Meeting Banner
Abstract #1747

Predicting Lesion Subtype and Response to Chemotherapy in Paediatric Wilms Tumours Using ADC Histogram Analysis

Patrick W. Hales1, ystein E. Olsen2, Neil J. Sebire3, Christopher A. Clark1

1Imaging and Biophysics, University College London, London, United Kingdom; 2Radiology Department, Great Ormond Street Hospital, London, United Kingdom; 3Histopathology Department, Great Ormond Street Hospital, London, United Kingdom

We investigated whether properties of the histogram of ADC values within a Wilms tumour, obtained from diffusion weighted imaging (DWI) at presentation, could be used to predict the histologically-determined tumour subtype, and response to chemotherapy. Median, skewness, kurtosis and full width at half maximum of the ADC histogram were used in a multinomial logistic regression model, to discriminate three tumour subtypes (stromal, blastemal, mixed), with 91% accuracy. The same parameters showed good predictive power when used in a multiple linear regression model, to predict the tumours response to chemotherapy, assessed by either change in volume, or shift in ADC.

Keywords

abdominal aimed alternatively apparent approximately assessed assessing audience available benefits biophysics breathing cells chemotherapy childhood children clinicians coefficient college common completely composed comprised considered consultant correctly correlated days defined defining definitions derived determine determined diagnosed diagnosis diffuse diffusion distribution distributions drawn either epithelial excess free full funded good grant great greatest hales half histogram histograms histological hospital identified indicative induced initial interval investigate investigated kingdom kurtosis lesion lesions logistic lowest malignancy malignant measured median might mixed model models months multinomial necrosis necrotic negatively oncology operative parameterized patients patterns pediatric percentage positive post potential predict predicted predicting predominantly presence prior produced properties protocol radiology received reduce reduction regression regressive regressors response revealed reviewed risk scientists shrink shrinkage skew skewness slices still street substantial subtype subtypes suggested suggests surgically table target terms therapy tissue tool towards traditionally treated treatment tumor tumors unreliable valuable volume whose width year