Evaluation of a knowledge-based planning solution for head and neck cancer

Area of Science:

• Radiation Oncology
• Medical Physics
• Cancer Treatment Planning

Background:

• Automated and knowledge-based planning aim to standardize radiotherapy plan quality.
• RapidPlan utilizes patient plan libraries to predict achievable dose-volume histograms (DVHs) and set optimization objectives.

Purpose of the Study:

• To benchmark RapidPlan's knowledge-based planning against traditional clinical plans for head and neck cancer patients.
• To evaluate RapidPlan's performance using different patient plan libraries and patient groups.

Main Methods:

• Volumetric modulated arc therapy (VMAT) plans from 60 head and neck cancer patients were used.
• RapidPlan models were trained on subsets of patient data and tested on distinct evaluation groups.
• Plan quality was assessed using homogeneity index (HI) and mean organ at risk (OAR) doses.

Main Results:

• RapidPlan demonstrated improved homogeneity index (HI) values compared to clinical plans in recent patients.
• Mean doses to salivary glands and swallowing muscles were comparable or reduced with RapidPlan.
• Oral cavity dose increased in one model, and plan consistency varied for outlier patients.

Conclusions:

• RapidPlan-generated plans are comparable to clinical plans when patient geometry aligns with model data.
• Models trained for comprehensive sparing can be applied to patients requiring only specific OAR sparing, enabling library sharing.
• Further research is needed to optimize outlier detection and model library population for knowledge-based planning.