DVH Prediction Model for Radiation Therapy Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting Dose Volume Histograms (DVH) in radiation therapy treatment planning rely on geometric data, which is challenging to modify to accurately predict achievable dose metrics for new patients, especially in complex treatments like simultaneous integrated boost (SIB) and multimodality therapies.
Innovation Solution
The method involves using protocol DVH curves based on dose distribution data from previous patients, trained through a prediction model that maps protocol DVH curves to estimated clinical DVH curves, focusing on realistic dose predictions rather than geometric similarities, and is applicable to SIB and multimodality treatments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geometric data (DTH, OVH) is used to predict DVH curves, then the prediction model can be trained on previous patient data, but the prediction accuracy is insufficient especially for complex treatments like SIB and multimodality therapies
Solution Approach 1:
The patent changes the fundamental parameter basis from geometric data (distance to target histograms, overlap volume histograms) to dose distribution data (protocol DVH curves). This parameter transformation enables the prediction model to accurately capture the complex dose dependencies in SIB and multimodality treatments, where geometric similarity does not guarantee dose similarity. The protocol DVH curves encode the actual dose delivery characteristics including modality-specific effects, making the predictions both accurate and universally applicable across different treatment types.
2Reliability
If manual treatment planning is performed, then the operator can adjust the plan based on experience, but the process is time-consuming and introduces human error and inconsistencies
Solution Approach 1:
The patent implements preliminary action by pre-calculating protocol DVH curves using standardized plan generation protocols applied to training data from previous patients. These pre-computed protocol DVH curves serve as the foundation for training the prediction model, enabling rapid predictions for new patients without requiring time-consuming manual planning. The system performs the computationally intensive work in advance, storing the results for efficient retrieval and application during actual treatment planning.
Solution Approach 2:
The patent uses copying by creating a prediction model that replicates the dose delivery characteristics of successful treatment plans from previous patients. Instead of manually recreating plans, the system copies the essential dose distribution patterns encoded in the protocol DVH curves and applies them to new patients with similar anatomical and treatment characteristics, ensuring consistency while eliminating manual variability.
3Measurement precision
If protocol DVH curves based on dose distribution data are used, then the prediction reflects actual dose dependencies, but the calculation and training process becomes more complex
Solution Approach 1:
The patent applies universality by designing a plan generation protocol that can generate protocol plans and calculate protocol DVH curves across multiple treatment modalities and complex treatment types (including SIB and multimodality therapies) using a unified approach. This universal protocol framework simplifies the training process by providing consistent data structures and calculation methods, even though the underlying dose distribution physics varies across different treatment types. The same protocol-based approach works for all modalities, reducing overall system complexity.
Data Source
Figure 1~3
AI summary
The invention facilitates the prediction of achievable dose distributions in radiation therapy treatment planning by training a prediction model for clinical DVH curves based on simplified DVH curves obtained using standardized methods. Machine learning is applied on pairs of clinical and simplified DVH curves to enable the future prediction of actual clinical DVH curves based on simplified DVH curves.