We use cookies to ensure that we give you the best experience on our website Learn more

May 21, 2019

PTCOG58

02 - Shape and texture analysis of skull-base chordomas to predict the outcome of pencil beam scanning proton therapy

;

Chordomas

proton therapy

machine learning

shape analysis

texture analysis

clustering

artificial intelligence

Abstract

Abstract

thumbnail

Keywords

Chordomas

proton therapy

machine learning

shape analysis

texture analysis

clustering

artificial intelligence

Abstract

Purpose: Skull-base chordomas (sbC) are rare bone tumors presenting within the clivus or spinal axis, characterized by significant tissue heterogeneity. Their shapes are variable, likely determined by proximity to surrounding anatomic barriers, making them a typical indication for proton therapy (PT). Here we investigate whether shape and/or textural features of the pre-treatment tumour correlate with clinical outcome. Methods: We retrospectively analyzed 50 sbC patients treated using PBS-PT at our institute. Pre-PT tumor shape and texture (Tab.1) were evaluated on DWI- and T2w-MRI images using in-house developed software and were classified using clustering algorithms (Kmeans) considering both tumor and organs-at-risk (OARs) features, which were cross-correlated with recurrences. Results: Fig,1 shows two contrasting tumours. (a) is a case with high sphericity and small surface-area (0,86 and 2409mm^2) which did not recur. In contrast the tumour in (b), with values of 0,27 and 8846mm^2 respectively, did recur. Overall, clustering analysis uniquely identified 4 out of 5 recurrences based on these morphological features. However, due to the small number of recurrences, this result is not statistically significant. Other features, such as compactness and signature-mean were significantly correlated (correlation-coeff.>0.85; Fig.1c) but no correlation was observed between textural features and clinical outcome. Conclusion: Sphericity and surface-area have been found to be potential predictive factors for treatment outcome in sbC. Greater surface values and protrusions (e.g. Fig.1b) could indicate a larger area of contact between tumor and normal-tissue, thus increasing the probability of tumor infiltration. We are currently extending our analysis to a larger patient cohort.

Discover over 20,000 new abstracts, posters and presentations from leading academic conferences every month. Stay on top of the latest findings, methodologies and discussions happening in your research field around the world.

Company

Legal

Follow us

© Copyright 2019 Morressier GmbH. All rights reserved.

© Copyright 2019 Morressier GmbH.
All rights reserved.