Meeting Banner
Abstract #0547

Diagnostic Modelling of Multi-Parametric MRI as a Radiological Tool to Predict Transition Zone Prostate Cancer

Nikolaos Dikaios1, Taiki Fujiwara2, Mohamed Abd Alazeez3, David Atkinson4, Shonit Punwani2

1Department of Medical Physics and Bioengineering, University College London, London, Greater London, United Kingdom; 2Department of Radiology, University College London Hospital; 3Department of Urology, University College London Hospital; 4Centre for Medical Imaging, University College London

Multi-parametric MRI has a reported sensitivity of 73% and specificity of 89% for detection of tumour within the peripheral zone of the prostate. However, benign prostatic hypertrophy as commonly found in the transition zone produces signal changes that make the radiologists detection of anterior gland tumour more difficult. Our study derived a predictive model for classification of suspected sites of anterior gland disease based on clinical (age, PSA, gland volume, PSA density), quantitative MRI parameters (ADC, contrast enhanced, T2 image signal) and textural features of MR images (entropy, contrast, co-occurance); and compared the performance of this model for detection of tumour against a consensus radiologist opinion.

Abstract PDF Presentation Video

Keywords

ability able almost anterior antigen apparent application applied approximately arterial ascribing assessed assigned assignment axial bioengineering biopsy black bootstrapped bootstrapping built cancer cancers characteristic circular classification classified classify clinical clinically coefficient cohort college commonly component consensus contain containing contoured contrast contribute contribution correctly correlation cross curve curves definite density derived despite detection diagnosed diagnosis diagnostic diameter difficulty diffusion discriminant disease distribution enhanced entropy experienced fact features gland grade graded grades greater histogram hospital identified illustrates included indeed independent indeterminate individual kingdom largest leave likelihood likely linear local localization location mapping matched matrix medical model months negative operator opinion overall parametric patient patients percentage performance peripheral physics planned population positive predictive presence produces prospectively prostate quantitative radiologist radiologists radiology receiver reported reporting revealed samples sensitivity separated significance site sites specificity staging status subsequent subsequently suggest summary suspected table template textural third tool transition unaware uncertain urology validation variables volume year years zone