Automated radiotherapy treatment planning for head and neck cancer patients

Abstract

The aim of this project is to develop high quality automated radiotherapy treatment plans for H&N patients.

The principal research questions are:

  1. Are Artificial Intelligence (AI) generated structures of sufficient quality for use in H&N routine clinical use?
  2. Can the quality of AI generated structures be independently & automatically quality assured?
  3. Is a combination of knowledge based treatment planning (KBP) with additional automated optimisation superior to KBP alone?

Proposed methods: A group of previously treated head and neck treatment plans will be selected for the project, and automation of the various stages of the radiotherapy treatment planning process will be developed as follows:

  1. AI auto-contouring software will be utilised to generate relevant structures which will be compared to a gold standard set of structures. These gold standard structures will be produced using multi-modality imaging and with the assistance of a clinical oncologist.
  2. Automated methods will be used to perform quality assurance checks on the structures produced by the auto-contouring system. The aim of this checking process will be to identify gross errors by determining if structures fit within expected normal ranges based on their geometry.
  3. Treatment plans will be produced using the Varian EclipseTM Treatment Planning System with the aid of the Eclipse Scripting API (ESAPI) and the Varian RapidPlanTM KBP solution, all of which are already available for use at the centre. Plans will initially be created using KBP alone and then a method of automating the addition of further optimisation parameters to the existing KBP solution using the ESAPI will be investigated. If automation proves problematic then a manual method will be used to compare plan quality between standard KBP and KBP with additional plan optimisation. Patient opinions regarding the use of artificial intelligence and computer automation in the production of their own radiotherapy treatment plans will be captured with the aid of a questionnaire, in either paper or electronic format.

Research impact: This research will inform regarding possible quality and efficiency improvements at all stages of the treatment planning process. It is anticipated that some of the research will lead to the adoption of new planning processes at the trust.