Deep Convolutional Neural Network for Radiation Therapy Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radiation therapy treatment planning is time-consuming and complex, particularly due to the need to balance target dose delivery and organ-at-risk sparing, often requiring trial-and-error processes and skilled dosimetrist judgment, which can lead to suboptimal plans and prolonged treatment times.
Innovation Solution
A deep convolutional neural network is trained on patient data to predict machine parameters for radiation therapy treatment plans, enabling real-time generation of high-quality plans that optimize target irradiation while minimizing exposure to healthy tissues, using imaging information and treatment constraints to automate the planning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional offline treatment planning is used, then treatment plans can be thoroughly optimized, but treatment time is prolonged and productivity is reduced
Solution Approach 1:
The patent uses deep learning models to copy and learn from high-quality treatment plans generated by expert dosimetrists. The neural network is trained on previously optimized treatment plans and their corresponding patient data, enabling it to generate new treatment plans by copying the patterns and principles from the training data, thus achieving both high quality and fast generation times
Solution Approach 2:
The patent replaces the manual mechanical process of dosimetrist planning with an automated neural network system. The neural network substitutes for the human expert's iterative optimization process, using computational algorithms to generate treatment plans automatically, thereby dramatically increasing productivity while maintaining quality through learned patterns from expert examples
2Reliability
If trial-and-error planning process is used, then treatment objectives can be met, but time consumption increases and productivity decreases
Solution Approach 1:
The patent performs preliminary action by pre-training the neural network on extensive datasets of treatment plans and patient data before actual treatment planning is needed. This pre-training allows the system to make immediate, informed decisions during real treatment planning without requiring time-consuming trial-and-error iterations, as the neural network has already learned optimal solutions from previous cases
Solution Approach 2:
The patent incorporates feedback mechanisms where treatment plan quality metrics and clinical outcomes are fed back into the training process. The neural network learns from the results of treatment plans and adjusts its parameters accordingly, enabling it to meet treatment objectives more reliably while reducing the need for iterative adjustments during actual planning
3Adaptability or versatility
If manual delineation by health care provider is used, then treatment plans can be customized, but the process becomes time-consuming and productivity is reduced
Solution Approach 1:
The patent implements self-service by enabling the neural network to automatically perform the delineation and treatment plan generation process without requiring manual input from health care providers for each patient. The system self-adjusts to individual patient anatomies and treatment requirements by processing their specific imaging data, thereby maintaining customization while dramatically improving planning efficiency
Solution Approach 2:
The patent utilizes parameter changes by dynamically adjusting the neural network's output parameters (beam angles, doses, MLC positions) based on the specific patient's anatomical parameters extracted from imaging data. This allows the system to customize treatment plans for each patient while maintaining high productivity, as the parameters are automatically optimized rather than manually adjusted
Data Source
AI summary
Systems and methods can include a method for training a deep convolutional neural network to provide a patient radiation treatment plan, the method comprising collecting patient data from a group of patients, the patient data including at least one image of patient anatomy and a prior treatment plan, wherein the treatment plan includes predetermined machine parameters, and training a deep convolution neural network for regression by using the prior treatment plans and the corresponding collected patient data to determine a new treatment plan. Systems and methods can also include a method of using a deep convolutional neural network to provide a radiation treatment plan, the method comprising retrieving a trained deep convolution neural network previously trained on patient data from a group of patients, collecting new patient data, wherein the new patient data includes at least one image of patient anatomy, and determining a new treatment plan for the new patient using the trained deep convolutional neural network for regression, wherein the new treatment plan has a new set of machine parameters.