target pose with an average targeting error of 1.34 0.52 (25th=1.02, 75th=1.36) mm in position and 3.16 1.06 (25th=2, 75th=4.94) degrees in orientation under a deformable simulated environment, with a re-planning time of 0.02 sec and a success rate of 100%.

CONCLUSION

With this work, we demonstrate that the presented intra-operative steerable needle path planner is able to avoid anatomical obstacles while optimising surgical criteria.

SIGNIFICANCE

The results demonstrate that the proposed method is fast and can securely steer flexible needles with high accuracy and robustness.