| Intensity modulated radiation therapy(IMRT)has the ability to ensure proper dose coverage and conformity for target while provide more dose sparing for organs-at-risk and adjacent normal tissue,by the modulation to nonuniform photon fluence maps.It has been widely used in clinical.In IMRT treatment planning,the ideal optimizing goals are unknown before planning,a medical physicist would usually pick optimization objectives based on population-based clinic protocol and later have them tweaked repeatedly through manual try-and-error until finding the optimal plan.However,such try-and-error process leads to the situation that the IMRT plan quality and planning efficiency are limited by the actual investment(e.g.time,energy)can be put into planning or knowledge richness of planners.Knowledge-based optimization methods can effectively ensure the planning efficiency and homogeneity of plan quality by predicting the appropriate dosimetric goals for novel patients by building the geometry-dosimetry correlation model based on prior plans learning and use such dosimetric goals as optimization initial objectives.But most of predictions were formed of DVH or endpoints,which is cumulative distribution summed over the spatial dose.Using DVH or endpoints as dosimetric goals in optimization would make the optimization lose the exquisite control to spatial dose and stay on an organ level.Predicting voxel-based dose distribution and using it as optimization objective should be most desired solution for treatment planning.Most researches are focusing on building as more precise prediction model as possible,but the application of prediction model in optimization was barely investigated.But considering the immanent uncertainty of prediction and complexity of spatial dose information,exploiting full advantages of prediction in planning optimization regimes should be more investigated and paid attention to in current prediction guided optimization method research.Therefore,based on the dose distribution prediction model that we had already built,in order to make the most of dose distribution prediction into planning optimization and provide optimal solution to the most extent,we had proposed a dose distribution prediction and gEUD based treatment planning optimization method for IMRT,which utilizes dose distribution prediction to formulate a voxel-based objective for optimization to approach the prediction,and introduces equivalent uniform dose sparing for OAR for compensating for prediction error while expanding solution space of the optimization.From the comparison between optimized plan and original plan of 10 selected GYN IMRT,we can see that proposed method could ensure a proper PTV dose coverage,homogeneity while significantly reduce the OAR doses.Moreover,into the consideration of what effect the prediction error may bring into optimization,,we had proposed a predicted dose sequence-based optimization method,which could increase the freedom of by building more inclusive prediction applying optimizing objective.The validity of proposed method was evaluated by comparing its resulting plan quality with original plan as well as the resulting plan quality of voxel-matching dose distribution prediction guided optimization.Based on the comparison results of 10 GYN IMRT plan,we can know that proposed method could ensure a plan quality that is comparable as or ever better than the original plan,however there’re trade-off between proposed optimized plan and voxel-matching dose guided plan.We had successfully proposed a dose distribution prediction and gEUD based treatment planning optimization method for IMRT,which is able to make full use of the 3D dose prediction while ensure the output plan.quality;And also proposed a predicted dose sequence-based optimization method by exploring a new prediction dose distribution guided optimization objective,this method could effectively ensure the quality homogeneity of output plan. |
