MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Incorporating a Novel Radiomics Framework for Biologically Optimised Prostate RadioTherapy (BiRT)

A Haworth1*, Y Sun2 , M Ebert3 , H Reynolds4 , J Betts5 , D Wraith6 , C Mitchell7 , D Murphy8 , S Williams9 , (1) University of Sydney, Sydney, New South Wales, (2) University of Sydney, Sydney, New South Wales, (3) The University of Western Australia, Nedlands, ,(4) Peter MacCallum Cancer Centre, Melbourne, Victoria, (5) Monash University, Clayton, Vic, (6) Queensland University of Technology, Kelvin Grove, Qld, (7) Peter MacCallum Cancer Centre, Parkville, Vic, (8) Peter MacCallum Cancer Centre, Parkville, Vic, (9) Peter MacCallum Cancer Centre, Melbourne, Victoria

Presentations

(Monday, 7/30/2018) 4:30 PM - 6:00 PM

Room: Davidson Ballroom A

Purpose: To describe a framework for Biologically Optimised prostate RadioTherapy (BiRT) that incorporates multiparametric MRI (mpMRI). The model requires knowledge of tumour location, tumour cell density, tumour aggressiveness and hypoxia. We describe a quantitative, voxel-by-voxel approach for extracting tumour characteristics from mpMRI to inform a radiomics framework for biological optimisation of focal brachytherapy treatments.

Methods: In vivo mpMRI scans were acquired from 30 patients prior to radical prostatectomy. Whole mount histology obtained after surgery was used to retrieve the “ground truth� information of tumour location and characteristics for the radiomics framework development. The prostate was contoured on T2-weighted images by a radiation oncologist. To achieve a high spatial accuracy, in vivo mpMRI data was co-registered with ex vivo histology using a 3D deformable registration framework. Voxels within the prostate contour were extracted as independent samples, with mpMRI data as features and ground truth from histology as labels. This provides the material to developed specific predictive models with suitable machine learning algorithms.

Results: A multi-purpose radiomics framework was developed and applied in four different settings, including predictions for tumour location, cell density, tumour aggressiveness and hypoxia. Tumour location prediction achieved an area under the receiver operating characteristics curve (AUC of ROC) ranging from 0.81 to 0.94. A predictive model for cell density was also developed and gave a root mean square error of 1.06 x 103 cells/mm2 (relative error of 13.25%). Results for predicting tumour aggressiveness achieved an AUC of 0.91. Finally, radiogenomics analysis of prostate hypoxia revealed a selection of 16 imaging features which showed significant correlations with hypoxia-related gene expressions.

Conclusion: We present a framework for a contemporary approach to prostate focal BT that incorporates mpMRI and a biological optimisation approach to treatment planning.

Keywords

Brachytherapy, Image Analysis, Radiobiology

Taxonomy

TH- Brachytherapy: prostate brachytherapy

Contact Email