Generative AI for Heterogeneous Radiotherapy Dose Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic dose prediction models in radiotherapy are limited by their focus on single modalities and beam configurations, restricting their versatility and ability to adapt to various clinical scenarios, leading to increased time and computing resource demands.
Innovation Solution
A conditional generative AI model is trained using a dataset of clinically approved plans and patient data, enabling it to predict three-dimensional dose maps for new patients across different radiotherapy modalities and beam geometries, including hybrid approaches, thus improving adaptability and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If algorithmic methods are used to calculate predicted dose distribution, then accuracy of treatment planning is improved, but time and computing resources are increased
Solution Approach 1:
The system performs preliminary action by pre-calculating and storing dose distribution data from treatment plans in a training dataset before actual treatment planning. The AI model is trained in advance on this pre-prepared data, enabling rapid predictions during actual treatment planning without performing complex algorithmic calculations in real-time, thus reducing planning time while maintaining accuracy
Solution Approach 2:
The system creates a copy of the complex algorithmic dose calculation process by training an AI model on pre-calculated dose distribution data. The trained model replicates the dosimetry calculations through pattern recognition rather than direct computation, providing accurate dose predictions without requiring the original complex algorithmic processing during treatment planning
2Adaptability or versatility
If existing automatic dose prediction models focus on single modality and beam configuration, then model simplicity is maintained, but versatility and adaptability are restricted
Solution Approach 1:
The system implements universality by training a single AI model on a diverse training dataset that includes multiple treatment modalities (IMRT, VMAT, hybrid) and various beam geometry configurations. This enables the model to function universally across different radiotherapy scenarios, adapting to various modalities and beam arrangements without requiring separate specialized models for each configuration
Solution Approach 2:
The system applies parameter changes by incorporating modality indicators and beam geometry parameters as input conditions to the AI model. The model receives these parameter variations and adjusts its predictions accordingly, allowing it to handle different treatment modalities and beam configurations through parameter-based adaptation rather than structural complexity
3Measurement precision
If clinically approved plans with diverse modalities are used for training, then model accuracy across different scenarios is improved, but training data complexity is increased
Solution Approach 1:
The system applies segmentation by organizing the diverse training data into distinct categories with specific labels indicating treatment modality (IMRT, VMAT, hybrid) and beam geometry characteristics. This structured segmentation allows the AI model to learn patterns specific to each category while maintaining the ability to generalize across all modalities, improving accuracy without being overwhelmed by data complexity
Data Source
AI summary
Provided herein are methods and systems for training and executing an AI model to generate a predicted dose map. In an example, a method comprises generating a training dataset comprising at least a medical image, structure mask(s) for structure(s) within the medical image, an indication of radiotherapy treatment modality, and beam geometry associated with a set of patients, wherein the training dataset comprises data associated with treatments implemented using at least one of volumetric modulated arc therapy or intensity-modulated radiotherapy modalities; training an artificial intelligence model using the training dataset, such that the artificial intelligence model is configured to receive data associated with a new patient and generate a predicted dose map for the new patient indicating dosage received by one or more internal structures of the new patient; executing the artificial intelligence model for the new patient to receive the predicted dose map for the new patient; and outputting the result.