Standardisation of Breast Radiotherapy Structure Nomenclature using Machine Learning

When planning radiotherapy treatment for breast cancer, clinicians contour important structures on scans, such as the lungs, heart, and treatment target areas. These contours are saved with names, but the same structure can be labelled in many different ways (e.g. “Left Lung,” “L-Lung,” or “Lung-Left”). This lack of consistency makes it very difficult to combine large sets of patient data across hospitals, which is important for research, clinical trials, and quality assurance. In this project, we are investigating the following questions:

1. How different are naming conventions for breast radiotherapy structures across different hospitals and datasets?

2. Can an automated system accurately identify and standardise these structures, reducing the need for manual work?

3. Does such an automated approach perform reliably across multiple hospitals with different practices?

4. What efficiencies can be gained in research and clinical practice through automated standardisation?

By developing an automated approach to standardise breast radiotherapy structure naming, this project will make it easier to combine data from different centres. This will support larger-scale research, enable easier participation in clinical trials, and improve consistency in clinical practice. Eventually, it will save clinicians time, reduce errors, and help deliver more reliable and comparable evidence for improving breast cancer treatment outcomes.