Establishing a Machine Learning Pipeline to Automate Tumor Segmentation from Written Radiology Report Data

Tumour segmentation facilitates precise disease localisation/staging to inform appropriate treatment. Manual segmentation performed by radiologists is time consuming, expensive, susceptible to intra- and interobserver variability, and can lack precision. Development of artificial intelligence (AI) models capable of performing automated tumour segmentation across a range of imaging modalities may help to overcome these limitations and improve patient care. These models will also be suitable for use across a range of research applications.