Meeting Banner
Abstract #1114

Supervised machine-learning enables segmentation and evaluation of heterogeneous post-treatment changes in multi-parametric MRI of soft-tissue sarcoma

Matthew David Blackledge1,2, Jessica M Winfield1,2, Aisha Miah3, Dirk Strauss4, Khin Thway5, Veronica A Morgan1,2, David J Collins1,2, Dow-Mu Koh1,2, Martin O Leach1,2, and Christina Messiou1,2

1Cancer Research UK Cancer Imaging Centre, The Institute of Cancer Research, London, United Kingdom, 2MRI Unit, The Royal Marsden NHS Foundation Trust, Sutton, United Kingdom, 3Department of Radiotherapy, The Royal Marsden NHS Foundation Trust, London, United Kingdom, 4Department of Surgery, The Royal Marsden NHS Foundation Trust, London, United Kingdom, 5Department of Histopathology, The Royal Marsden NHS Foundation Trust, London, United Kingdom

Multi-parametric MRI enables non-invasive response assessment in heterogeneous soft-tissue sarcomas, but evaluation of post-treatment changes in MRI parameters requires segmentation of cellular tumour-tissue, which might be expected to respond to treatment, from necrotic/cystic regions and fat. Six supervised Machine-Learning methods were explored using a randomized cross-validation approach, from which a candidate method (automatic Kernel Density Estimation) was selected owing to its high validation accuracy and automatic selection of hyper-parameters. The automatic-KDE method enabled evaluation of post-radiotherapy changes in volumes and ADCs of each tumour component, and provided visual depiction of heterogeneous changes in multi-parametric MR-images.

This abstract and the presentation materials are available to members only; a login is required.

Join Here