Objective risk stratification of prostate cancer using machine learning and radiomics applied to multiparametric magnetic resonance images

Bino Varghese,Suzanne L Palmer,Frank Chen,Gaurav Pandey,Darryl Hwang,Andre Luis De Castro Abreu,Vinay Duddalwar,Manju Aron,Osamu Ukimura,Monish Aron,Inderbir Gill

doi:10.1038/s41598-018-38381-x

Abstract

Multiparametric magnetic resonance imaging (mpMRI) has become increasingly important for the clinical assessment of prostate cancer (PCa), but its interpretation is generally variable due to its relatively subjective nature. Radiomics and classification methods have shown potential for improving the accuracy and objectivity of mpMRI-based PCa assessment. However, these studies are limited to a small number of classification methods, evaluation using the AUC score only, and a non-rigorous assessment of all possible combinations of radiomics and classification methods. This paper presents a systematic and rigorous framework comprised of classification, cross-validation and statistical analyses that was developed to identify the best performing classifier for PCa risk stratification based on mpMRI-derived radiomic features derived from a sizeable cohort. This classifier performed well in an independent validation set, including performing better than PI-RADS v2 in some aspects, indicating the value of objectively interpreting mpMRI images using radiomics and classification methods for PCa risk assessment.

Highlights

Prostate cancer (PCa) is the third most common cause of death and the most prevalent male malignancy worldwide[1]
This study presents a systematic and rigorous Machine Learning (ML)-based framework comprised of classification[30], cross-validation[31] and statistical analyses[32] designed to identify the best performing classifier for prostate cancer (PCa) risk stratification based on multi-parametric magnetic resonance imaging (mpMRI)-derived radiomic features derived from a sizeable cohort
We present the results of the application of our ML-based framework and its components on radiomics features derived from our various study cohorts

Summary

Introduction

Prostate cancer (PCa) is the third most common cause of death and the most prevalent male malignancy worldwide[1]. T2W and DW (including apparent diffusion coefficient [ADC] map) signal intensities have been shown to correlate with histopathology-based nuclear cell-density[9], as well as PCa aggressiveness[10,11]. The underlying assumption is that images collected during routine clinical care contain latent information regarding tumor behavior that can be extracted using a variety of quantitative image characterization algorithms[14]. The extraction of these radiomic features enables the conversion of collections of digital clinical images into structured quantitative data that can help model tumor behavior. Lv et al used fractal-based features to distinguish prostate tumor tissue from normal peripheral zone tissue[18]

Methods

Results

Conclusion