Neural ODE Planning for Minimally Invasive Thermal Ablation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current thermal ablation planning for tumors larger than what can be treated with a single antenna is challenging, time-consuming, operator-dependent, and often results in suboptimal ablation margins, leading to high local recurrence rates and prolonged procedural times.
Innovation Solution
A system and method using neural ordinary differential equations (ODE) for automatically planning minimally invasive thermal ablation, determining the number and trajectories of RF ablation needles, along with ablation power and duration, based on pre-operative imaging data to ensure a standardized, repeatable, and operator-independent planning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual planning by clinicians is used, then operator experience can be applied, but the process becomes time-consuming and operator-dependent leading to suboptimal ablation margins
Solution Approach 1:
The patent replaces the manual mechanical planning process with an automated computational system using neural networks and optimization algorithms. The system automatically generates ablation plans by processing imaging data and calculating optimal needle trajectories, eliminating dependence on operator experience and manual time investment while maintaining or improving ablation margin accuracy.
Solution Approach 2:
The planning system performs self-service by automatically generating ablation plans without requiring clinician intervention in the trajectory calculation process. The system independently processes imaging data, determines optimal needle paths, and produces complete ablation plans, freeing clinicians from time-consuming manual planning while ensuring consistent quality.
2Reliability
If multiple needles are used to treat larger tumors, then complete tumor coverage can be achieved, but the spatial planning complexity increases significantly
Solution Approach 1:
The patent replaces complex manual spatial planning with automated computational optimization. The system uses neural networks and optimization algorithms to automatically determine optimal needle trajectories and positioning, transforming an intractable spatial problem into a computationally solvable task that achieves complete tumor coverage with multiple needles while eliminating planning complexity for the operator.
Solution Approach 2:
The system transitions from 2D imaging views to 3D spatial optimization by processing volumetric imaging data and calculating three-dimensional needle trajectories. This dimensional transformation enables comprehensive tumor coverage planning in three-dimensional space, allowing multiple needles to be optimally positioned to cover irregularly shaped tumors while avoiding critical structures.
3Loss of information
If pre-operative CT or MR images are used for planning, then tumor visualization is achieved, but the planning must be mentally mapped onto the patient during intervention
Solution Approach 1:
The patent replaces manual mental mapping with automated image registration and navigation systems. The computational system automatically aligns pre-operative imaging data with intraoperative anatomy, providing real-time guidance for needle placement. This eliminates the cognitive burden of mental mapping and ensures accurate spatial correspondence between planning images and actual patient anatomy during intervention.
4Measurement precision
If patients remain under general anesthesia for prolonged periods for planning, then precise positioning is maintained, but procedural risk and time increase
Solution Approach 1:
The system performs all complex planning computations before the patient undergoes anesthesia or during a minimized anesthesia period. Pre-operative imaging data is processed offline to generate complete ablation plans, allowing the patient to remain under anesthesia only long enough for needle placement according to the pre-computed trajectories, thereby minimizing anesthesia duration while maintaining positioning precision.
Solution Approach 2:
The patent replaces time-consuming manual planning during anesthesia with rapid automated computational planning performed beforehand. The system generates complete ablation plans in advance, reducing the time the patient must remain under general anesthesia from hours to minutes, while maintaining the same level of positioning precision through automated image-guided navigation.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
The present invention concerns a method (300) and a system (200) for automatic planning of a thermal ablation of a target object (105), called hereafter TO, located within a biological body (101), the method comprising: - acquiring (301) one or several images of said TO (105) within said biological body (101); - determining (302) a position of said TO (105) within the biological body (101) from said acquired images; - determining (303) a position of an external surface (102) of the biological body (101) with respect to said position of the TO (105) from said acquired images; - acquiring (305), for an initial set of N RF ablation needles (106) comprising RF ablation needles of one or several types that are usable for carrying out said thermal ablation, and for each of said types of RF ablation needles of said initial set, a set of characterizing features that are common to all RF ablation needles (106) of a same type, said set of characterizing features comprising at least one fixed parameter and/or at least one variable parameter; - feeding (306) into a neural ODE algorithm at least one of said characterizing feature, said position of the external surface (102), said position of the TO (105), wherein the neural ODE algorithm is configured for outputting at least one thermal ablation plan, each thermal ablation plan comprising a final set of said RF ablation needles (106) required for ablating the TO (105), said final set comprising N_F ≤ N RF ablation needles (106), and for each RF ablation needle (106) of said final set, its type, its trajectory (T) from said external surface (102), and optionally, a value for said variable parameter(s); - providing (307) said plan via an interface (204) configured for guiding a clinician to realize said thermal ablation of said TO (105).