Stability of radiomic features in CT perfusion maps

Bogowicz, Marta; Bundschuh, Ralph A.; Veit-Haibach, Patrick; Hüllner, Martin; Studer, G.; Stieb, Sonja; Glatz, S.; Pruschy, Martin; Gückenberger, Matthias; Tanadini‐Lang, Stephanie

doi:10.1088/1361-6560/61/24/8736

Cited by 48 publications

(49 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Published studies show promising results in NSCLC treatments, yet there are still major challenges and limitations to resolve before they can be translated into reliable clinical application. A number of experimental clinical studies have found that current restrictions of radiomics in its therapeutic application are mainly subjected to (1) image standardization, 29 , 91 (2) image registration, 92 - 94 and (3) data sharing. 5 The on-going research activities focusing on tackling these limitations are elaborated subsequently.…”

Section: Current Research Challenges and Prospectsmentioning

confidence: 99%

Radiomics for Response and Outcome Assessment for Non-Small Cell Lung Cancer

Shi

Yuan

et al. 2018

Technol Cancer Res Treat

View full text Add to dashboard Cite

Routine follow-up visits and radiographic imaging are required for outcome evaluation and tumor recurrence monitoring. Yet more personalized surveillance is required in order to sufficiently address the nature of heterogeneity in nonsmall cell lung cancer and possible recurrences upon completion of treatment. Radiomics, an emerging noninvasive technology using medical imaging analysis and data mining methodology, has been adopted to the area of cancer diagnostics in recent years. Its potential application in response assessment for cancer treatment has also drawn considerable attention. Radiomics seeks to extract a large amount of valuable information from patients’ medical images (both pretreatment and follow-up images) and quantitatively correlate image features with diagnostic and therapeutic outcomes. Radiomics relies on computers to identify and analyze vast amounts of quantitative image features that were previously overlooked, unmanageable, or failed to be identified (and recorded) by human eyes. The research area has been focusing on the predictive accuracy of pretreatment features for outcome and response and the early discovery of signs of tumor response, recurrence, distant metastasis, radiation-induced lung injury, death, and other outcomes, respectively. This review summarized the application of radiomics in response assessments in radiotherapy and chemotherapy for non-small cell lung cancer, including image acquisition/reconstruction, region of interest definition/segmentation, feature extraction, and feature selection and classification. The literature search for references of this article includes PubMed peer-reviewed publications over the last 10 years on the topics of radiomics, textural features, radiotherapy, chemotherapy, lung cancer, and response assessment. Summary tables of radiomics in response assessment and treatment outcome prediction in radiation oncology have been developed based on the comprehensive review of the literature.

show abstract

Section: Current Research Challenges and Prospectsmentioning

confidence: 99%

Radiomics for Response and Outcome Assessment for Non-Small Cell Lung Cancer

Shi

Yuan

et al. 2018

Technol Cancer Res Treat

View full text Add to dashboard Cite

show abstract

“…38 In the single study that focused on the acquisition parameters at risk of influencing perfusion maps, Bogowicz et al found that the percentage of unstable texture features ranged from 56-98% with different artery contouring, and from 58-75% with different temporal resolution, which is higher than in our study. 27 This highest variability in their study may be explained by the difference in histological types, in imaging modality (CT vs. MRI), in the texture features that were calculated, and in the perfusion maps that were studied (blood flow, mean time transit, and blood volume vs. AUC and K trans in our studies).…”

Section: Discussionmentioning

confidence: 83%

“…Previous studies have demonstrated that temporal parameters (ie, scan duration and temporal resolution) could significantly modify the ability to discriminate benign from malignant prostate or breast lesions, but they were based on average values of DCE‐MRI indices or morphology of the time–intensity curves. Only one study has focused on the stability of texture features extracted from computed tomography (CT) perfusion maps identifying an influence of temporal resolution . Hence, data regarding the influence of temporal parameters on texture features extracted from DCE‐MRI parametric maps are lacking.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Influence of temporal parameters of DCE‐MRI on the quantification of heterogeneity in tumor vascularization

Crombé

Saut

Guigui

et al. 2019

Magnetic Resonance Imaging

View full text Add to dashboard Cite

Background Evaluating heterogeneity in tumor vascularization through texture analysis could improve predictions of patients' outcome and response evaluation. Purpose To investigate the influence of temporal parameters on texture features extracted from dynamic contrast‐enhanced (DCE)‐MRI parametric maps. Study type Prospective cross‐sectional study. Subjects Twenty‐five adults with soft‐tissue sarcoma (STS), median age: 68 years. Field Strength/Sequence DCE‐MRI acquisition using a CAIPIRINHA‐Dixon‐TWIST‐VIBE sequence at 1.5T (temporal resolutions: 2 sec, duration: 5 min). Assessment The area under time–intensity curve (AUC) and Ktrans maps were generated for several temporal resolution (dt = 2 sec, 4 sec, 6 sec, 8 sec, 10 sec, 12 sec, 20 sec) and scan durations (T = 3 min, 4 min, 5 min for a 6‐sec sampling) by downsampling and truncating the initial DCE‐MRI sequence. Tumor volume was manually segmented and propagated on all parametric maps. Thirty‐two first‐ and second order‐texture features were extracted per map to quantify the intratumoral heterogeneity. Statistical Tests The influence of temporal parameters on texture features was studied with repeated‐measures analysis of variance (or nonparametric equivalent). The dispersion of each texture feature depending on temporal parameters was estimated with coefficients of variation (CVs). The performances of multivariate models to predict the response to chemotherapy (ie, binary logistic regression based on the baseline texture features) were compared. Results The temporal resolution had a significant influence on 12/32 (37.5%) and 14/32 (43.8%) texture features evaluated on AUC and Ktrans maps, respectively (range of P < 0.0001–0.0395). Scan duration had a significant influence on 23/32 (71.9%) texture features from Ktrans map (range of P < 0.0001–0.0321). Dispersion was high (mean CV >0.5) with sampling for 2/32 (6.3%) and 10/32 (31.3%) features from AUC and Ktrans maps, respectively; and with truncating for 6/32 (18.8%) features from Ktrans map. The area under the receiver operating characteristics curve of predictive models ranged from 0.77 (95% confidence interval [CI] = [0.54–1.00], with dt = 6 sec T = 4 min) to 0.90 (95% CI = [0.74–1.00], with dt = 6 sec T = 5 min). Data Conclusion The values of texture features extracted from DCE‐MRI parametric maps can be influenced by temporal parameters, which can lead to variations in performance of predictive models. Level of Evidence: 2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1773–1788.

show abstract

“…The influence of image processing and of feature extraction algorithms and implementation on the feature variation is tackled by the Image Biomarker Standardization Initiative (IBSI) [16]. Various studies have evaluated the reproducibility and stability of texture features and the influence of scanner variation and reconstruction settings [5,6,8,14,[17][18][19][20]. These studies generally aim at selecting stable and repeatable texture features for a given task with test-retest and inter-rater reliability analysis, without proposing a method to standardize unstable features.…”

Section: Introductionmentioning

confidence: 99%

Neural network training for cross-protocol radiomic feature standardization in computed tomography

2019

View full text Add to dashboard Cite

Radiomics has shown promising results in several medical studies, yet it suffers from a limited discrimination and informative capability as well as a high variation and correlation with the tomographic scanner types, pixel spacing, acquisition protocol and reconstruction parameters. This paper proposes and compares two methods to transform quantitative image features in order to improve their stability across varying image acquisition parameters while preserving the texture discrimination abilities. In this way, variations in extracted features are representative of true physio-pathological tissue changes in the scanned patients. A first approach is based on a two-layer neural network that can learn a non-linear standardization transformation of various types of features including hand-crafted and deep features. Second, domain adversarial training is explored to increase the invariance of the transformed features to the scanner of origin. The generalization of the proposed approach to unseen textures and unseen scanners is demonstrated by a set of experiments using a publicly available CT texture phantom dataset scanned with various imaging devices and parameters.

show abstract

Stability of radiomic features in CT perfusion maps

Cited by 48 publications

References 48 publications

Radiomics for Response and Outcome Assessment for Non-Small Cell Lung Cancer

Radiomics for Response and Outcome Assessment for Non-Small Cell Lung Cancer

Influence of temporal parameters of DCE‐MRI on the quantification of heterogeneity in tumor vascularization

Neural network training for cross-protocol radiomic feature standardization in computed tomography

Contact Info

Product

Resources

About