Abstract #2802

Gaussian mixture modelling of combined functional imaging parameters provides new insight into tumour heterogeneity

Jessica M Winfield^1,2, Matthew D Blackledge², Aisha Miah³, Dirk Strauss⁴, Khin Thway⁵, David J Collins^1,2, Martin O Leach^1,2, Sharon L Giles^1,2, Daniel Henderson³, and Christina Messiou^1,2

¹MRI, Royal Marsden Hospital, Sutton, United Kingdom, ²Division of Radiotherapy and Imaging, Cancer Research UK Cancer Imaging Centre, Institute of Cancer Research, London, United Kingdom, ³Department of Radiotherapy, Royal Marsden Hospital, London, United Kingdom, ⁴Department of Surgery, Royal Marsden Hospital, London, United Kingdom, ⁵Department of Histopathology, Royal Marsden Hospital, London, United Kingdom

Multi-parametric functional imaging may enable non-invasive assessment of response to treatment in soft tissue sarcomas. Image analysis is complicated, however, by the highly heterogeneous nature of these tumours, which can include regions of cellular tumour, fat, necrosis and cystic change that may respond differently to treatment. In this study, patients with retroperitoneal sarcoma were imaged before and after radiotherapy using DW-MRI, Dixon and pre-/post-contrast T₁-w imaging for evaluation of enhancing fraction (EF). Gaussian mixture modelling was applied to classify pixels in the tumour volume according to their functional imaging behaviour, combining ADC, fat fraction and EF to characterise tumour components. This method enabled segmentation of highly heterogeneous tumours and estimation of mean ADC and volume of each tumour component. Heterogeneous changes post-radiotherapy were summarised in tissue classification maps, which combine multiple functional imaging parameters. Combined analysis of functional imaging parameters may provide greater insight into tumour behaviour, for example identification of viable tumour.

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

Join Here