Learning objectives
We provide a practical approach for radiologists to undertake radiomic and radiogenomic studies.
The process outlines the preliminary steps of: selection of imaging data,
pseudonymisation,
normalisation,
registration and tumour segmentation.
This is followed by extraction of quantitative intensity,
texture,
shape and wavelet-based features.
Machine-learning analysis of radiomic,
genomic,
clinical data and multiparametric MRI data is performed,
involving feature selection and classification.
The trained classifier model then requires testing on an independent dataset for validation.1
Background
Phenotypic information is routinely being extracted through imaging non-invasively which can be used for precision medicine.2 It is critically important that radiologists lead this artificial intelligence (AI) revolution.
They have developed clinical skills and experience through generations of accumulated knowledge.
Clinically useful,
predictive radiomic properties must be clearly linked to meaningful biologic characteristics and qualitative imaging properties familiar to radiologists.
Radiomics and radiogenomics incorporates several important disciplines,
including radiology (imaging interpretation),
computer vision (quantitative feature extraction) and machine learning (classifier evaluation).3
Radiomics has shown the...
Findings and procedure details
Identifying a Clinical Question and the Research Team
Selection of the patient population and clinical question should be explored between clinical teams and radiologists to ensure appropriate and clinically focused research studies are undertaken.
The research should address an unmet clinical need that aims to bring direct patient benefit.
A multidisciplinary team approach to research utilising skills of radiologists,
clinicians,
physicists,
computer scientists,
data scientists and statisticians is essential and should be identified prior to commencing the study.
Radiologists have the unique position of having...
Conclusion
Through the convergence of radiology,
computer vision and machine learning techniques,
radiomics provides a mechanism for a multidisciplinary approach in imaging.
When radiomic models align well with disease biology,
then only radiomic findings maximise their likelihood for clinical utility.
Without this,
there is the risk of drowning into the plethora of clinically unsupervised informatics.
This newly developing field should form a part of the Radiology Training Programmes.
References
Erickson BJ,
Korfiatis P,
Akkus Z,
Kline TL.
Machine Learning for Medical Imaging.
Radiographics.
2017;37(2):505-515.
Gillies RJ,
Kinahan PE,
Hricak H.
Radiomics: Images Are More than Pictures,
They Are Data.
Radiology.
2016;278(2):563-77.
Zhou M,
Scott J,
Chaudhury B,
et al.
Radiomics in Brain Tumor: Image Assessment,
Quantitative Feature Descriptors,
and Machine-Learning Approaches.
AJNR Am J Neuroradiol.
2017.
Narang,
S.,
Lehrer,
M.,
Yang,
D.,
Lee,
J.,
& Rao,
A.
(2016).
Radiomics in glioblastoma: current status,
challenges and potential opportunities.
Translational Cancer Research,
5(4),
383-397.
doi:10.21037/8806
Avanzo...