Adaptive Management of Patient Anatomical Variation Using Image Feedback

 

Di Yan, DSc.

Radiation Oncology, William Beaumont Hospital, Royal Oak, Michigan, USA

 

Image guided adaptive radiotherapy (IGART) is an application of dynamic feedback control in radiation treatment, which utilizes the image feedback of patient-specific anatomical & biological variations to frequently evaluate treatment quality and optimize the treatment plan if necessary by including the variations in the design of dose distribution. IGART expands the imaging/correct/delivery sequence of image guided intervention to a broader and more efficient methodology, in which evolving knowledge of patient-specific anatomical & biological variations assessed by multiple patient image feedback is used to dynamically optimize the treatment. A typical IGART model consists of two parts, (1) a patient-specific variation process as well as the corresponding estimator for process parameter and dose estimation, and (2) an adaptive controller to manage patient variations. This lecture outlines the IGART model for patient anatomical variation management alone.

Patient anatomical variation during the radiotherapy includes inter- and intra-treatment variations of patient anatomical shape and position caused by patient setup, beam placement, and patient organ physiological motion and deformation. In addition, tumor and normal organ dose response can also induce significant variations in tissue shape and position during the treatment course. Random process has been used to model patient-specific anatomical variation in radiotherapy. Two major parameters, the mean and the standard deviation of the variation, have been used to characterize this variation process. These two parameters could be either time-invariant indicating a stationary random process, or time-variant indicting a non-stationary random process. It will be relatively easy to estimate the parameters in a stationary process, but less straightforward to perform the estimation if the variation process is non-stationary. Probability distribution function (pdf) of patient-specific anatomical variation can also be estimated occasionally for certain type of variation i.e. patient respiratory induced organ variation. However, pdf estimation could not be reasonably accomplished for most patient-specific variations due to the limit number of samples in image feedback. Therefore, it is practically important to use these two process parameters directly to estimate the cumulative dose.

Adaptive controller provides an optimal control law that maps the estimated process parameters into the space of control or adjustment variables, such as machine output intensity map, patient position and etc. In theory of nonlinear stochastic control, the control law can be derived using the dynamic programming and solving the Bellman equation numerically. However, the numerical solution exists only for very simple cases. For radiotherapy process, a simplified and idealized control law can only be used for the education purpose, but not in the clinic implementation to manage the actual patient treatment. A practical controller in IGART can consists of a decision-making function to determine if the ongoing treatment plan needs to be modified, and a 4D adaptive planning engine to modify the ongoing treatment plan.

Patient anatomical variation in radiotherapy can be systematically managed using image feedback adaptive treatment technique. The fundamental difference between adaptive treatment technique and other image-guided techniques is the use of patient-specific treatment information. Adaptive technique intends to use all patient-specific dose information - including what has been delivered in the previous treatments, what can be delivered at the present treatment, and what would be delivered in future treatments - in the design of treatment plan. Therefore, treatment plan designed in adaptive radiotherapy is called 4D adaptive plan, which is in principle a treatment control law to manage treatment process. The objectives in adaptive planning optimization are constructed based on a selection of control strategies that could be either the online or the offline with one control action, multiple actions or continue actions. Selection of control strategy and number of control actions is, of course, dependent on the nature of patient variation process as well as the estimation uncertainties, and has to be determined considering also the clinical practice.