Skip to main content
. 2021 Jul 7;22(8):16–44. doi: 10.1002/acm2.13337

TABLE 1.

Traditional KBP studies aimed to predict dose–volume histograms (DVHs) for providing a starting point for the plan optimization process

Ref. Method Approach/Model Key features Purpose
46 MB Support Vector Regression Organ volumes, shape and DTH To model functional relationship between DVH and patient anatomical shape information.
4 MB Fitting using least square min. OAR distance to PTV To translate key feature correlation to mathematical relationships between OAR geometry and expected dose.
47 MB Stepwise multiple regression DTH To build feature models to identify the variation of anatomical features contributing to OAR dose sparing.
48 MB Stepwise multiple regression Target, OARs, overlap volumes and DTH Extension of Yuan et al. for intra‐treatment‐modality model (IMRT – Tomotherapy)
49 MB

Stepwise multiple regression

 Target, OARs, overlap volumes and DTH, fraction of OAR outside treatment field To build two predictive models (single‐sparing and standard model) to characterize the dependence of parotid dose sparing on patient anatomical features in the summed (primary + boost) plan, rather than two completely separate models.
50 AB Direct Overlapping volume

To select a reference plan from a library of clinically approved/delivered plans with similar medical conditions and geometry
51 AB Direct PTV shape, volume, three spherical coordinates of PTV with respect to OAR OVH To develop a knowledge‐driven decision support system to assist clinicians to pick plan parameters and assess radiation dose distribution for a perspective patient
52 MB Kernel Density Estimate Distance to PTV

To develop an automated treatment planning solution that iteratively

  1. optimize training set

  2. predicts DVHs for OARs

  3. generates clinically acceptable plans
53 MB Ensemble Anatomical features, DTH To combine strengths of various linear regression models to build a more robust model
54 MB K‐nearest neighbors Generalized‐DTH To characterize DVH variance in multiple target plans
33, 34, 35, 36, 37, 38, 39, 40, 42, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78

RapidPlanTM Eclipse® treatment planning software:

Algorithm is divided into two components: 1) Model configuration and 2) DVH estimation.

  1. Mode configuration is divided into data extraction phase and model training phase

  2. DVH estimation consists of DVH estimation phase and objective generation phase

Abbreviations: AB, atlas‐based (Direct or Indirect method); DTH, distance‐to‐target histogram; MB, model‐based; OVH, overlap volume histogram.