Radiomics and “radi-…omics” in cancer immunotherapy: a guide for clinicians
Introduction
In recent years the concept of precision medicine has become an area of immense interest (Burki, 2017; Ashley, 2016) in medical oncology (Le Tourneau et al., 2018). In this field, emerging mechanisms of resistance (Khan and Spicer, 2019; Lim and Ma, 2019; Narayanan et al., 2020; Kalbasi and Ribas, 2020), financial costs and toxicities (Postow et al., 2018; Porcu et al., 2020; Solinas et al., 2018; Marin-Acevedo et al., 2018; Porcu et al., 2019) are important elements to take into account for guiding treatment decisions, in order to identify ideal candidates that might benefit from a variety of novel and expensive therapies. Besides the identification of new molecular prognostic and predictive biomarkers (Dumitrescu, 2018) and the development of new targeted (Dugger et al., 2018) and immunotherapeutic drugs (Hegde and Chen, 2020; Dobosz and Dzieciątkowski, 2019), imaging has started to play a pivotal role in the evolution towards precision medicine (Acharya et al., 2018).
A fruitful collaboration between radiologists and medical oncologists is essential due to the growing dependency on imaging as a therapeutic biomarker. This is particularly relevant in this new era of cancer immunotherapy with immune checkpoint blockade (ICB) targeting a variety of immune checkpoint molecules that physiologically modulate the immune response (Solinas et al., 2019a; Solinas et al., 2019b; Solinas et al., 2019c; Solinas et al., 2020; ElTanbouly et al., 2020; Rowshanravan et al., 2018). These new drug regimens have a novel mechanism of action. Instead of directly killing tumor cells, they are able to harness the patient’s own immune response against cancer, making unique and challenging the assessment of treatment response (Litière et al., 2017; Seymour et al., 2017; Solinas et al., 2017; Porcu et al., 2018). This requires a deep knowledge for the specific features that characterize immune-related phenomena in each organ site. The contribution of imaging is also important for differential diagnosis when facing with immune related adverse events (irAEs) (Porcu et al., 2020; Solinas et al., 2018; Porcu et al., 2019).
Radiological sciences are thus moving from a dependency on basic visual features in medical images to new non-visual features buried in the pixel data that are not routinely detectable with the human eye (Gillies et al., 2016). This is obtained thanks to the improvement of computational power of informatic systems and to the use of artificial intelligence (AI) (Obermeyer and Emanuel, 2016), specifically machine learning techniques (Choy et al., 2018) such as deep learning (LeCun et al., 2015).
With their ability to combine quantitative data obtained from imaging together with those derived from genomic analyses, “radiomics” and “radiogenomics” metadata are becoming of common use in clinical research. Hopefully, they will soon be part of daily routine clinical practice with the aim to allow oncologists and radiologists to optimize patient selection and management for medical therapy and to better assess treatment responses (Lambin et al., 2012).
The aims of this narrative review are to provide a simplified guide for clinicians to these new concepts, including some important technical aspects, and to summarize the existing evidence on radiomics in cancer immunotherapy.
Section snippets
Definitions
“Radiomics” is defined as the analysis of imaging data through the use of specific algorithms aimed at identifying quantitative features otherwise not identifiable with the simple visual analysis, in order to create enhanced data models for improving medical decision support (Gillies et al., 2016; Lambin et al., 2012). Images from ultrasound (US), computed tomography (CT), magnetic resonance (MR) and positron emission tomography (PET) combined with CT (PET-CT) or MR (PET-MR) can be analyzed
Feature identification in radiomics
Radiomics is based on the discovery of imaging features not identifiable with the simple visual analysis, through the use of specific and sophisticated algorithms (Gillies et al., 2016; Lambin et al., 2012). From a technical point of view, the whole process can be divided into three steps (Gillies et al., 2016; Mazurowski, 2015) (Fig. 1):
- 1)
Image acquisition, database creation and data selection: all imaging examination can be “radiomically” analyzed, including MR, CT, US and PET (Gillies et al.,
The model development in radiomics and “radi…-omics”
Once radiomic and non-radiomic features (such as those derived from histologic, genomic, proteomic or transcriptomic analyses) are collected, they can be used for data mining, representing the process of discovering patterns in datasets (Gillies et al., 2016; Incoronato et al., 2017). By exploring the mutual relationships between radiomic and non-radiomic features with a selected outcome variable, AI modeling can be used to develop models that, once validated, could be used for different
Radiomics and “radi…-omics” in cancer immunotherapy
Radiomics could be a helpful tool in cancer immunotherapy for optimal patient selection (El Naqa I, Ten Haken RK, 2018). It may act as a non-invasive digital biopsy technique able to quantify tumor T-cell infiltration, to support personalized immunotherapy interventions, and to longitudinally monitor the therapeutic response above the traditional dimensional criteria (Solinas et al., 2017; Porcu et al., 2018). In particular, digital biopsy thorugh the combination of radiomics and pathomics
Current situation and future prospects
Radiomics is changing radiology and oncology (Ha, 2019; Trebeschi et al., 2019; Bera et al., 2018; Du et al., 2019; Banna et al., 2019b), but the process of translation from pure research to clinical practice is facing some difficulties (Park et al., 2020; Deutsch and Paragios, 2019). In a recent article by Sollini et al (Choi et al., 2016), authors adopted a phase classification criteria for radiomic studies similar to the one used for drug development, and the relative paucity of phase III
Conclusions
Radiomics and “radi…-omics” techniques leverage imaging information beyond those attained through visual inspection. In the field of cancer immunotherapy, radiomics has already shown value in characterizing the tumor immune-environment and to stratify prognosis by easily evaluating the baseline imaging investigations.
Even though the process of translation of these models from pure research to clinical practice is still in progress, they represent an innovative tool that once applied in clinical
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or non-profit sectors.
Declaration of Competing Interest
Matteo Lambertini acted as consultant for Roche and Lilly, and received speaker honoraria from Roche, Takeda and Theramex outside the submitted work. The other authors declare that they have no conflict of interests.
References (124)
- et al.
Radiogenomics - current status, challenges and future directions
Cancer Lett.
(2016) Defining precision medicine
Lancet Oncol.
(2017)- et al.
Radiomics to predict response to immunotherapy, bridging the gap from proof of concept to clinical applicability?
Ann Oncol.
(2019) - et al.
Noninvasive imaging in cancer immunotherapy: The way to precision medicine
Cancer Lett.
(2019) - et al.
New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)
Eur J Cancer.
(2009) - et al.
Radiomics-based features for pattern recognition of lung cancer histopathology and metastases
Comput Methods Programs Biomed.
(2018) Texture analysis using gray level run lengths
Comput. Graph. Mage Process.
(1975)- et al.
Top 10 Challenges in Cancer Immunotherapy
Immunity
(2020) - et al.
Radiomics: extracting more information from medical images using advanced feature analysis
Eur J Cancer.
(2012) - et al.
Fractal and multifractal analysis: a review
Med Image Anal.
(2009)
Radiogenomics: what it is and why it is important
J Am Coll Radiol.
Basic principles of ROC analysis
Semin Nucl Med.
Targeting the ubiquitin-proteasome pathway to overcome anti-cancer drug resistance
Drug Resist Updat.
Radiological evaluation of response to immunotherapy in brain tumors: Where are we now and where are we going?
Crit Rev Oncol Hematol.
CTLA-4: a moving target in immunotherapy
Blood
The present and future of deep learning in radiology
Eur J Radiol.
iRECIST: guidelines for response criteria for use in trials testing immunotherapeutics [published correction appears in Lancet Oncol. 2019 May;20(5):e242]
Lancet Oncol.
Towards precision medicine: from quantitative imaging to radiomics
J Zhejiang Univ Sci B.
Defining a Radiomic Response Phenotype: A Pilot Study using targeted therapy in NSCLC [published correction appears in Sci Rep. 2017 Feb 17;7:41197]
Sci Rep.
Imaging Biomarkers and Imaging Biobanks
Textural features corresponding to textural properties
IEEE Trans. Syst. Man Cybern.
Towards precision medicine
Nat Rev Genet.
Proteomics: Technologies and Their Applications
J Chromatogr Sci.
Imaging genomics in cancer research: limitations and promises
Br J Radiol.
The Promise of Digital Biopsy for the Prediction of Tumor Molecular Features and Clinical Outcomes Associated With Immunotherapy
Front Med (Lausanne).
The Promise of Digital Biopsy for the Prediction of Tumor Molecular Features and Clinical Outcomes Associated With Immunotherapy
Front Med (Lausanne).
Novel Quantitative Imaging for Predicting Response to Therapy: Techniques and Clinical Applications
Am Soc Clin Oncol Educ Book.
MRI radiomic features are associated with survival in melanoma brain metastases treated with immune checkpoint inhibitors
Neuro Oncol.
CT radiomics and PET radiomics: ready for clinical implementation?
Q J Nucl Med Mol Imaging.
Group. A collaborative enterprise for multi-stakeholder participation in the advancement of quantitative imaging
Radiology.
Radiomics Evaluation of Histological Heterogeneity Using Multiscale Textures Derived From 3D Wavelet Transformation of Multispectral Images
Front Oncol.
An investigation of machine learning methods in delta-radiomics feature analysis
PLoS One.
Pretreatment prediction of immunoscore in hepatocellular cancer: a radiomics-based clinical model based on Gd-EOB-DTPA-enhanced MRI imaging
Eur Radiol.
Does the Reporting Quality of Diagnostic Test Accuracy Studies, as Defined by STARD 2015, Affect Citation?
Korean J Radiol.
Current Applications and Future Impact of Machine Learning in Radiology
Radiology.
Radiomics to predict immunotherapy-induced pneumonitis: proof of concept
Invest New Drugs.
Quasi-experimentation: design & analysis issues for field settings
18F-FDG PET Dissemination Features in Diffuse Large B-Cell Lymphoma Are Predictive of Outcome
J Nucl Med.
Fractal-based radiomic approach to predict complete pathological response after chemo-radiotherapy in rectal cancer
Radiol Med.
Assessment of tumor heterogeneity: an emerging imaging tool for clinical practice?
Insights Imaging.
Artificial neural networks: opening the black box
Cancer
Machine Learning in Medicine
Circulation.
Radiogenomics and Radioproteomics
OMICS J Radiology.
The Intriguing History of Cancer Immunotherapy
Front Immunol.
Drug development in the era of precision medicine
Nat Rev Drug Discov.
Early Epigenetic Markers for Precision Medicine
Methods Mol Biol.
Can radiomics personalise immunotherapy?
Lancet Oncol.
VISTA: Coming of age as a multi-lineage immune checkpoint [published online ahead of print, 2020 Jan 13]
Clin Exp Immunol.
Delta-radiomics features for the prediction of patient outcomes in non-small cell lung cancer
Sci Rep.
External validity, generalizability, and knowledge utilization
J Nurs Scholarsh.
Cited by (24)
Peeling the I-Onion to Demystify Pseudoprogression
2023, Cancer CellArtificial intelligence-assisted selection and efficacy prediction of antineoplastic strategies for precision cancer therapy
2023, Seminars in Cancer BiologyRadiomics signature based on CECT for non-invasive prediction of response to anti-PD-1 therapy in patients with hepatocellular carcinoma
2023, Clinical RadiologyCitation Excerpt :The radiomics features were extracted lesion-per-lesion. Electronic Supplementary Material S2 describes the detailed process of feature extraction.24 Lesion-level progressive disease (PD) was defined as an increase in the diameter of the target lesion by at least 20% compared with the baseline value according to the enhanced imaging during the arterial phase (Fig 2).
When artificial intelligence meets PD-1/PD-L1 inhibitors: Population screening, response prediction and efficacy evaluation
2022, Computers in Biology and MedicineCitation Excerpt :Big data generated in clinical oncology, which may come from a variety of sources such as molecular, imaging, pharmacological, and demographic information, presents clinicians and data scientists with challenges for analysis while also provides opportunities for knowledge discovery in tumor biology and therapy [8]. Clinical oncology research is increasingly focusing on the full exploitation of clinical big data and multi-omics research represented by radiomics studies, which occupies an important aspect of prognostic analysis of immunotherapy [9]. As an important derivative of Precision Medicine in immunotherapy, medical AI in omics research allows machine learning to play an important role in therapy discovery and response prediction [10], on the basis of which numerous exciting results have been generated, among which the application of AI in anti-PD-1/PD-L1 therapy is an essential and representative aspect (Fig. 1-(B)).
- 1
Contributed equally: co-first authors.