The Dream of Hassle-Free Radiotherapy

Is It Possible to Transform a Single MRI Image into a Complete 3D Scan?

Inside the Extraordinary World of SynthRAD2023

Can deep learning manage to synthesize CT including bones? Image from MRI? Image created by author.

Imagine if doctors treating cancer could rely on a single daily MRI or cone beam CT (CBCT) scan to plan radiotherapy, without the added time and inconvenience of bringing in a separate conventional CT. In principle, this could streamline treatment and spare patients multiple scans, but it also raises a crucial question: how do we get the necessary “electron density” information if no CT is taken? A ground-breaking approach has emerged: synthetic CT (sCT). Instead of acquiring a new CT scan each time, scientists and engineers have begun using advanced machine learning methods to transform MRI or CBCT images into a precise CT-like representation. This process, called CT synthesis, provides the density information radiotherapy planning needs.

The Time is Now

Over half of all cancer patients receive radiotherapy, which works by aiming targeted radiation beams at tumors. However, healthy organs need protection, so precision is key. Conventional CT (Computed Tomography) scans directly measure Hounsfield Units (a measure of electron density), which helps plan how much the radiation beam will weaken or…