Use of radiation to treat cancer
Future of Radiotherapy
Potential of AI in Radiotherapy
One of the most challenging aspects which we need to overcome before beginning treatment is the delineation of the target volume. This involves creating a model of the tumour, outlining the target and any organs that may be at risk. Currently, this task is performed manually by an oncologist (a doctor who treats cancer) using specially designed software to draw contours around the regions of interest. While the task demands considerable clinical judgement, it is also laborious and repetitive. Consequently, it is an extremely time-consuming process, often taking up to several hours per patient.
​
However, the recent implementation of a machine learning model to create auto segmented volumes showed a reduction in contouring time of 93%, as the planning oncologist was required only to edit the volumes [5].
This highlights the future potential of AI in reducing planning time whilst maintaining the high standard of treatment, allowing resources to be better used elsewhere alongside an increase in workflow efficiency.
Figure 1) Example of tumour volume delineation in two patients. A, B, and C are the observers (different physicians). The upper row shows a tumour with differences in the contouring of the volume, whereas the differences are more limited in the patient represented in the lower row.
FLASH Radiotherapy
Another exciting technological advance is ultra-high dose rate radiotherapy (FLASH). Conventional radiotherapy is typically delivered with dose rates around 0·03 Gy/s, over 2–7 minutes. But when radiotherapy is delivered in ultra-high dose rates (>40 Gy/s) in less than 1 s [2], minimal deleterious effect has been observed on healthy tissue. ​Studies demonstrate that alongside the reduction in toxicity, FLASH also retains the same detrimental effects to the tumour as conventional radiotherapy. This has been proven in mice, but only a few human trials. Before this becomes available for commercial use, more data will be required to further confirm the safety of this technique in humans, as the available data is limited.
​
It is not fully understood how the mechanism works as of yet. One explanation for this effect could be that hypoxic tissues (tissues that are deprived of oxygen) are more resistant to radiation and therefore less likely to become damaged than well-oxygenated tissues (demonstrated by figure 2).
Figure 2)Possible mechanisms for the protection of normal tissue during FLASH -RT treatment.
​
If this becomes implemented alongside the machine learning model (which can quickly and accurately generate a target volume for the dose), the efficiency and success rate of treatment may drastically improve.
MR-LINAC and adaptive Radiotherapy
The MR-linac design combines a linear accelerator and an MRI system. This integration aims to facilitate high-resolution MRI imaging directly from the treatment table, allowing for real-time visualization of anatomical changes throughout radiotherapy [1]. MRI serves to monitor both inter-fraction and intra-fraction motion. Inter-fraction motion refers to changes in the patient's anatomy from one treatment session to another, while intra-fraction motion pertains to movements that occur within a single treatment session, such as respiration or random movement if the patient is still for an extended period. This data provides valuable insights for adaptive radiotherapy.
​
​
Figure 3) The MR-Linac in the Manchester cancer research centre, the second site in the UK and sixth in the world at the time of writing. [3]
​
​
Daily MRI scans can verify soft-tissue positioning or prompt daily treatment adjustments, while MRI during beam delivery enables dynamic dose tracking and real time treatment adaptation. Regular anatomical and functional MRI scans can evaluate treatment response, meaning the clinical team can decide if the current dosage and location is the best option. The ultimate goal is to get to real-time plan adaptation while accounting for the dose delivered so far, as to not cause radiation toxicity. Some problems with this method are that, in some cases, the time involved in the day to day modifications may not provide a noticeable difference. This could result in a waste of time and resources, which are both of extreme value in a medical environment - especially nowadays with the increased pressure on the healthcare system.
