The Technome - A Predictive Internal Calibration Approach for Quantitative Imaging Biomarker Research

Alexander Mühlberg,Martine Rémy-Jardin,Oliver Taubmann,Julian Holch,Rainer Kärgel,Michael Sühling,Stefan Maurus,Alexander Katzmann,Thomas Huber,Jean-Baptiste Faivre,Dominik Nörenberg,Félix Lades,Volker Heinemann,Michael Wels

doi:10.1038/s41598-019-57325-7

Abstract

The goal of radiomics is to convert medical images into a minable data space by extraction of quantitative imaging features for clinically relevant analyses, e.g. survival time prediction of a patient. One problem of radiomics from computed tomography is the impact of technical variation such as reconstruction kernel variation within a study. Additionally, what is often neglected is the impact of inter-patient technical variation, resulting from patient characteristics, even when scan and reconstruction parameters are constant. In our approach, measurements within 3D regions-of-interests (ROI) are calibrated by further ROIs such as air, adipose tissue, liver, etc. that are used as control regions (CR). Our goal is to derive general rules for an automated internal calibration that enhance prediction, based on the analysed features and a set of CRs. We define qualification criteria motivated by status-quo radiomics stability analysis techniques to only collect information from the CRs which is relevant given a respective task. These criteria are used in an optimisation to automatically derive a suitable internal calibration for prediction tasks based on the CRs. Our calibration enhanced the performance for centrilobular emphysema prediction in a COPD study and prediction of patients’ one-year-survival in an oncological study.

Highlights

The goal of radiomics is to convert medical images into a minable data space by extraction of quantitative imaging features for clinically relevant analyses, e.g. survival time prediction of a patient
While scientists in the field of image acquisition and image reconstruction deal with such non-reducible technical variation decreasing its qualitative influence with advanced techniques, it is not known which impact non-reducible technical variation has on derived quantitative features, e.g. radiomics, in relation to the examined biological or pathological variation
On computed tomography (CT) images, all measurements should be seen as relative – a notion that is already accepted for bone mineral density measurements that are calibrated by in-scan phantoms as seen, e.g., in the work by Kalender[36]

Summary

Introduction

The goal of radiomics is to convert medical images into a minable data space by extraction of quantitative imaging features for clinically relevant analyses, e.g. survival time prediction of a patient. In CT, a corpulent patient with a larger cross-section will usually have a higher noise level within his body than a slimmer patient This effect appears despite constant scan parameters as less quanta arrive at the detector. Another example is beam hardening that is stronger if the cross-section of the patient is larger: photons in the center have a higher average energy than in the periphery. The effects are estimated by an empirical Bayes fit and removed from the feature by subtraction and division These are statistical methods, i.e. a calibration is learned on the same data as it is applied and their main focus is the stabilisation of intensity distributions or features with regards to technical variation

Objectives

Methods

Results

Conclusion