Virtual Chimera Anatomy Generation for In-Silico Trial Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional clinical trials for medical devices are limited by patient variability, leading to insufficient validation of safety and efficacy, and costly late-phase failures, necessitating a more efficient and diverse evaluation method.
Innovation Solution
A generative framework that combines disparate medical datasets with non- or partially-overlapping anatomical structures using composition neural networks to create virtual chimera populations, enabling simulation of multi-part organs and organs assemblies through graph convolutional neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional in-vivo clinical trials are used to evaluate medical devices, then regulatory approval can be obtained, but the process is time-consuming and costly, and patient variability limits the validation of safety and efficacy
Solution Approach 1:
The patent creates virtual copies of patient anatomy and physiology through computational models. These virtual patients replicate the variability and characteristics of real patients, allowing repeated simulations without time loss. The virtual population models capture anatomical variations, disease states, and response patterns, enabling comprehensive safety and efficacy evaluation parallel to real-world trials.
Solution Approach 2:
The patent divides the clinical trial evaluation process into separate modular components: virtual patient population generation, device simulation, and outcome analysis. This segmentation allows independent optimization of each component and enables parallel processing of multiple evaluation scenarios simultaneously, reducing overall time while maintaining comprehensive validation.
2Reliability
If traditional in-vivo clinical trials are used to evaluate medical devices, then regulatory approval can be obtained, but the process is costly and late-phase failures result in significant financial losses
Solution Approach 1:
The patent performs preliminary virtual testing and simulation before conducting real-world clinical trials. Virtual patients are used to pre-evaluate device performance, identify potential safety issues, and optimize trial design parameters. This preliminary action filters out problematic designs early, reducing the likelihood of costly late-phase failures while maintaining comprehensive validation through the virtual population.
Solution Approach 2:
By creating virtual copies of patient populations with diverse characteristics and disease states, the patent enables repeated virtual testing without the financial risk of real-world trial failures. The virtual population can be simulated indefinitely to explore edge cases and rare events, providing comprehensive validation at minimal cost compared to real clinical trials.
3Adaptability or versatility
If clinical trials are designed to cover patient variability, then comprehensive validation is achieved, but sufficient patients cannot be recruited to cover the full range of target populations
Solution Approach 1:
The patent generates virtual copies of patients with diverse anatomical characteristics, disease states, and demographic parameters. These virtual patients replicate the variability of real populations without recruitment constraints, enabling comprehensive coverage of target populations including rare subgroups that would be difficult to recruit in traditional trials.
Solution Approach 2:
The patent transitions from a one-dimensional real patient recruitment approach to a multi-dimensional virtual population generation approach. Virtual patients can be systematically varied across multiple dimensions (anatomy, physiology, disease severity, demographics) to cover comprehensive patient variability, exceeding the capabilities of traditional recruitment methods.
4Adaptability or versatility
If in-silico trials are used to evaluate medical devices, then diverse patient characteristics can be explored, but the approach requires complex computational modelling and simulation frameworks
Solution Approach 1:
The patent divides the complex computational modelling into separate modular components: virtual patient population generation modules, device simulation modules, and outcome analysis modules. Each module can be independently developed, validated, and optimized. This segmentation reduces the perceived complexity by allowing focused development and easier maintenance of individual components.
Solution Approach 2:
The patent introduces standardized data formats and interfaces as intermediaries between different computational modules. These intermediaries simplify the integration of diverse data types and models, reducing the complexity of connecting virtual patient generation with device simulation and outcome analysis components.
Data Source
AI summary
There is provided a computer-implemented method for generating virtual chimera populations of multi-part organ shapes for use in an in-silico trial, wherein organ shapes are represented as surface or volumetric meshes, the method comprising using a part-aware generative model to learn a latent representation of each part of a multi-part organ shape for inclusion in a virtual population and output synthesised parts of the multi-part organ shape, using a spatial composition model to align the outputted synthesised parts of the multi-part organ shape and output an anatomically meaningful example of an overall multi-part organ shape as a virtual chimera and storing the virtual chimera in the virtual population, for use in the in-silico study. There is also provided an apparatus and computer readable medium embodying the method.