Dosimetric features‐driven machine learning model for DVH prediction in VMAT treatment planning

Abstract

Purpose Few features characterizing the dosimetric properties of the patients are included in currently available dose‐volume histogram (DVH) prediction models, making it intractable to build a correlative relationship between the input and output parameters. Here, we use planning target volume (PTV)‐only treatment plans of the patients (i.e., the achievable dose distribution in the absence of organs‐at‐risk (OAR) constraints) to estimate the potentially achievable quality of treatment plans and establish a machine learning‐based DVH prediction framework with the use of the dosimetric metric as model input parameters. Methods A support vector regression (SVR) approach was used as the backbone of our machine learning model. A database containing volumetric modulated arc therapy (VMAT) plans of 63 prostate cancer patients were used. For each patient, the PTV‐only plan was generated first. A correlative relationship between the OAR DVH of the PTV‐only plan (model input) and the corresponding DVH of the clinical treatment plan (CTP) (model output) was then established with the 53 training cases. The prediction model was tested by the validation cohort of ten cases. Results For the training cohort, the checks of dosimetric endpoints (DEs) indicated that 52 of 53 plans (98%) were within the 10% error bound for bladder, and 45 of 53 plans (85%) were within the 10% error bound for rectum. In the validation tests, 92% and 96% of the DEs were within the 10% error bounds for bladder and rectum, respectively, and eight of ten validation plans (80%) were within the 10% error bound for both the bladder and rectum. The sum of absolute residuals (SAR) achieved a mean of 0.034 ± 0.028 and 0.046 ± 0.021 for the bladder and rectum, respectively. Conclusions A novel dosimetric features‐driven machine learning model with the use of PTV‐only plan has been established for DVH prediction. The framework is capable of efficiently generating best achievable DVHs for VMAT planning.

Development and evaluation of machine learning models for voxel dose predictions in online adaptive magnetic resonance guided radiation therapy

“…By contrast, most previous KB prediction models aimed at identifying planning improvements related to OAR dose sparing while maintaining uniform target coverage, eg PTV V95% > 95%. 23,24,26,27,31…”

“…But the most common method used recently falls under the general term of knowledgebased (KB) prediction models. 22 KB methods have been used to predict achievable dose-volume histograms (DVHs), 23,24 plan metrics, 25 and full 3D dose distributions. [26][27][28][29][30][31] In fact, using information from previously treated patients and plans to better inform future procedures is now present in nearly all aspects of RT.…”

“…Furthermore, 3D dose predictions could serve as novel inputs for treatment plan optimization that may completely alter the paradigm under which RT planning currently occurs. 26,28,30 Despite the growth in applications of KB methods to solving interesting RT problems, a majority of the most relevant work has focused on treatment sites such as prostate, 23,24,26,28,31 head and neck, 25,29,30 breast, 28 and lung. 36 There are limited studies dedicated to abdominal cancers 15,27 because they are notoriously difficult to treat and there is minimal consensus regarding the best RT treatment approaches.…”

Application of dose-volume histogram prediction in biologically related models for nasopharyngeal carcinomas treatment planning

“…However, one concern regarding the DTH and OVH is that their simplicity may lead to inaccurate presentation of the interpatient variations in anatomical features, which may have an impact on the organ dose deposition [ 5 , 6 ], especially for complex tumour volumes in close proximity to critical structures such as those observed in nasopharyngeal carcinomas (NPCs). The dose deposited in an OAR voxel depends not only on its distance from the PTV surface but also on the treatment beam orientation [ 5 , 7 , 8 ].…”

Dose distribution prediction in isodose feature‐preserving voxelization domain using deep convolutional neural network

“…Knowledge‐based planning (KBP) typically involves predicting the achievable dose–volume histograms (DVHs) or dose distribution using machine learning algorithms or searching population‐based plan library, aiming to improve the plan quality and planning efficiency . A variety of KBP‐based methods have been proposed, including the overlap volume histogram (OVH) method, distance‐to‐target histogram (DTH) method, differential distance to target histogram (dDTH) method, contextual atlas regression forest (cARF) method,and dosimetric features‐driven method …”