Machine learning based dose guided real-time adaptive radiotherapy

A non-transitory computer readable medium comprising non-transitory computer-readable instructions, the computer-readable instructions comprising instructions for performing operations comprising:

• obtaining data representing a training patient anatomy at a first time within a training radiotherapy treatment fraction after a training radiotherapy treatment dose has been delivered by a radiotherapy device;
• computing a deviation between the data representing the training patient anatomy at the first time and reference data representing a reference training patient anatomy during the training radiotherapy treatment fraction, wherein the reference data representing the reference training patient anatomy indicates a prescribed training dose parameter to be delivered within the training radiotherapy treatment fraction, wherein computing the deviation comprises:
• determining a first amount of overlap between a tumor in a training patient corresponding to the data representing the training patient anatomy and an organ at risk in the training patient based on the obtained data representing the training patient anatomy at the first time, the first amount of overlap indicating a first quantity of radiation exposure to the tumor relative to the organ at risk;
• and
• determining a reference amount of overlap between the tumor in the training patient and the organ at risk based on the reference data, the reference amount of overlap indicating a reference quantity of radiation exposure to the tumor relative to the organ at risk;
• applying the computed deviation to a machine learning model to estimate one or more intra- fraction radiotherapy treatment parameters of a given function that provides a radiotherapy device parameter adjustment based on the one or more intra-fraction radiotherapy treatment parameters;
• computing, utilizing the machine learning model, a dose amount per time parameter and an amount of overlap parameter, a result of the given function specifying a dose amount per time factor as a function of the dose amount per time parameter and the amount of overlap parameter, the machine learning model trained to establish a relationship between the computed deviation and the one or more intra-fraction radiotherapy treatment parameters;
• determining a new patient anatomy data at a second time within a given radiotherapy treatment fraction after a new radiotherapy treatment dose has been delivered by the radiotherapy device;
• retrieving reference patient anatomy data for the given radiotherapy treatment fraction that indicates a prescribed dose parameter to be delivered within the given radiotherapy treatment fraction;
• computing a new deviation between the new patient anatomy data at the second time and the reference patient anatomy data during the given radiotherapy treatment fraction;
• and
• adjusting a parameter of the radiotherapy device based on a result of applying the new deviation to the given function to change an amount of radiotherapy treatment dose delivered at a third time, following the second time, within the given radiotherapy treatment fraction in accordance with the prescribed dose parameter to be delivered within the given radiotherapy treatment fraction.

A system comprising:

• a memory for storing instructions;
• and
• one or more processors for executing the instructions stored in the memory for performing operations comprising:
• obtaining data representing a patient anatomy at a first time within a training radiotherapy treatment fraction after a radiotherapy treatment dose has been delivered by a radiotherapy device;
• computing a deviation between the data representing the patient anatomy at the first time and reference data representing a reference patient anatomy during the radiotherapy treatment fraction, wherein the reference data indicates a prescribed dose parameter to be delivered within the radiotherapy treatment fraction;
• applying the computed deviation to a machine learning model to estimate one or more intra-fraction radiotherapy treatment parameters of a given function that provides a radiotherapy device parameter adjustment based on the one or more intra-fraction radiotherapy treatment parameters, wherein computing the deviation comprises:
• determining a first amount of overlap between a tumor in a training patient corresponding to data representing training patient anatomy and an organ at risk in the training patient based on the obtained data representing the training patient anatomy at the first time, the first amount of overlap indicating a first quantity of radiation exposure to the tumor relative to the organ at risk;
• and
• determining a reference amount of overlap between the tumor in the training patient and the organ at risk based on the reference data, the reference amount of overlap indicating a reference quantity of radiation exposure to the tumor relative to the organ at risk;
• computing a dose amount per time parameter and an amount of overlap parameter, a result of the given function specifying a dose amount per time factor as a function of the dose amount per time parameter and the amount of overlap parameter, the machine learning model comprising an artificial neural network trained to establish a relationship between a training deviation between a training patient anatomy and reference training patient anatomy and a set of intra-fraction radiotherapy treatment parameters;
• determining a new patient anatomy data at a second time within a given radiotherapy treatment fraction after a new radiotherapy treatment dose has been delivered by the radiotherapy device;
• retrieving reference patient anatomy data for the given radiotherapy treatment fraction that indicates the prescribed dose parameter to be delivered within the given radiotherapy treatment fraction;
• computing a new deviation between the new patient anatomy data at the second time and the reference patient anatomy data during the given radiotherapy treatment fraction;
• and
• adjusting a parameter of the radiotherapy device based on a result of applying the new deviation to the given function to change an amount of radiotherapy treatment dose delivered at a third time, following the second time, within the given radiotherapy treatment fraction in accordance with the prescribed dose parameter to be delivered within the given radiotherapy treatment fraction.