The Image Biomarker Standardization Initiative validated consensus-based reference values for 169 radiomics features, thus enabling calibration and verification of radiomics software. Key results: • research teams found agreement for calculation of 169 radiomics features derived from a digital phantom and a human lung cancer on CT scan. • Of these 169 candidate radiomics features, good to excellent reproducibility was achieved for 167 radiomics features using MRI, 18F-FDG PET and CT images obtained in 51 patients with soft-tissue sarcoma.

Textural and shape analysis is gaining considerable interest in medical imaging, particularly to identify parameters characterizing tumor heterogeneity and to feed radiomic models. Here, we present a free, multiplatform, and easy-to-use freeware called LIFEx, which enables the calculation of conventional, histogram-based, textural, and shape features from PET, SPECT, MR, CT, and US images, or from any combination of imaging modalities. The application does not require any programming skills and was developed for medical imaging professionals. The goal is that independent and multicenter evidence of the usefulness and limitations of radiomic features for characterization of tumor heterogeneity and subsequent patient management can be gathered. Many options are offered for interactive textural index calculation and for increasing the reproducibility among centers. The software already benefits from a large user community (more than 800 registered users), and interactions within that community are part of the development strategy. Significance: This study presents a user-friendly, multi-platform freeware to extract radiomic features from PET, SPECT, MR, CT, and US images, or any combination of imaging modalities. Cancer Res; 78(16); 4786–9. ©2018 AACR.

Texture indices are of growing interest for tumor characterization in 18 F-FDG PET. Yet, on the basis of results published in the literature so far, it is unclear which indices should be used, what they represent, and how they relate to conventional indices such as standardized uptake values (SUVs), metabolic volume (MV), and total lesion glycolysis (TLG). We investigated in detail 31 texture indices, 5 firstorder statistics (histogram indices) derived from the gray-level histogram of the tumor region, and their relationship with SUV, MV, and TLG in 3 different tumor types. Methods: Three patient groups corresponding to 3 cancer types at baseline were studied independently: patients with metastatic colorectal cancer (72 lesions), nonsmall cell lung cancer (24 lesions), and breast cancer (54 lesions). Thirty-one texture indices were studied in addition to SUVs, MV, and TLG, and 5 indices extracted from histogram analysis were also investigated. The relationships between indices were studied as well as the robustness of the various texture indices with respect to the parameters involved in the calculation method (sampling schemes and tumor volume of interest). Results: Regardless of the patient group, many indices were highly correlated (Pearson correlation coefficient jrj ≥ 0.80), making it desirable to focus on only a few uncorrelated indices. Three histogram indices were highly correlated with SUVs (jrj ≥ 0.84). Four texture indices were highly correlated with MV, and none was highly correlated with SUVs (jrj ≤ 0.55). The resampling formula used to calculate texture indices had a substantial impact, and resampling using at least 32 discrete values should be used for texture indices calculation. The texture indices change as a function of the segmentation method was higher than that of peak and maximum SUVs but less than mean SUV for 5 texture indices and was larger than that of MV for 14 texture indices and for the 5 histogram indices. All these results were extremely consistent across the 3 tumor types and explained many of the observations reported in the literature so far. Conclusion: None of the histogram indices and only 17 of 31 texture indices were robust with respect to the tumor-segmentation method. An appropriate resampling formula with at least 32 gray levels should be used to avoid introducing a misleading relationship between texture indices and SUV. Some texture indices are highly correlated or strongly correlate with MV whatever the tumor type.Such correlation should be accounted for when interpreting the usefulness of texture indices for tumor characterization, which might call for systematic multivariate analyses.

Several reports have shown that radiomic features are affected by acquisition and reconstruction parameters, thus hampering multicenter studies. We propose a method that, by removing the center effect while preserving patient-specific effects, standardizes features measured from PET images obtained using different imaging protocols. PretreatmentF-FDG PET images of patients with breast cancer were included. In one nuclear medicine department (department A), 63 patients were scanned on a time-of-flight PET/CT scanner, and 16 lesions were triple-negative (TN). In another nuclear medicine department (department B), 74 patients underwent PET/CT on a different brand of scanner and a different reconstruction protocol, and 15 lesions were TN. The images from department A were smoothed using a gaussian filter to mimic data from a third department (department A-S). The primary lesion was segmented to obtain a lesion volume of interest (VOI), and a spheric VOI was set in healthy liver tissue. Three SUVs and 6 textural features were computed in all VOIs. A harmonization method initially described for genomic data was used to estimate the department effect based on the observed feature values. Feature distributions in each department were compared before and after harmonization. In healthy liver tissue, the distributions significantly differed for 4 of 9 features between departments A and B and for 6 of 9 between departments A and A-S ( < 0.05, Wilcoxon test). After harmonization, none of the 9 feature distributions significantly differed between 2 departments ( > 0.1). The same trend was observed in lesions, with a realignment of feature distributions between the departments after harmonization. Identification of TN lesions was largely enhanced after harmonization when the cutoffs were determined on data from one department and applied to data from the other department. The proposed harmonization method is efficient at removing the multicenter effect for textural features and SUVs. The method is easy to use, retains biologic variations not related to a center effect, and does not require any feature recalculation. Such harmonization allows for multicenter studies and for external validation of radiomic models or cutoffs and should facilitate the use of radiomic models in clinical practice.

Background: Radiomics extracts features from medical images more precisely and more accurately than visual assessment. However, radiomics features are affected by CT scanner parameters such as reconstruction kernel or section thickness, thus obscuring underlying biologically important texture features. Purpose: To investigate whether a compensation method could correct for the variations of radiomic feature values caused by using different CT protocols. Materials and Methods: Phantom data involving 10 texture patterns and 74 patients in cohorts 1 (19 men; 42 patients; mean age, 60.4 years; September-October 2013) and 2 (16 men; 32 patients; mean age, 62.1 years; January-September 2007) scanned by using different CT protocols were retrospectively included. For any radiomic feature, the compensation approach identified a protocol-specific transformation to express all data in a common space that were devoid of protocol effects. The differences in statistical distributions between protocols were assessed by using Friedman tests before and after compensation. Principal component analyses were performed on the phantom data to evaluate the ability to distinguish between texture patterns after compensation. Results: In the phantom data, the statistical distributions of features were different between protocols for all radiomic features and texture patterns (P , .05). After compensation, the protocol effect was no longer detectable (P. .05). Principal component analysis demonstrated that each texture pattern was no longer displayed as different clusters corresponding to different imaging protocols, unlike what was observed before compensation. The correction for scanner effect was confirmed in patient data with 100% (10 of 10 features for cohort 1) and 98% (87 of 89 features for cohort 2) of P values less than .05 before compensation, compared with 30% (three of 10) and 15% (13 of 89) after compensation. Conclusion: Image compensation successfully realigned feature distributions computed from different CT imaging protocols and should facilitate multicenter radiomic studies.

PurposeTexture indices (TI) calculated from 18F-FDG PET tumor images show promise for predicting response to therapy and survival. Their calculation involves a resampling of standardized uptake values (SUV) within the tumor. This resampling can be performed differently and significantly impacts the TI values. Our aim was to investigate how the resampling approach affects the ability of TI to reflect tissue-specific pattern of metabolic activity.Methods18F-FDG PET were acquired for 48 naïve-treatment patients with non-small cell lung cancer and for a uniform phantom. We studied 7 TI, SUVmax and metabolic volume (MV) in the phantom, tumors and healthy tissue using the usual relative resampling (RR) method and an absolute resampling (AR) method. The differences in TI values between tissue types and cancer subtypes were investigated using Wilcoxon’s tests.ResultsMost RR-based TI were highly correlated with MV for tumors less than 60 mL (Spearman correlation coefficient r between 0.74 and 1), while this correlation was reduced for AR-based TI (r between 0.06 and 0.27 except for RLNU where r = 0.91). Most AR-based TI were significantly different between tumor and healthy tissues (pvalues <0.01 for all 7 TI) and between cancer subtypes (pvalues<0.05 for 6 TI). Healthy tissue and adenocarcinomas exhibited more homogeneous texture than tumor tissue and squamous cell carcinomas respectively.ConclusionTI computed using an AR method vary as a function of the tissue type and cancer subtype more than the TI involving the usual RR method. AR-based TI might be useful for tumor characterization.

BackgroundThere is currently little support to understand which pathological factors led to differences in tumor texture as measured from FDG PET/CT images. We studied whether tumor heterogeneity measured using texture analysis in FDG-PET/CT images is correlated with pathological prognostic factors in invasive breast cancer.MethodsFifty-four patients with locally advanced breast cancer who had an initial FDG-PET/CT were retrospectively included. In addition to SUVmax, three robust textural indices extracted from 3D matrices: High-Gray-level Run Emphasis (HGRE), Entropy and Homogeneity were studied. Univariate and multivariate logistic regression was used to identify PET parameters associated with poor prognosis pathological factors: hormone receptor negativity, presence of HER-2 and triple negative phenotype. Receiver operating characteristic (ROC) curves and the (AUC) analysis, and reclassification measures, were performed in order to evaluate the performance of combining texture analysis and SUVmax for characterizing breast tumors.ResultsTumor heterogeneity, measured with HGRE, was higher in negative estrogen receptor (p = 0.039) and negative progesterone receptor tumors (p = 0.036), and in Scarff-Bloom-Richardson grade 3 tumors (p = 0.047). None of the PET indices could identify HER-2 positive tumors. Only SUVmax was positively correlated with Ki-67 (p<0.0004). Triple negative breast cancer (TNBC) exhibited higher SUVmax (Odd Ratio = 1.22, 95%CI [1.06–1.39],p = 0.004), lower Homogeneity (OR = 3.57[0.98–12.5],p = 0.05) and higher HGRE (OR = 8.06[1.88–34.51],p = 0.005) than non-TNBC. Multivariate analysis showed that HGRE remained associated with TNBC (OR = 5.27[1.12–1.38],p = 0.03) after adjustment for SUVmax. Combining SUVmax and HGRE yielded in higher area under the ROC curves (AUC) than SUVmax for identifying TNBC: AUC =  0.83 and 0.77, respectively. Probability of correct classification also increased in 77% (10/13) of TNBC and 71% (29/41) of non-TNBC (p = 0.003), when combining SUVmax and HGRE.ConclusionsTumor heterogeneity measured on FDG-PET/CT was higher in invasive breast cancer with poor prognosis pathological factors. Texture analysis might be used, in addition to SUVmax, as a new tool to assess invasive breast cancer aggressiveness.

Objective: Test a practical realignment approach to compensate the technical variability of MR radiomic features.Methods: T1 phantom images acquired on 2 scanners, FLAIR and contrast enhanced T1-weighted (CE-T1w) images of 18 brain tumor patients scanned on both 1.5T and 3T scanners, and 36 T2weighted (T2w) images of prostate cancer patients scanned in one of two centers were investigated. The ComBat procedure was used for harmonizing radiomic features. Differences in statistical distributions in feature values between 1.5 and 3T images were tested before and after harmonization. The prostate studies were used to determine the impact of harmonization to distinguish between Gleason grades (GG).Results: In the phantom data, 40 out of 42 radiomic feature values were significantly different between the 2 scanners before harmonization and none after. In white matter regions, the statistical distributions of features were significantly different (P<0.05) between the 1.5 and 3T images for 37 out of 42 features in both FLAIR and CE-T1w images. After harmonization, no statistically significant differences were observed. In brain tumors, 41 (FLAIR) or 36 (CE-T1w) out of 42 features were significantly different between the 1.5 and 3T images without harmonization, against 1 (FLAIR) or none (CE-T1w) with harmonization. In prostate studies, 636 radiomic features were significantly different between GG after harmonization against 461 before. The ability to distinguish between GG using radiomic features was increased after harmonization. Conclusion:ComBat harmonization efficiently removes inter-center technical inconsistencies in radiomic feature values and increases the sensitivity of studies using data from several scanners.