top of page

Radiation therapy patient outcomes – what can you control and measure?

29 Sep 2021

Is it possible to identify the variables that most affect patient outcomes given the multidisciplinary nature of radiation therapy and its workflow?

The sum of its parts A paper in 2012 on Image-Guided Radiotherapy: Has It Influenced Patient Outcomes? (Bujold et al. 2012), recognized the intricate multi-process and multidisciplinary nature of radiation therapy and asked whether IGR influenced patient outcomes. The radiation therapy workflow is much like the often quoted or misquoted Aristotle statement, “the whole is greater than the sum of its parts”, which some have argued should be translated as, “the whole is something besides the parts” (SE-Scholar, 2019). The RT workflow relies on each system, technology, and person to be exacting, accurate, discerning, and efficient. Each person or system works beside one another to achieve one aim: Create an optimal outcome for its patients.

Dose and its impact on patient outcomes Effective dosing and managing toxicity levels in the tumor and sparing healthy organs is vital to patients’ longer–term health outcomes. This view was echoed at a recent User Community EMEA Symposium II hosted by Mirada, with 15 external attendees representing ten different institutions. An RT Oncologist, in a breakout session, claimed that he hoped to find research to link and find direct correlations between parts of the RT workflow, specifically organ-at risk delineation and its impact on the dose and patient outcome. The oncologist is not alone in this aim.

UMCG, or the University MC Groningen cancer unit, were one of the first clinical hospitals to trial both the head and neck and prostate AI auto-contouring solution, DLCExpert, in 2018. Their vital concern was managing high patient toxicity levels as it affects patients‘ quality of life. Yet, UMCG, like all good scientific institutions, focused on the variables that they could control: inter-person variability via organ-at-risk contouring. UMGC had used Atlas-based contouring, but in collaboration with Mirada, trialed DLCExpert AI’s models and automation platform and then peer-reviewed the results using ESTRO guidelines. In terms of dose, their qualitative view was that they had more confidence around toxicity levels. It is suggested that this result was linked to the consistency of delineation created by the autocontouring provided by DLCExpert, which improved results from Atlas to AI.

In science, both qualitative and quantitative metrics are ideal. What was quantitatively measured was the time savings: UMCG reported a 57% drop in time spent for prostate. But, does saving time matter for patient outcomes?

Why does saving time matter? Cancer Research UK predicts there will be 514,000 new cases per year by 2035 (UK). Radiation therapy is one of the most effective cancer treatments in oncology, but effectiveness only matters if more patients can be seen and treated in a timely way. According to a survey published by ESTRO in 2015, just over half of the 3.41 million new cases which should receive radiation therapy actually did; this is a patient throughput issue. This statistic illustrating patient throughput problems is supported by data by the European Cancer Observatory, claiming that most countries achieve only a 50% success rate in treating patients with radiation therapy. However, given the ongoing strain that COVID is placing on every health care provider and their staff, looking for ways to increase the throughput for patients and achieve efficiencies must indeed be just as important as dose in effecting patient outcomes?

Returning to the analogy of the sum of its parts, many RT departments have broken down and investigated all the variables that affect the RT workflows. What can be humanly improved upon to increase accuracy, consistency and efficiency and what can be improved by AI and automation? Do these gains – big and small – impact the whole, and can algorithmic solutions and automation help humans be more discerning, accurate, and save time?

As we saw with the example of UMCG and AI autocontouring; they felt more confident around reducing inter-person variability and reported increased consistency, accuracy and also time saving after using DLCExpert. Autocontouring can take away the manual intensive work that can take up to one hour or 90 minutes for complicated cases like with the head and neck. Yasmin McQuinlan, who works at Guys’ NHS Hospital, London and an inhouse Research Dosimetrist at Mirada, believes AI is a valuable assistant to experts. AI auto-contouring provides more time for quality checks and complex cases so the experts act as gatekeepers to their clinic’s standards. DLCExpert is not just an autocontouring product, but it also comes with a free Zero-Click automation platform. This platform provides the benefits of speed and efficiency. It can be plugged directly into existing clinical systems, so experts will not experience any changes to their existing RT workflow. Once a patients simulation scan is complete, this will land at the editing workstation or TPS of choice, already contoured, saving time. DLCExpert and its automation platform may help your RT workflow increase time savings and accuracy, allowing you to put your people where they matter most, with patients and working on complex cases.

If you want to put the DLCExpert’s algorithm to test, Take the Turing test: See if you can see the difference between a machine or human contours? Mirada aims to do more research to help clinics link patient outcomes and dose more closely to AI autocontouring. If you have an opinion on this, please let us know.

References: Image-Guided Radiotherapy: Has It Influenced Patient Outcomes? Alexis Bujold, Tim Craig, David Jaffray, Laura A. Dawson, Seminars in Radiation Oncology, Volume 22, Issue 1, 2012, Pages 50-61, ISSN 1053-4296,



bottom of page