AI Rad Companion Organ RT - Improving the Radiotherapy workflow and reduce interobserver variability: NCCR experience

Initial Implementation 

Prior to the clinical implementation of the AI-Rad Companion software, an in-house pilot study was completed investigating both the spatial accuracy and the precision of AI-generated contours. The study involved the comparison of AI-Rad Companion generated contours on previously treated, or already contoured patients, for all potential sites and indications.

The oncologist’s feedback with regards to the various AI-generated contours was collected and assessed. The overall consensus was that it was safe to clinically proceed with the AI-Rad Companion software, as long as it was followed by a rigorous assessment of all contours by the primary oncologist, and whilst simultaneously adhering to strict pre-existing peer review guidelines.

Below is a case study example of a head and neck patient that was contoured using the AI-Rad Companion software

Case Study

History

A 67-year-old male patient presented with a one year duration of voice hoarseness. Head and neck cancer multidisciplinary team (MDT) confirmed diagnosis of moderately differentiated invasive squamous cell carcinoma of the left vocal cord with a TNM Staging of cT4aN0M0. Due to the patient’s age and extralaryngeal extension through the cricothyroid membrane they preferred a laryngeal preservation protocol. Recommendations from the meeting was that the patient was to undergo definitive chemoradiotherapy.

Process

A radiotherapy planning computed tomography (RTCT) scan with the patient in an immobilization mask was acquired to assist with planning of the radiotherapy treatment.

The resulting images were then sent to the Siemens Healthineers AI-Rad Companion software for contouring of organs at risk and regional lymph node groups. The original CT and structure set are then imported into the planning software to allow for contour modification and approval.

The biggest advantage in this instance and in general is the reduction in the time it takes to manually contour the organs and risk and the regional lymph node groups (Fig.1, Fig.2, Fig.3). The average time saving is about 50%, which then allows more time to generate the clinical plan and the potential to reduce the time to start the radiotherapy treatment. From our experience, we have found the process to take approximately 20 minutes.

Currently within our clinic, we are using the AI-Rad Companion for all radical radiotherapy treatments including head and neck (Fig.4), thorax, abdomen and pelvic cases. In addition to the time saving feature, it also reduces the inter-user variability in the generation of contours.

The naming of structures defaults to the Radiation Therapy Oncology Group (RTOG) guidelines, however, this is an editable feature which allows the user to customize the naming nomenclature to suit your department’s current structure.

We have also been in the position to compare the AI-Rad companion with another clinically approved AI contouring system developed which has been designed specifically for radiotherapy use.

In our experience, the AI-Rad Companion has a lot more auto contouring options and is ahead in accuracy. It also has the added benefit of producing more contours.