Cone Beam Computed Tomography guided Palliative Radiotherapy

Erasmus MC and Elekta worked together to speed up the treatment preparation of cancer patients. More specifically, patients that are treated for bone metastases with radiotherapy. Often these patients are in pain, and they benefit from a fast start of the treatment. Currently the treatment preparation takes a full day, while the partnership aims to reduce it to one hour.  

Each year in the Netherlands, about 120.000 patients get the diagnosis cancer. More than 50% of these patients are treated with radiotherapy as part of their treatment. At Erasmus MC, 20% to 30% of all radiotherapy patient are treated with palliative intent. The majority of the patients treated with palliative intent are treated for bone metastasis. Bone metastases are often painful, and therefore patients benefit from a fast start of treatment.  

Currently the preparation of palliative radiotherapy takes approximately one day. The patient arrives at the hospital in the morning, has an intake with the radiation oncologist and a CT scan. And while the patients wait until the end of the afternoon, the tumor and healthy organs are delineated on the CT scan, and the treatment plan is optimized. Right before the treatment starts, while the patient in lying on the treatment couch, a cone-beam CT scan is acquired. The cone-beam CT scan is acquired to verify if the patient is positioned correctly. The scan quality of the cone-beam CT is not sufficient for tumor delineation or treatment optimization. Within this project Erasmus MC and Elekta explored deep learning to improve the image quality of the cone-beam CT scan, such that it can be used for tumour delineation and treatment planning.  

For 70% of the patients the improvement was sufficient to use the scans for tumour delineation. This would mean that for these patients the original CT scan can be omitted, and the following workflow would be possible. Patients would arrive in the hospital, have an intake, will be positioned on the treatment table, a cone-beam CT scan will be acquired, that is improved with deep learning, tumour delineation and treatment planning take place, and the patient is treated right away. This would reduce the treatment preparation time to approximately one hour. See here a picture of the original cone-beam CT scan with poor image quality, the cone-beam CT scan improved with deep learning and a conventional CT scan (as golden standard). Tumour delineation on the improved cone-beam CT was, for this patient, as good as with the reference CT.