The latest forecast predicts that for 2023, Australia will record the highest number of prostate cancer cases ever recorded, with 25,487 cases expected to be diagnosed at a rate of 154.6 cases per 100,000 males.

— Prostrate Cancer Foundation of Australia
Overviews

The Challange

Around 4750 Queensland males are diagnosed with prostate cancer every year, driving a need to streamline the treatment process so that it is as efficient and effective as possible.

A common treatment for prostate cancer is radiation therapy. In this treatment, a controlled dose of radiation is used to kill or damage cancer cells. However, imprecise dosage of radiation can effect nearby tissues and organs, such as the bladder and bowel. Avoiding these negative impacts of radiation relies on the use of imaging throughout a patient’s treatment, or image guided radiation therapy, to ensure the correct radiation dose is delivered. These images are currently labour intensive to review due to the limited tools available to clinicians.

Currently, an initial CT scan of the treatment area of the patient before they start their course of radiation therapy is used to plan a complete course of radiation therapy. This plan provides information for all the treatment machine settings to deliver the required radiation dose based on the expected changes of the cancer and other changes in the patient’s body over the treatment period

Before treatment commences on any given day, the clinicians use a new CT scan to assess how closely the patient’s cancer and body matches their expected plan. If there are major differences, radiation therapists who deliver the treatment may need to add additional procedures before treatment can be delivered such as repeating patient treatment preparation then re-imaging to re-planning the day’s treatment. This is time consuming, and associated changes in the body during while these additional processes are happening, such as bladder filling, can become problematic.

A system that can automatically check patient progression and alert clinicians to unacceptable changes in comparison with the treatment plan and assist them with making optimal decisions will minimise the time on the bed that patients endure during treatment.

Project
Project

The Solution

PAG and radiation therapists based in QLD Health are researching the use of AI assisted segmentation techniques that will be used with image review tools to determine if the current state of a patient’s prostate cancer is within a suitable level of tolerance in relation to the treatment plan and the day’s treatment can proceed, or if additional procedures are required first.

We are using cutting edge machine learning to build the most accurate AI assistant for the clinicians, which is restricted by several clinical and logistical requirements.

The adopted segmentation method must be accurate when applied to the lower quality CT scan that is that is taken at the beginning of each new treatment day. These treatment CT scans are lower quality due to CT equipment modifications required when it is used on radiation therapy treatment machines.

The software under construction must provide easily interpreted information in both interactive visual and tabular format. The high pressure environment on the treatment floor necessitates that all relevant information is clearly conveyed so that decisions in treatment are optimal and not prone to error. To this end, the software is designed in collaboration with QLD Health radiation therapists to ensure that information is most useful, intuitive and accessible to the end users.

The extremely busy and high pressure nature of this work, means providing clarity to clinicians so they can make the right decision as quickly as possible is our key objective.

The Impact

The goal of this research is to improve the treatment times experienced by patients, without compromising the quality of treatment received.

The AI assisted segmentation program used alongside image review tools will result in faster decisions and throughput of patients. This time saving between scanning and treatment commencement each day will enable more accurate treatments, as changes in the body whilst the patient is on the bed compromise the accuracy of the treatment.

The AI assisted program is intended to allow radiation therapists to focus more on patients whose cancer and body are diverging dramatically from their plan, which will reduce stress and pressure in the workplace, and result in overall better outcomes for all patients.

Currently proposed fully automated solutions are not widely implemented due to concerns of safety, or being impractical. Our approach, being AI assisted, rather than fully automated, bypasses these current concerns, with the clinicians retaining the overriding treatment decision.

The software is customised to the current treatment approaches of QLD Health. It is being designed to enhance current practice, and, potentially, moving towards even more automated approaches that may be possible in the future.

Project