A novel inverse treatment planning training method for inexperienced treatment planners.

Abstract

Knowledge-based treatment planning (KBP) has been traditionally employed to improve consistency and efficiency in radiation treatment planning. This work introduces a unique use of the KBP model to build site-specific treatment planning modules to provide inexperienced treatment planners with planning intuitions and enhanced learning experience. Our experimental design divided seven planners into two main categories: (a) three planners with more than 6 months of experience and (b) four planners with less than a month of experience. Prior to going through the training module, each planner optimized a treatment plan until they were satisfied with the plan quality on two validation cases. These two plans effectively recorded their baseline planning performance. This was followed by the initiation of treatment planning training on three different training cases. The training process required each planner to first reach their best plan quality, which was then evaluated by the training module. The training module generates a spatial representation of areas of improvement in the form of structures with instructions regarding the current dose value and its ideal dose value. Each planner then focused on incorporating the suggestions made by the training module (i.e., iteration 1, 2…). After training, each planner transferred their knowledge from training to re-optimize the two validation cases. The performance of each planner was evaluated based on the improvements in plan quality before and after the training via clinically relevant dose-volume metrics. The results of this heuristic approach show plan quality improvements after the training for less experienced planners. For instance, the mean dose to bladder, rectum, and penile bulb was reduced by 3.0 Gy, 4.0 Gy, and 2.5 Gy, respectively. An average reduction of 4.6% and 8.2% was achieved for bladder V_40Gy and rectum V_45Gy. The quality of treatment plans improved as planners accounted for suggestions made by the training module. As anticipated, the treatment plan quality of more experienced planners was comparable before and after the training. Overall results of this work demonstrate the feasibility of training inexperienced planners with the help of the training module implemented in Eclipse treatment planning system.