Objectives To evaluate feasibility and performance of a radiogenomics model based on ovarian US images predicting germline BRCA1/2 gene status.
Methods This retrospective study included 255 patients who were addressed to germline BRCA1/2 testing and pelvic US documenting normal ovaries. Four imaging feature groups were extracted from each normalized US image with manually segmented regions of interest. Feature selection for univariate analysis was carried out via correlation analysis.
Multivariable analysis for classification of germline BRCA1/2 status was then carried out via logistic regression, support vector machine, ensemble of decision trees and automated machine learning pipelines. Data were split into a training (75%) and a testing (25%) set.
The performance of the models was assessed with respect to negative and positive capability to predict germline BRCA1/2 status and compared with NGS data.
Results The four strategies obtained a similar performance in terms of accuracy on the testing set, varying from 0.54 of logistic regression to 0.64 of the auto-machine learning pipeline. The latter showed also the highest value of specificity on the testing set (0.91) and a negative predictive value of 0.65. Data coming only from the Voluson US machine showed generally higher performances, particularly with the auto-machine learning pipeline (testing set specificity 0.87, negative predictive value 0.73, accuracy value 0.72 and 0.79 on training set).
Conclusions The study shows that a radiogenomics-based model on machine learning techniques is feasible when applied to US images. Future investigations are warranted to make it a reliable screening tool for gBRCA1/2 status.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.