Radiomics and artificial intelligence for soft-tissue sarcomas: Current status and perspectives

This article proposes a summary of the current status of the research regarding the use of radiomics and artificial intelligence to improve the radiological assessment of patients with soft tissue sarcomas (STS), a heterogeneous group of rare and ubiquitous mesenchymal malignancies. After a first part explaining the principle of radiomics approaches, from raw image post-processing to extraction of radiomics features mined with unsupervised and supervised machine-learning algorithms, and the current research involving deep learning algorithms in STS, especially convolutional neural networks, this review details their main research developments since the formalisation of 'radiomics' in oncologic imaging in 2010. This review focuses on CT and MRI and does not involve ultrasonography. Radiomics and deep radiomics have been successfully applied to develop predictive models to discriminate between benign soft-tissue tumors and STS, to predict the histologic grade (i.e., the most important prognostic marker of STS), the response to neoadjuvant chemotherapy and/or radiotherapy, and the patients' survivals and probability for presenting distant metastases. The main findings, limitations and expectations are discussed for each of these outcomes. Overall, after a first decade of publications emphasizing the potential of radiomics through retrospective proof-of-concept studies, almost all positive but with heterogeneous and often non-replicable methods, radiomics is now at a turning point in order to provide robust demonstrations of its clinical impact through open-science, independent databases, and application of good and standardized practices in radiomics such as those provided by the Image Biomarker Standardization Initiative, without forgetting innovative research paths involving other '-omics' data to better understand the relationships between imaging of STS, gene-expression profiles and tumor microenvironment.

Impact of Vaccination and the Omicron Variant on COVID-19-related Chest CT Findings: A Multicenter Study

Background The SARS-CoV-2 Omicron variant has a higher infection rate than previous variants but results in less severe disease. However, the effects of Omicron and vaccination on chest CT findings are difficult to evaluate. Purpose To investigate the effect of vaccination status and predominant variant on chest CT findings, diagnostic scores, and severity scores in a multicenter sample of consecutive patients referred to emergency departments for proven COVID-19. Materials and Methods This retrospective multicenter study included adults referred to 93 emergency departments with SARS-CoV-2 infection according to a reverse-transcriptase polymerase chain reaction test and known vaccination status between July 2021 and March 2022. Clinical data and structured chest CT reports, including semiquantitative diagnostic and severity scores following the French Society of Radiology-Thoracic Imaging Society guidelines, were extracted from a teleradiology database. Observations were divided into Delta-predominant, transition, and Omicron-predominant periods. Associations between scores and variant and vaccination status were investigated with χ² tests and ordinal regressions. Multivariable analyses evaluated the influence of Omicron variant and vaccination status on the diagnostic and severity scores. Results Overall, 3876 patients were included (median age, 68 years [quartile 1 to quartile 3 range, 54-80]; 1695 women). Diagnostic and severity scores were associated with the predominant variant (Delta vs Omicron, χ² = 112.4 and 33.7, respectively; both P < .001) and vaccination status (χ² = 243.6 and 210.1; both P < .001) and their interaction (χ² = 4.3 [P = .04] and 28.7 [P < .001], respectively). In multivariable analyses, Omicron variant was associated with lower odds of typical CT findings than was Delta variant (odds ratio [OR], 0.46; P < .001). Two and three vaccine doses were associated with lower odds of demonstrating typical CT findings (OR, 0.32 and 0.20, respectively; both P < .001) and of having high severity score (OR, 0.47 and 0.33, respectively; both P < .001), compared with unvaccinated patients. Conclusion Both the Omicron variant and vaccination were associated with less typical chest CT manifestations of COVID-19 and lesser extent of disease. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Yoon and Goo in this issue.

Impact of Vaccination and the Omicron Variant on COVID-19-related Chest CT Findings: A Multicenter Study

Background Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) omicron variant has a higher infection rate than previous variants but results in less severe disease. However, the impacts of omicron and vaccination on chest CT findings are difficult to evaluate. Purpose To investigate the impact of vaccination status and predominant variant on chest CT findings, diagnostic and severity scores in multicenter sample of consecutive patients referred to emergency departments for proven COVID-19. Materials and Methods This retrospective, multicenter study included adults referred to 93 emergency departments with SARS-CoV-2 infection according to RT-PCR and known vaccination status between July 2021 and March 2022. Clinical data and structured chest CT reports including semiquantitative diagnostic and severity scores following the French Society of Radiology-Thoracic Imaging Society guidelines were extracted from a teleradiology database. Observations were divided into 'delta-predominant', 'transition', and 'omicron-predominant' periods. Associations between scores and variant and vaccination status were investigated with Chi-square tests and ordinal regressions. Multivariable analyses evaluated the influence of omicron variant and vaccination status on the diagnostic and severity scores. Results Overall, 3876 patients were included (median age: 68 years [Q1-Q3: 54-80], 1695 females). Diagnostic and severity scores were associated with the predominant variant (delta- versus omicron-predominant, Chi-square=112.4 and 33.7, both P<.001) and vaccination (Chi-square=243.6 and 210, both P<.001) and their interaction (Chi-square=4.3, P=.04 and Chi-square=28.7, P<.001, respectively). In multivariable analyses, omicron variant was associated with lower odds of typical CT findings than delta variant (OR=0.46, P<.001). Two and three vaccine doses were associated with lower odds of demonstrating typical CT findings (OR=0.32 and OR=0.20, both P<.001), and of having high severity score (OR=0.47 and OR=0.33, both P<.001), compared with unvaccinated patients. Conclusion Both the omicron variant and vaccination were associated with less typical chest CT manifestations for COVID-19 and lesser extent of disease. See also the editorial by Yoon and Goo in this issue.

Role of Imaging in Initial Prognostication of Locally Advanced Soft Tissue Sarcomas

BACKGROUND

Although imaging is central in the initial staging of patients with soft tissue sarcomas (STS), it remains underused and few radiological features are currently used in practice for prognostication and to help guide the best therapeutic strategy. Yet, several prognostic qualitative and quantitative characteristics from magnetic resonance imaging (MRI) and positron emission tomography (PET) have been identified over these last decades.

OBJECTIVE

After an overview of the current validated prognostic features based on baseline imaging and their integration into prognostic tools, such as nomograms used by clinicians, the aim of this review is to summarize more complex and innovative MRI, PET, and radiomics features, and to highlight their role to predict indirectly (through histologic grade) or directly the patients' outcomes.

Distinct patterns of the natural evolution of soft tissue sarcomas on pre-treatment MRIs captured with delta-radiomics correlate with gene expression profiles

OBJECTIVES

Radiomics of soft tissue sarcomas (STS) is assumed to correlate with histologic and molecular tumor features, but radiogenomics analyses are lacking. Our aim was to identify if distinct patterns of natural evolution of STS obtained from consecutive pre-treatment MRIs are associated with differential gene expression (DGE) profiling in a pathway analysis.

METHODS

All patients with newly diagnosed STS treated in a curative intent in our sarcoma reference center between 2008 and 2019 and with two available pre-treatment contrast-enhanced MRIs were included in this retrospective study. Radiomics features (RFs) were extracted from fat-sat contrast-enhanced T1-weighted imaging. Log ratio and relative change in RFs were calculated and used to determine grouping of samples based on a consensus hierarchical clustering. DGE and oncogenesis pathway analysis were performed in the delta-radiomics groups identified in order to detect associations between delta-radiomics patterns and transcriptomics features of STS. Secondarily, the prognostic value of the delta-radiomics groups was investigated.

RESULTS

Sixty-three patients were included (median age: 63 years, interquartile range: 52.5-70). The consensus clustering identified 3 reliable delta-radiomics patient groups (A, B, and C). On imaging, group B patients were characterized by increase in tumor heterogeneity, necrotic signal, infiltrative margins, peritumoral edema, and peritumoral enhancement before the treatment start (p value range: 0.0019-0.0244), and, molecularly, by downregulation of natural killer cell-mediated cytotoxicity genes and upregulation of Hedgehog and Hippo signaling pathways. Group A patients were characterized by morphological stability of pre-treatment MRI traits and no local relapse (log-rank p = 0.0277).

CONCLUSIONS

This study highlights radiomics and transcriptomics convergence in STS. Proliferation and immune response inhibition were hyper-activated in the STS that were the most evolving on consecutive imaging.

KEY POINTS

• Three consensual and stable delta-radiomics clusters were identified and captured the natural patterns of morphological evolution of STS on pre-treatment MRIs. • These 3 patterns were explainable and correlated with different well-known semantic radiological features with an ascending gradient of pejorative characteristics from the A group to C group to B group. • Gene expression profiling stressed distinct patterns of up/downregulated oncogenetic pathways in STS from B group in keeping with its most aggressive radiological evolution.

Soft-tissue sarcoma in adults: Imaging appearances, pitfalls and diagnostic algorithms

This article provides an overview of the current knowledge regarding diagnostic imaging of patients with soft-tissue sarcomas, which is a heterogeneous group of rare mesenchymal malignancies. After an initial contextualization, diagnostic flow-chart based on initial radiological findings of soft-tissue masses (with specific focus on adipocytic soft-tissue tumors [STTs], hemorragic STTs and retroperitoneal STTs) are provided considering relevant results from novel researches, guidelines, and experts' viewpoints, with the aim to help radiologists and clinicians in their practice. Particularly, the central place of sarcoma reference centers in the diagnostic and therapeutic management is highlighted, as well as the pivotal role that radiologists should play to correctly identify patients with soft-tissue sarcoma at the initial stage of the disease. Indications and methods for performing imaging-guided biopsies are also discussed, as well as clues to improve soft-tissue sarcoma grading with conventional and quantitative imaging.

Natural speed of growth of untreated soft-tissue sarcomas: A dimension-based imaging analysis

PURPOSE

The interval from first symptoms to diagnosis, staging and referral to reference center can last months for soft-tissue sarcoma (STS) patients. Meanwhile, patients can undergo different imaging that capture the 'natural' tumor changes, before medical intervention. Aim was to depict these 'natural' dimensional variations and to correlate them with patients' outcome.

METHODS

Single-center retrospective study including all consecutive adults with newly-diagnosed STS, ≥2 pre-treatment imaging (CT-scan or MRI) on the tumor (Exam-0 and Exam-1), and managed in reference center between 2007 and 2018. Longest diameter (LD) and volume were calculated on both examinations to obtain the naïve dimensional growth before any intervention. SARCULATOR nomogram was applied on data at Exam-0 and Exam-1. Correlations with overall, metastatic and local relapse-free survivals (OS, MFS and LFS), and with pre-treatment pathological features were performed.

RESULTS

137 patients were included (median age: 65 years). Average naïve growth was +39.4% in LD and +503% in volume during an average Exam-0-to-Exam-1 interval of 130 days. The 10-year distant metastasis and OS predictions were different at Exam-0 and Exam-1 (P < 0.0001 for both). All the changes in radiological measurements significantly correlated with pre-treatment number of mitosis, grade and complex genomic (P-value range: <0.0001-0.0481). Multivariate Cox modeling identified the relative change in LD/month and absolute change in LD/month as independent predictors for OS and LFS, respectively (P = 0.0003 and 0.0001, respectively).

CONCLUSION

When available, the natural speed of growth on pre-treatment imaging should be evaluated to improve the estimation of pre-treatment histological grade and patients' OS and LFS.

Intensity harmonization techniques influence radiomics features and radiomics-based predictions in sarcoma patients

Intensity harmonization techniques (IHT) are mandatory to homogenize multicentric MRIs before any quantitative analysis because signal intensities (SI) do not have standardized units. Radiomics combine quantification of tumors' radiological phenotype with machine-learning to improve predictive models, such as metastastic-relapse-free survival (MFS) for sarcoma patients. We post-processed the initial T2-weighted-imaging of 70 sarcoma patients by using 5 IHTs and extracting 45 radiomics features (RFs), namely: classical standardization (IHT_std), standardization per adipose tissue SIs (IHT_fat), histogram-matching with a patient histogram (IHT_HM.1), with the average histogram of the population (IHT_HM.All) and plus ComBat method (IHT_HM.All.C), which provided 5 radiomics datasets in addition to the original radiomics dataset without IHT (No-IHT). We found that using IHTs significantly influenced all RFs values (p-values: < 0.0001-0.02). Unsupervised clustering performed on each radiomics dataset showed that only clusters from the No-IHT, IHT_std, IHT_HM.All, and IHTHM.All.C datasets significantly correlated with MFS in multivariate Cox models (p = 0.02, 0.007, 0.004 and 0.02, respectively). We built radiomics-based supervised models to predict metastatic relapse at 2-years with a training set of 50 patients. The models performances varied markedly depending on the IHT in the validation set (range of AUROC from 0.688 with IHT_std to 0.823 with IHT_HM.1). Hence, the use of intensity harmonization and the related technique should be carefully detailed in radiomics post-processing pipelines as it can profoundly affect the reproducibility of analyses.

Systematic review of sarcomas radiomics studies: Bridging the gap between concepts and clinical applications?

OBJECTIVES

Sarcomas are a model for intra- and inter-tumoral heterogeneities making them particularly suitable for radiomics analyses. Our purposes were to review the aims, methods and results of radiomics studies involving sarcomas METHODS: Pubmed and Web of Sciences databases were searched for radiomics or textural studies involving bone, soft-tissues and visceral sarcomas until June 2020. Two radiologists evaluated their objectives, results and quality of their methods, imaging pre-processing and machine-learning workflow helped by the items of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2), Image Biomarker Standardization Initiative (IBSI) and 'Radiomics Quality Score' (RQS). Statistical analyses included inter-reader agreements, correlations between methodological assessments, scientometrics indices, and their changes over years, and between RQS, number of patients and models performance.

RESULTS

Fifty-two studies were included involving: soft-tissue sarcomas (29/52, 55.8 %), bone sarcomas (15/52, 28.8 %), gynecological sarcomas (6/52, 11.5 %) and mixed sarcomas (2/52, 3.8 %), mostly imaged with MRI (36/52, 69.2 %), for a total of distinct patients. Median RQS was 4.5 (28.4 % of the maximum, range: -7 - 17). Performances of predictive models and number of patients negatively correlated (p = 0.027). None of the studies detailed all the items from the IBSI guidelines. There was a significant increase in studies' impact factors since the establishing of the RQS in 2017 (p = 0.038).

CONCLUSION

Although showing promising results, further efforts are needed to make sarcoma radiomics studies reproducible with an acceptable level of evidence. A better knowledge of the RQS and IBSI reporting guidelines could improve the quality of sarcoma radiomics studies and accelerate clinical applications.

High-Grade Soft-Tissue Sarcomas: Can Optimizing Dynamic Contrast-Enhanced MRI Postprocessing Improve Prognostic Radiomics Models?

BACKGROUND

Heterogeneity on pretreatment dynamic contrast-enhanced (DCE)-MRI of sarcomas may be prognostic, but the best technique to capture this characteristic remains unknown.

PURPOSE

To investigate the best method to extract prognostic data from baseline DCE-MRI.

STUDY TYPE

Retrospective, single-center.

POPULATION

Fifty consecutive uniformly-treated adults with nonmetastatic high-grade sarcomas.

FIELD STRENGTH/SEQUENCE

1.5T; T₂ -weighted-imaging, fat-suppressed fast spoiled gradient echo DCE-MRI.

ASSESSMENT

Ninety-two radiomics features (RFs) were extracted at each DCE-MRI phase (11, from t = 0-88 sec). Relative changes in RFs (rRFs) since the acquisition baseline were calculated (11 × 92 rRFs). Curves of rRF as function of time postinjection were integrated (92 integrated-rRFs [irRFs]). K^trans and area under the time-intensity curve at 88-sec parametric maps were computed and 2 × 92 parametric-RFs (pRFs) were extracted. Five DCE-MRI-based radiomics models were built on: an RFs subset (32 sec, 64 sec, 88 sec); all rRFs; all irRFs; and all pRFs. Two models were elaborated as reference, on: conventional radiological features; and T₂ -WI RFs.

STATISTICAL TESTS

A common machine-learning approach was applied to radiomics models. Features with P < 0.05 at univariate analysis were entered in a LASSO-penalized Cox regression including bootstrapped 10-fold cross-validation. The resulting radiomics scores (RScores) were dichotomized per their median and entered in multivariate Cox models for predicting metastatic relapse-free survival. Models were compared with integrative area under the curve (AUC) and concordance index.

RESULTS

Only dichotomized RScores from models based on rRFs subset, all rRFS and irRFS correlated with prognostic (P = 0.0107-0.0377). The models including all rRFs and irRFs had the highest c-index (0.83), followed by the radiological model. The radiological model had the highest integrative AUC (0.87), followed by models including all rRFs and irRFs. The radiological and full rRFs models were significantly better than the T₂ -based radiomics model (P = 0.02).

DATA CONCLUSION

The initial DCE-MRI of STS contains prognostic information. It seems more relevant to make predictions on rRFs instead of pRFs. Evidence Level: 3 Technical Efficacy: 3 J. Magn. Reson. Imaging 2020;52:282-297.