Assessing uncertainties in image registration

Researchers in the US and Denmark have demonstrated a new way to evaluate the uncertainty of deformable image registration (DIR), which tracks the changes of tumours in medical images. According to the study, which was based on real images of patients with prostate cancer, a “distance discordance metric” (DDM) is better able to evaluate uncertainties in DIR than existing methods. The new method should allow radiotherapists to provide cancer patients with more accurate treatment plans (Phys. Med. Biol. 61 6172).

The aim of radiotherapy is to localize radiation to a tumour while sparing healthy tissue, so that a patient does not have to suffer more radiation than necessary. To do this, radiotherapists make use of medical imaging, such as CT and MRI, to identify a tumour’s boundaries. But this is difficult if the relevant part of the body undergoes motion, or if the patient loses weight or the tumour itself begins to change in size. […]

The rest of this article is available here.