Quantitative radiomics: impact of stochastic effects on textural feature analysis implies the need for standards

Repeatability of 18 F-FDG PET radiomic features: A phantom study to explore sensitivity to image reconstruction settings, noise, and delineation method

BACKGROUND

18 F-fluoro-2-deoxy-D-Glucose positron emission tomography (18 F-FDG PET) radiomics has the potential to guide the clinical decision making in cancer patients, but validation is required before radiomics can be implemented in the clinical setting. The aim of this study was to explore how feature space reduction and repeatability of 18 F-FDG PET radiomic features are affected by various sources of variation such as underlying data (e.g., object size and uptake), image reconstruction methods and settings, noise, discretization method, and delineation method.

METHODS

The NEMA image quality phantom was scanned with various sphere-to-background ratios (SBR), simulating different activity uptakes, including spheres with low uptake, that is, SBR smaller than 1. Furthermore, images of a phantom containing 3D printed inserts reflecting realistic heterogeneity uptake patterns were acquired. Data were reconstructed using various matrix sizes, reconstruction algorithms, and scan durations (noise). For every specific reconstruction and noise level, ten statistically equal replicates were generated. The phantom inserts were delineated using CT and PET-based segmentation methods. A total of 246 radiomic features was extracted from each image dataset. Images were discretized with a fixed number of 64 bins (FBN) and a fixed bin width (FBW) of 0.25 for the high and a FBW of 0.05 for the low uptake data. In terms of feature reduction, we determined the impact of these factors on the composition of feature clusters, which were defined on the basis of Spearman's correlation matrices. To assess feature repeatability, the intraclass correlation coefficient was calculated over the ten replicates.

RESULTS

In general, larger spheres with high uptake resulted in better repeatability compared to smaller low uptake spheres. In terms of repeatability, features extracted from heterogeneous phantom inserts were comparable to features extracted from bigger high uptake spheres. For example, for an EARL-compliant reconstruction, larger and smaller high uptake spheres yielded good repeatability for 32% and 30% of the features, while the heterogeneous inserts resulted in 34% repeatable features. For the low uptake spheres, this was the case for 22% and 20% of the features for bigger and smaller spheres, respectively. Images reconstructed with point-spread-function (PSF) resulted in the highest repeatability when compared with OSEM or time-of-flight, for example, 53%, 30%, and 32% of repeatable features, respectively (for unsmoothed data, discretized with FBN, 300 s scan duration). Reducing image noise (increasing scan duration and smoothing) and using CT-based segmentation for the low uptake spheres yielded improved repeatability. FBW discretization resulted in higher repeatability than FBN discretization, for example, 89% and 35% of the features, respectively (for the EARL-compliant reconstruction and larger high uptake spheres).

CONCLUSION

Feature space reduction and repeatability of 18 F-FDG PET radiomic features depended on all studied factors. The high sensitivity of PET radiomic features to image quality suggests that a high level of image acquisition and preprocessing standardization is required to be used as clinical imaging biomarker.

Activity painting: PET images of freely defined activity distributions applying a novel phantom technique

The aim of this work was to develop a novel phantom that supports the construction of highly reproducible phantoms with arbitrary activity distributions for PET imaging. It could offer a methodology for answering questions related to texture measurements in PET imaging. The basic idea is to move a point source on a 3-D trajectory in the field of view, while continuously acquiring data. The reconstruction results in a 3-D activity concentration map according to the pathway of the point source. A 22Na calibration point source was attached to a high precision robotic arm system, where the 3-D movement was software controlled. 3-D activity distributions of a homogeneous cube, a sphere, a spherical shell and a heart shape were simulated. These distributions were used to measure uniformity and to characterize reproducibility. Two potential applications using the lesion simulation method are presented: evaluation in changes of textural properties related to the position in the PET field of view; scanner comparison based on visual and quantitative evaluation of texture features. A lesion with volume of 50x50x50 mm3 can be simulated during approximately 1 hour. The reproducibility of the movement was found to be >99%. The coefficients of variation of the voxels within a simulated homogeneous cube was 2.34%. Based on 5 consecutive and independent measurements of a 36 mm diameter hot sphere, the coefficient of variation of the mean activity concentration was 0.68%. We obtained up to 18% differences within the values of investigated textural indexes, when measuring a lesion in different radial positions of the PET field of view. In comparison of two different human PET scanners the percentage differences between heterogeneity parameters were in the range of 5-55%. After harmonizing the voxel sizes this range reduced to 2-16%. The general activity distributions provided by the two different vendor show high similarity visually. For the demonstration of the flexibility of this method, the same pattern was also simulated on a small animal PET scanner giving similar results, both quantitatively and visually. 3-D motion of a point source in the PET field of view is capable to create an irregular shaped activity distribution with high reproducibility.

Characterizing Trastuzumab-Induced Alterations in Intratumoral Heterogeneity with Quantitative Imaging and Immunohistochemistry in HER2+ Breast Cancer

The purpose of this study is to investigate imaging and histology-based measurements of intratumoral heterogeneity to evaluate early treatment response to targeted therapy in a murine model of HER2+ breast cancer. BT474 tumor-bearing mice (N = 30) were treated with trastuzumab or saline and imaged longitudinally with either dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) or 18F-fluoromisonidazole (FMISO) positron emission tomography (PET). At the imaging study end point (day 4 for MRI or 7 for PET), each tumor was excised for immunohistochemistry analysis. Voxel-based histogram analysis was performed on imaging-derived parametric maps (i.e., Ktrans and ve from DCE-MRI, SUV from 18F-FMISO-PET) of the tumor region of interest to measure heterogeneity. Image processing and histogram analysis of whole tumor slice immunohistochemistry data were performed to validate the in vivo imaging findings. Trastuzumab-treated tumors had increased heterogeneity in quantitative imaging measures of cellularity (ve), with a mean Kolmogorov-Smirnov (K-S) distance of 0.32 (P = .05) between baseline and end point distributions. Trastuzumab-treated tumors had increased vascular heterogeneity (Ktrans) and decreased hypoxic heterogeneity (SUV), with a mean K-S distance of 0.42 (P < .01) and 0.46 (P = .047), respectively, between baseline and study end points. These observations were validated by whole-slice immunohistochemistry analysis with mean interquartile range of CD31 distributions of 1.72 for treated and 0.95 for control groups (P = .02). Quantitative longitudinal changes in tumor cellular and vascular heterogeneity in response to therapy may provide evidence for early prediction of response and guide therapy for patients with HER2+ breast cancer.

Deep learning for patient-specific quality assurance: Identifying errors in radiotherapy delivery by radiomic analysis of gamma images with convolutional neural networks

PURPOSE

Patient-specific quality assurance (QA) for intensity-modulated radiation therapy (IMRT) is a ubiquitous clinical procedure, but conventional methods have often been criticized as being insensitive to errors or less effective than other common physics checks. Recently, there has been interest in the application of radiomics, quantitative extraction of image features, to radiotherapy QA. In this work, we investigate a deep learning approach to classify the presence or absence of introduced radiotherapy treatment delivery errors from patient-specific QA.

METHODS

Planar dose maps from 186 IMRT beams from 23 IMRT plans were evaluated. Each plan was transferred to a cylindrical phantom CT geometry. Three sets of planar doses were exported from each plan corresponding to (a) the error-free case, (b) a random multileaf collimator (MLC) error case, and (c) a systematic MLC error case. Each plan was delivered to the electronic portal imaging device (EPID), and planned and measured doses were used to calculate gamma images in an EPID dosimetry software package (for a total of 558 gamma images). Two radiomic approaches were used. In the first, a convolutional neural network with triplet learning was used to extract image features from the gamma images. In the second, a handcrafted approach using texture features was used. The resulting metrics from both approaches were input into four machine learning classifiers (support vector machines, multilayer perceptrons, decision trees, and k-nearest-neighbors) in order to determine whether images contained the introduced errors. Two experiments were considered: the two-class experiment classified images as error-free or containing any MLC error, and the three-class experiment classified images as error-free, containing a random MLC error, or containing a systematic MLC error. Additionally, threshold-based passing criteria were calculated for comparison.

RESULTS

In total, 303 gamma images were used for model training and 255 images were used for model testing. The highest classification accuracy was achieved with the deep learning approach, with a maximum accuracy of 77.3% in the two-class experiment and 64.3% in the three-class experiment. The performance of the handcrafted approach with texture features was lower, with a maximum accuracy of 66.3% in the two-class experiment and 53.7% in the three-class experiment. Variability between the results of the four machine learning classifiers was lower for the deep learning approach vs the texture feature approach. Both radiomic approaches were superior to threshold-based passing criteria.

CONCLUSIONS

Deep learning with convolutional neural networks can be used to classify the presence or absence of introduced radiotherapy treatment delivery errors from patient-specific gamma images. The performance of the deep learning network was superior to a handcrafted approach with texture features, and both radiomic approaches were better than threshold-based passing criteria. The results suggest that radiomic QA is a promising direction for clinical radiotherapy.

Variation in algorithm implementation across radiomics software

Given the increased need for consistent quantitative image analysis, variations in radiomics feature calculations due to differences in radiomics software were investigated. Two in-house radiomics packages and two freely available radiomics packages, MaZda and IBEX, were utilized. Forty 256 × 256 - pixel regions of interest (ROIs) from 40 digital mammograms were studied along with 39 manually delineated ROIs from the head and neck (HN) computed tomography (CT) scans of 39 patients. Each package was used to calculate first-order histogram and second-order gray-level co-occurrence matrix (GLCM) features. Friedman tests determined differences in feature values across packages, whereas intraclass-correlation coefficients (ICC) quantified agreement. All first-order features computed from both mammography and HN cases (except skewness in mammography) showed significant differences across all packages due to systematic biases introduced by each package; however, based on ICC values, all but one first-order feature calculated on mammography ROIs and all but two first-order features calculated on HN CT ROIs showed excellent agreement, indicating the observed differences were small relative to the feature values but the bias was systematic. All second-order features computed from the two databases both differed significantly and showed poor agreement among packages, due largely to discrepancies in package-specific default GLCM parameters. Additional differences in radiomics features were traced to variations in image preprocessing, algorithm implementation, and naming conventions. Large variations in features among software packages indicate that increased efforts to standardize radiomics processes must be conducted.

Radiomics: the facts and the challenges of image analysis

Radiomics is an emerging translational field of research aiming to extract mineable high-dimensional data from clinical images. The radiomic process can be divided into distinct steps with definable inputs and outputs, such as image acquisition and reconstruction, image segmentation, features extraction and qualification, analysis, and model building. Each step needs careful evaluation for the construction of robust and reliable models to be transferred into clinical practice for the purposes of prognosis, non-invasive disease tracking, and evaluation of disease response to treatment. After the definition of texture parameters (shape features; first-, second-, and higher-order features), we briefly discuss the origin of the term radiomics and the methods for selecting the parameters useful for a radiomic approach, including cluster analysis, principal component analysis, random forest, neural network, linear/logistic regression, and other. Reproducibility and clinical value of parameters should be firstly tested with internal cross-validation and then validated on independent external cohorts. This article summarises the major issues regarding this multi-step process, focussing in particular on challenges of the extraction of radiomic features from data sets provided by computed tomography, positron emission tomography, and magnetic resonance imaging.

Radiomics: the facts and the challenges of image analysis

Radiomics is an emerging translational field of research aiming to extract mineable high-dimensional data from clinical images. The radiomic process can be divided into distinct steps with definable inputs and outputs, such as image acquisition and reconstruction, image segmentation, features extraction and qualification, analysis, and model building. Each step needs careful evaluation for the construction of robust and reliable models to be transferred into clinical practice for the purposes of prognosis, non-invasive disease tracking, and evaluation of disease response to treatment. After the definition of texture parameters (shape features; first-, second-, and higher-order features), we briefly discuss the origin of the term radiomics and the methods for selecting the parameters useful for a radiomic approach, including cluster analysis, principal component analysis, random forest, neural network, linear/logistic regression, and other. Reproducibility and clinical value of parameters should be firstly tested with internal cross-validation and then validated on independent external cohorts. This article summarises the major issues regarding this multi-step process, focussing in particular on challenges of the extraction of radiomic features from data sets provided by computed tomography, positron emission tomography, and magnetic resonance imaging.

Spatial-temporal variability of radiomic features and its effect on the classification of lung cancer histology

The purpose of this research was to study the sensitivity of Computed Tomography (CT) radiomic features to motion blurring and signal-to-noise ratios (SNR), and investigate its downstream effect regarding the classification of non-small cell lung cancer (NSCLC) histology. Forty-three radiomic features were considered and classified into one of four categories: Morphological, Intensity, Fine Texture, and Coarse Texture. First, a series of simulations were used to study feature-sensitivity to changes in spatial-temporal resolution. A dynamic digital phantom was used to generate images with different breathing amplitudes and SNR, from which features were extracted and characterized relative to initial simulation conditions. Stage I NSCLC patients were then retrospectively identified, from which three different acquisition-specific feature-spaces were generated based on free-breathing (FB), average-intensity-projection (AIP), and end-of-exhalation (EOE) CT images. These feature-spaces were derived to cover a wide range of spatial-temporal tradeoff. Normalized percent differences and concordance correlation coefficients (CCC) were used to assess the variability between the 3D and 4D acquisition techniques. Subsequently, three corresponding acquisition-specific logistic regression models were developed to classify lung tumor histology. Classification performance was compared between the different data-dependent models. Simulation results demonstrated strong linear dependences (p  >  0.95) between respiratory motion and morphological features, as well as between SNR and texture features. The feature Short Run Emphasis was found to be particularly stable to both motion blurring and changes in SNR. When comparing FB-to-EOE, 37% of features demonstrated high CCC agreement (CCC  >  0.8), compared to only 30% for FB-to-AIP. In classifying tumor histology, EoE images achieved an average AUC, Accuracy, Sensitivity, and Specificity of [Formula: see text], respectively. FB images achieved respective values of [Formula: see text], and AIP images achieved respective values of [Formula: see text]. Several radiomic features have been identified as being relatively robust to spatial-temporal variations based on both simulation data and patient data. In general, features that were sensitive to motion blurring were not necessarily the same features that were sensitive to changes in SNR. Our modeling results suggest that the EoE phase of a 4DCT acquisition may provide useful radiomic information, particularly for features that are highly sensitive to respiratory motion.

A radiogenomic dataset of non-small cell lung cancer

Medical image biomarkers of cancer promise improvements in patient care through advances in precision medicine. Compared to genomic biomarkers, image biomarkers provide the advantages of being non-invasive, and characterizing a heterogeneous tumor in its entirety, as opposed to limited tissue available via biopsy. We developed a unique radiogenomic dataset from a Non-Small Cell Lung Cancer (NSCLC) cohort of 211 subjects. The dataset comprises Computed Tomography (CT), Positron Emission Tomography (PET)/CT images, semantic annotations of the tumors as observed on the medical images using a controlled vocabulary, and segmentation maps of tumors in the CT scans. Imaging data are also paired with results of gene mutation analyses, gene expression microarrays and RNA sequencing data from samples of surgically excised tumor tissue, and clinical data, including survival outcomes. This dataset was created to facilitate the discovery of the underlying relationship between tumor molecular and medical image features, as well as the development and evaluation of prognostic medical image biomarkers.

Radiomic Profiling of Head and Neck Cancer: 18F-FDG PET Texture Analysis as Predictor of Patient Survival

Background and Purpose

The accurate prediction of prognosis and pattern of failure is crucial for optimizing treatment strategies for patients with cancer, and early evidence suggests that image texture analysis has great potential in predicting outcome both in terms of local control and treatment toxicity. The aim of this study was to assess the value of pretreatment ¹⁸F-FDG PET texture analysis for the prediction of treatment failure in primary head and neck squamous cell carcinoma (HNSCC) treated with concurrent chemoradiation therapy.

Methods

We performed a retrospective analysis of 90 patients diagnosed with primary HNSCC treated between January 2010 and June 2017 with concurrent chemo-radiotherapy. All patients underwent ¹⁸F-FDG PET/CT before treatment. ¹⁸F-FDG PET/CT texture features of the whole primary tumor were measured using an open-source texture analysis package. Least absolute shrinkage and selection operator (LASSO) was employed to select the features that are associated the most with clinical outcome, as progression-free survival and overall survival. We performed a univariate and multivariate analysis between all the relevant texture parameters and local failure, adjusting for age, sex, smoking, primary tumor site, and primary tumor stage. Harrell c-index was employed to score the predictive power of the multivariate cox regression models.

Results

Twenty patients (22.2%) developed local failure, whereas the remaining 70 (77.8%) achieved durable local control. Multivariate analysis revealed that one feature, defined as low-intensity long-run emphasis (LILRE), was a significant predictor of outcome regardless of clinical variables (hazard ratio < 0.001, P=0.001).The multivariate model based on imaging biomarkers resulted superior in predicting local failure with a c-index of 0.76 against 0.65 of the model based on clinical variables alone.

Conclusion

LILRE, evaluated on pretreatment ¹⁸F-FDG PET/CT, is associated with higher local failure in patients with HNSCC treated with chemoradiotherapy. Using texture analysis in addition to clinical variables may be useful in predicting local control.

Voxel size and gray level normalization of CT radiomic features in lung cancer

Radiomic features are potential imaging biomarkers for therapy response assessment in oncology. However, the robustness of features with respect to imaging parameters is not well established. Previously identified potential imaging biomarkers were found to be intrinsically dependent on voxel size and number of gray levels (GLs) in a recent texture phantom investigation. Here, we validate the voxel size and GL in-phantom normalizations in lung tumors. Eighteen patients with non-small cell lung cancer of varying tumor volumes were analyzed. To compare with patient data, phantom scans were acquired on eight different scanners. Twenty four previously identified features were extracted from lung tumors. The Spearman rank (r_s) and interclass correlation coefficient (ICC) were used as metrics. Eight out of 10 features showed high (r_s > 0.9) and low (r_s < 0.5) correlations with number of voxels before and after normalizations, respectively. Likewise, texture features were unstable (ICC < 0.6) and highly stable (ICC > 0.8) before and after GL normalizations, respectively. We conclude that voxel size and GL normalizations derived from a texture phantom study also apply to lung tumors. This study highlights the importance and utility of investigating the robustness of radiomic features with respect to CT imaging parameters in radiomic phantoms.

LIFEx: A Freeware for Radiomic Feature Calculation in Multimodality Imaging to Accelerate Advances in the Characterization of Tumor Heterogeneity

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.

Positron emission tomography-based assessment of metabolic gradient and other prognostic features in sarcoma

Intratumoral heterogeneity biomarkers derived from positron emission tomography (PET) imaging with fluorodeoxyglucose (FDG) are of interest for a number of cancers, including sarcoma. A range of radiomic texture variables, adapted from general methodologies for image analysis, has shown promise in the setting. In the context of sarcoma, our group introduced an alternative model-based approach to the measurement of heterogeneity. In this approach, the heterogeneity of a tumor is characterized by the extent to which the 3-D FDG uptake pattern deviates from a simple elliptically contoured structure. By using a nonparametric analysis of the uptake profile obtained from this spatial model, a variable assessing the metabolic gradient of the tumor is developed. The work explores the prognostic potential of this new variable in the context of FDG-PET imaging of sarcoma. A mature clinical series involving 197 patients, 88 of whom have complete time-to-death information, is used. Texture variables based on the imaging data are also evaluated in this series and a range of appropriate machine learning methodologies are then used to explore the complementary prognostic roles for structure and texture variables. We conclude that both texture-based and model-based variables can be combined to achieve enhanced prognostic assessments of outcome for patients with sarcoma based on FDG-PET imaging information.

Radiomics in Nuclear Medicine Applied to Radiation Therapy: Methods, Pitfalls, and Challenges

Radiomics is a recent area of research in precision medicine and is based on the extraction of a large variety of features from medical images. In the field of radiation oncology, comprehensive image analysis is crucial to personalization of treatments. A better characterization of local heterogeneity and the shape of the tumor, depicting individual cancer aggressiveness, could guide dose planning and suggest volumes in which a higher dose is needed for better tumor control. In addition, noninvasive imaging features that could predict treatment outcome from baseline scans could help the radiation oncologist to determine the best treatment strategies and to stratify patients as at low risk or high risk of recurrence. Nuclear medicine molecular imaging reflects information regarding biological processes in the tumor thanks to a wide range of radiotracers. Many studies involving ¹⁸F-fluorodeoxyglucose positron emission tomography suggest an added value of radiomics compared with the use of conventional PET metrics such as standardized uptake value for both tumor diagnosis and prediction of recurrence or treatment outcome. However, these promising results should not hide technical difficulties that still currently prevent the approach from being widely studied or clinically used. These difficulties mostly pertain to the variability of the imaging features as a function of the acquisition device and protocol, the robustness of the models with respect to that variability, and the interpretation of the radiomic models. Addressing the impact of the variability in acquisition and reconstruction protocols is needed, as is harmonizing the radiomic feature calculation methods, to ensure the reproducibility of studies in a multicenter context and their implementation in a clinical workflow. In this review, we explain the potential impact of positron emission tomography radiomics for radiation therapy and underline the various aspects that need to be carefully addressed to make the most of this promising approach.

Error Detection in Intensity-Modulated Radiation Therapy Quality Assurance Using Radiomic Analysis of Gamma Distributions

PURPOSE

To improve the detection of errors in intensity-modulated radiation therapy (IMRT) with a novel method that uses quantitative image features from radiomics to analyze gamma distributions generated during patient specific quality assurance (QA).

METHODS AND MATERIALS

One hundred eighty-six IMRT beams from 23 patient treatments were delivered to a phantom and measured with electronic portal imaging device dosimetry. The treatments spanned a range of anatomic sites; half were head and neck treatments, and the other half were drawn from treatments for lung and rectal cancers, sarcoma, and glioblastoma. Planar gamma distributions, or gamma images, were calculated for each beam using the measured dose and calculated doses from the 3-dimensional treatment planning system under various scenarios: a plan without errors and plans with either simulated random or systematic multileaf collimator mispositioning errors. The gamma images were randomly divided into 2 sets: a training set for model development and testing set for validation. Radiomic features were calculated for each gamma image. Error detection models were developed by training logistic regression models on these radiomic features. The models were applied to the testing set to quantify their predictive utility, determined by calculating the area under the curve (AUC) of the receiver operator characteristic curve, and were compared with traditional threshold-based gamma analysis.

RESULTS

The AUC of the random multileaf collimator mispositioning model on the testing set was 0.761 compared with 0.512 for threshold-based gamma analysis. The AUC for the systematic mispositioning model was 0.717 versus 0.660 for threshold-based gamma analysis. Furthermore, the models could discriminate between the 2 types of errors simulated here, exhibiting AUCs of approximately 0.5 (equivalent to random guessing) when applied to the error they were not designed to detect.

CONCLUSIONS

The feasibility of error detection in patient-specific IMRT QA using radiomic analysis of QA images has been demonstrated. This methodology represents a substantial step forward for IMRT QA with improved sensitivity and specificity over current QA methods and the potential to distinguish between different types of errors.

Combination of computer extracted shape and texture features enables discrimination of granulomas from adenocarcinoma on chest computed tomography

Differentiation between benign and malignant nodules is a problem encountered by radiologists when visualizing computed tomography (CT) scans. Adenocarcinomas and granulomas have a characteristic spiculated appearance and may be fluorodeoxyglucose avid, making them difficult to distinguish for human readers. In this retrospective study, we aimed to evaluate whether a combination of radiomic texture and shape features from noncontrast CT scans can enable discrimination between granulomas and adenocarcinomas. Our study is composed of CT scans of 195 patients from two institutions, one cohort for training ([Formula: see text]) and the other ([Formula: see text]) for independent validation. A set of 645 three-dimensional texture and 24 shape features were extracted from CT scans in the training cohort. Feature selection was employed to identify the most informative features using this set. The top ranked features were also assessed in terms of their stability and reproducibility across the training and testing cohorts and between scans of different slice thickness. Three different classifiers were constructed using the top ranked features identified from the training set. These classifiers were then validated on the test set and the best classifier (support vector machine) yielded an area under the receiver operating characteristic curve of 77.8%.

Significance of the impact of motion compensation on the variability of PET image features

In lung cancer, quantification by positron emission tomography/computed tomography (PET/CT) imaging presents challenges due to respiratory movement. Our primary aim was to study the impact of motion compensation implied by retrospectively gated (4D)-PET/CT on the variability of PET quantitative parameters. Its significance was evaluated by comparison with the variability due to (i) the voxel size in image reconstruction and (ii) the voxel size in image post-resampling. The method employed for feature extraction was chosen based on the analysis of (i) the effect of discretization of the standardized uptake value (SUV) on complementarity between texture features (TF) and conventional indices, (ii) the impact of the segmentation method on the variability of image features, and (iii) the variability of image features across the time-frame of 4D-PET. Thirty-one PET-features were involved. Three SUV discretization methods were applied: a constant width (SUV resolution) of the resampling bin (method RW), a constant number of bins (method RN) and RN on the image obtained after histogram equalization (method EqRN). The segmentation approaches evaluated were 40[Formula: see text] of SUV_max and the contrast oriented algorithm (COA). Parameters derived from 4D-PET images were compared with values derived from the PET image obtained for (i) the static protocol used in our clinical routine (3D) and (ii) the 3D image post-resampled to the voxel size of the 4D image and PET image derived after modifying the reconstruction of the 3D image to comprise the voxel size of the 4D image. Results showed that TF complementarity with conventional indices was sensitive to the SUV discretization method. In the comparison of COA and 40[Formula: see text] contours, despite the values not being interchangeable, all image features showed strong linear correlations (r  >  0.91, [Formula: see text]). Across the time-frames of 4D-PET, all image features followed a normal distribution in most patients. For our patient cohort, the compensation of tumor motion did not have a significant impact on the quantitative PET parameters. The variability of PET parameters due to voxel size in image reconstruction was more significant than variability due to voxel size in image post-resampling. In conclusion, most of the parameters (apart from the contrast of neighborhood matrix) were robust to the motion compensation implied by 4D-PET/CT. The impact on parameter variability due to the voxel size in image reconstruction and in image post-resampling could not be assumed to be equivalent.

Pathologic stratification of operable lung adenocarcinoma using radiomics features extracted from dual energy CT images

PURPOSE

To evaluate the usefulness of surrogate biomarkers as predictors of histopathologic tumor grade and aggressiveness using radiomics data from dual-energy computed tomography (DECT), with the ultimate goal of accomplishing stratification of early-stage lung adenocarcinoma for optimal treatment.

RESULTS

Pathologic grade was divided into grades 1, 2, and 3. Multinomial logistic regression analysis revealed i-uniformity and 97.5th percentile CT attenuation value as independent significant factors to stratify grade 2 or 3 from grade 1. The AUC value calculated from leave-one-out cross-validation procedure for discriminating grades 1, 2, and 3 was 0.9307 (95% CI: 0.8514-1), 0.8610 (95% CI: 0.7547-0.9672), and 0.8394 (95% CI: 0.7045-0.9743), respectively.

MATERIALS AND METHODS

A total of 80 patients with 91 clinically and radiologically suspected stage I or II lung adenocarcinoma were prospectively enrolled. All patients underwent DECT and F-18-fluorodeoxyglucose (FDG) positron emission tomography (PET)/CT, followed by surgery. Quantitative CT and PET imaging characteristics were evaluated using a radiomics approach. Significant features for a tumor aggressiveness prediction model were extracted and used to calculate diagnostic performance for predicting all pathologic grades.

CONCLUSIONS

Quantitative radiomics values from DECT imaging metrics can help predict pathologic aggressiveness of lung adenocarcinoma.

Quantification of Iodine Concentration Using Single-Source Dual-Energy Computed Tomography in a Calf Liver

OBJECTIVE

To evaluate the accuracy of single-source dual-energy computed tomography (ssDECT) in iodine quantification using various segmentation methods in an ex vivo model.

METHODS

Ten sausages, injected with variable quantities of iodinated contrast, were inserted into 2 livers and scanned with ssDECT. Material density iodine images were reconstructed. Three radiologists segmented each sausage. Iodine concentration, volume, and absolute quantity were measured. Agreement between the measured and injected iodine was assessed with the concordance correlation coefficient (CCC). Intrareader agreement was assessed using the intraclass correlation coefficient (ICC).

RESULTS

Air bubbles were observed in sausage (IX). Sausage (X) was within the same view as hyper-attenuating markers used for localization. With IX and X excluded, CCC and ICC were greater than 0.98 and greater than 0.88. When included, CCC and ICC were greater than 0.94 and greater than 0.79.

CONCLUSIONS

Iodine quantification was reproducible and precise. However, accuracy reduced in sausages consisting of air filled cavities and within the same view as hyperattenuating markers.

A Postreconstruction Harmonization Method for Multicenter Radiomic Studies in PET

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. Methods: Pretreatment ¹⁸F-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. Results: 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 (P < 0.05, Wilcoxon test). After harmonization, none of the 9 feature distributions significantly differed between 2 departments (P > 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. Conclusion: 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.

Towards precision medicine: from quantitative imaging to radiomics

Radiology (imaging) and imaging-guided interventions, which provide multi-parametric morphologic and functional information, are playing an increasingly significant role in precision medicine. Radiologists are trained to understand the imaging phenotypes, transcribe those observations (phenotypes) to correlate with underlying diseases and to characterize the images. However, in order to understand and characterize the molecular phenotype (to obtain genomic information) of solid heterogeneous tumours, the advanced sequencing of those tissues using biopsy is required. Thus, radiologists image the tissues from various views and angles in order to have the complete image phenotypes, thereby acquiring a huge amount of data. Deriving meaningful details from all these radiological data becomes challenging and raises the big data issues. Therefore, interest in the application of radiomics has been growing in recent years as it has the potential to provide significant interpretive and predictive information for decision support. Radiomics is a combination of conventional computer-aided diagnosis, deep learning methods, and human skills, and thus can be used for quantitative characterization of tumour phenotypes. This paper discusses the overview of radiomics workflow, the results of various radiomics-based studies conducted using various radiological images such as computed tomography (CT), magnetic resonance imaging (MRI), and positron-emission tomography (PET), the challenges we are facing, and the potential contribution of radiomics towards precision medicine.

Accounting for reconstruction kernel-induced variability in CT radiomic features using noise power spectra

Large variability in computed tomography (CT) radiomics feature values due to CT imaging parameters can have subsequent implications on the prognostic or predictive significance of these features. Here, we investigated the impact of pitch, dose, and reconstruction kernel on CT radiomic features. Moreover, we introduced correction factors to reduce feature variability introduced by reconstruction kernels. The credence cartridge radiomics and American College of Radiology (ACR) phantoms were scanned on five different scanners. ACR phantom was used for 3-D noise power spectrum (NPS) measurements to quantify correlated noise. The coefficient of variation (COV) was used as the variability assessment metric. The variability in texture features due to different kernels was reduced by applying the NPS peak frequency and region of interest (ROI) maximum intensity as correction factors. Most texture features were dose independent but were strongly kernel dependent, which is demonstrated by a significant shift in NPS peak frequency among kernels. Percentage improvement in robustness was calculated for each feature from original and corrected %COV values. Percentage improvements in robustness of 19 features were in the range of 30% to 78% after corrections. We show that NPS peak frequency and ROI maximum intensity can be used as correction factors to reduce variability in CT texture feature values due to reconstruction kernels.

Investigating the Robustness Neighborhood Gray Tone Difference Matrix and Gray Level Co-occurrence Matrix Radiomic Features on Clinical Computed Tomography Systems Using Anthropomorphic Phantoms: Evidence From a Multivendor Study

OBJECTIVE

The aim of this study was to determine if optimized imaging protocols across multiple computed tomography (CT) vendors could result in reproducible radiomic features calculated from an anthropomorphic phantom.

METHODS

Materials with varying degrees of heterogeneity were placed throughout the lungs of the phantom. Twenty scans of the phantom were acquired on 3 CT manufacturers with chest CT protocols that had optimized protocol parameters. Scans were reconstructed using vendor-specific standards and lung kernels. The concordance correlation coefficient (CCC) was used to calculate reproducibility between features. For features with high CCC values, Bland-Altman analysis was also used to quantify agreement.

RESULTS

The mean Hounsfield unit (HU) was 32.93 HU (141.7 to -26.5 HU) for the rubber insert and 347.2 HU (-320.9 to -347.7 HU) for the wood insert. Low CCC values of less than 0.9 were calculated for all features across all scans.

CONCLUSIONS

Radiomic features that are derived from the spatial distribution of voxel intensities should be particularly scrutinized for reproducibility in a multivendor environment.

Tumor radiomic heterogeneity: Multiparametric functional imaging to characterize variability and predict response following cervical cancer radiation therapy

BACKGROUND

Robust approaches to quantify tumor heterogeneity are needed to provide early decision support for precise individualized therapy.

PURPOSE

To conduct a technical exploration of longitudinal changes in tumor heterogeneity patterns on dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI), diffusion-weighted imaging (DWI) and FDG positron emission tomography / computed tomography (PET/CT), and their association to radiation therapy (RT) response in cervical cancer.

STUDY TYPE

Prospective observational study with longitudinal MRI and PET/CT pre-RT, early-RT (2 weeks), and mid-RT (5 weeks).

POPULATION

Twenty-one FIGO IB₂ -IVA cervical cancer patients receiving definitive external beam RT and brachytherapy.

FIELD STRENGTH/SEQUENCE

1.5T, precontrast axial T₁ -weighted, axial and sagittal T₂ -weighted, sagittal DWI (multi-b values), sagittal DCE MRI (<10 sec temporal resolution), postcontrast axial T₁ -weighted.

ASSESSMENT

Response assessment 1 month after completion of treatment by a board-certified radiation oncologist from manually delineated tumor volume changes.

STATISTICAL TESTS

Intensity histogram (IH) quantiles (DCE SI_10% and DWI ADC_10% , FDG-PET SUV_max ) and distribution moments (mean, variance, skewness, kurtosis) were extracted. Differences in IH features between timepoints and modalities were evaluated by Skillings-Mack tests with Holm's correction. Area under receiver-operating characteristic curve (AUC) and Mann-Whitney testing was performed to discriminate treatment response using IH features.

RESULTS

Tumor IH means and quantiles varied significantly during RT (SUV_mean : ↓28-47%, SUV_max : ↓30-59%, SI_mean : ↑8-30%, SI_10% : ↑8-19%, ADC_mean : ↑16%, P < 0.02 for each). Among IH heterogeneity features, FDG-PET SUV_CoV (↓16-30%, P = 0.011) and DW-MRI ADC_skewness decreased (P = 0.001). FDG-PET SUV_CoV was higher than DCE-MRI SI_CoV and DW-MRI ADC_CoV at baseline (P < 0.001) and 2 weeks (P = 0.010). FDG-PET SUV_kurtosis was lower than DCE-MRI SI_kurtosis and DW-MRI ADC_kurtosis at baseline (P = 0.001). Some IH features appeared to associate with favorable tumor response, including large early RT changes in DW-MRI ADC_skewness (AUC = 0.86).

DATA CONCLUSION

Preliminary findings show tumor heterogeneity was variable between patients, modalities, and timepoints. Radiomic assessment of changing tumor heterogeneity has the potential to personalize treatment and power outcome prediction.

LEVEL OF EVIDENCE

2 Technical Efficacy: Stage 3 J. Magn. Reson. Imaging 2018;47:1388-1396.

Repeatability of Radiomic Features in Non-Small-Cell Lung Cancer [(18)F]FDG-PET/CT Studies: Impact of Reconstruction and Delineation

Purpose

To assess (1) the repeatability and (2) the impact of reconstruction methods and delineation on the repeatability of 105 radiomic features in non-small-cell lung cancer (NSCLC) 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG) positron emission tomorgraphy/computed tomography (PET/CT) studies.

Procedures

Eleven NSCLC patients received two baseline whole-body PET/CT scans. Each scan was reconstructed twice, once using the point spread function (PSF) and once complying with the European Association for Nuclear Medicine (EANM) guidelines for tumor PET imaging. Volumes of interest (n = 19) were delineated twice, once on PET and once on CT images.

Results

Sixty-three features showed an intraclass correlation coefficient ≥ 0.90 independent of delineation or reconstruction. More features were sensitive to a change in delineation than to a change in reconstruction (25 and 3 features, respectively).

Conclusions

The majority of features in NSCLC [¹⁸F]FDG-PET/CT studies show a high level of repeatability that is similar or better compared to simple standardized uptake value measures.

Electronic supplementary material

The online version of this article (doi:10.1007/s11307-016-0940-2) contains supplementary material, which is available to authorized users.

Radiomics strategies for risk assessment of tumour failure in head-and-neck cancer

Quantitative extraction of high-dimensional mineable data from medical images is a process known as radiomics. Radiomics is foreseen as an essential prognostic tool for cancer risk assessment and the quantification of intratumoural heterogeneity. In this work, 1615 radiomic features (quantifying tumour image intensity, shape, texture) extracted from pre-treatment FDG-PET and CT images of 300 patients from four different cohorts were analyzed for the risk assessment of locoregional recurrences (LR) and distant metastases (DM) in head-and-neck cancer. Prediction models combining radiomic and clinical variables were constructed via random forests and imbalance-adjustment strategies using two of the four cohorts. Independent validation of the prediction and prognostic performance of the models was carried out on the other two cohorts (LR: AUC = 0.69 and CI = 0.67; DM: AUC = 0.86 and CI = 0.88). Furthermore, the results obtained via Kaplan-Meier analysis demonstrated the potential of radiomics for assessing the risk of specific tumour outcomes using multiple stratification groups. This could have important clinical impact, notably by allowing for a better personalization of chemo-radiation treatments for head-and-neck cancer patients from different risk groups.

Imaging features from pretreatment CT scans are associated with clinical outcomes in nonsmall-cell lung cancer patients treated with stereotactic body radiotherapy

PURPOSE

To investigate whether imaging features from pretreatment planning CT scans are associated with overall survival (OS), recurrence-free survival (RFS), and loco-regional recurrence-free survival (LR-RFS) after stereotactic body radiotherapy (SBRT) among nonsmall-cell lung cancer (NSCLC) patients.

PATIENTS AND METHODS

A total of 92 patients (median age: 73 yr) with stage I or IIA NSCLC were qualified for this study. A total dose of 50 Gy in five fractions was the standard treatment. Besides clinical characteristics, 24 "semantic" image features were manually scored based on a point scale (up to 5) and 219 computer-derived "radiomic" features were extracted based on whole tumor segmentation. Statistical analysis was performed using Cox proportional hazards model and Harrell's C-index, and the robustness of final prognostic model was assessed using tenfold cross validation by dichotomizing patients according to the survival or recurrence status at 24 months.

RESULTS

Two-year OS, RFS and LR-RFS were 69.95%, 41.3%, and 51.85%, respectively. There was an improvement of Harrell's C-index when adding imaging features to a clinical model. The model for OS contained the Eastern Cooperative Oncology Group (ECOG) performance status [Hazard Ratio (HR) = 2.78, 95% Confidence Interval (CI): 1.37-5.65], pleural retraction (HR = 0.27, 95% CI: 0.08-0.92), F2 (short axis × longest diameter, HR = 1.72, 95% CI: 1.21-2.44) and F186 (Hist-Energy-L1, HR = 1.27, 95% CI: 1.00-1.61); The prognostic model for RFS contained vessel attachment (HR = 2.13, 95% CI: 1.24-3.64) and F2 (HR = 1.69, 95% CI: 1.33-2.15); and the model for LR-RFS contained the ECOG performance status (HR = 2.01, 95% CI: 1.12-3.60) and F2 (HR = 1.67, 95% CI: 1.29-2.18).

CONCLUSIONS

Imaging features derived from planning CT demonstrate prognostic value for recurrence following SBRT treatment, and might be helpful in patient stratification.

The impact of image reconstruction settings on 18F-FDG PET radiomic features: multi-scanner phantom and patient studies

OBJECTIVES

The purpose of this study was to investigate the robustness of different PET/CT image radiomic features over a wide range of different reconstruction settings.

METHODS

Phantom and patient studies were conducted, including two PET/CT scanners. Different reconstruction algorithms and parameters including number of sub-iterations, number of subsets, full width at half maximum (FWHM) of Gaussian filter, scan time per bed position and matrix size were studied. Lesions were delineated and one hundred radiomic features were extracted. All radiomics features were categorized based on coefficient of variation (COV).

RESULTS

Forty seven percent features showed COV ≤ 5% and 10% of which showed COV > 20%. All geometry based, 44% and 41% of intensity based and texture based features were found as robust respectively. In regard to matrix size, 56% and 6% of all features were found non-robust (COV > 20%) and robust (COV ≤ 5%) respectively.

CONCLUSIONS

Variability and robustness of PET/CT image radiomics in advanced reconstruction settings is feature-dependent, and different settings have different effects on different features. Radiomic features with low COV can be considered as good candidates for reproducible tumour quantification in multi-center studies.

KEY POINTS

• PET/CT image radiomics is a quantitative approach assessing different aspects of tumour uptake. • Radiomic features robustness is an important issue over different image reconstruction settings. • Variability and robustness of PET/CT image radiomics in advanced reconstruction settings is feature-dependent. • Robust radiomic features can be considered as good candidates for tumour quantification.

Intrinsic dependencies of CT radiomic features on voxel size and number of gray levels

PURPOSE

Many radiomics features were originally developed for non-medical imaging applications and therefore original assumptions may need to be reexamined. In this study, we investigated the impact of slice thickness and pixel spacing (or pixel size) on radiomics features extracted from Computed Tomography (CT) phantom images acquired with different scanners as well as different acquisition and reconstruction parameters. The dependence of CT texture features on gray-level discretization was also evaluated.

METHODS AND MATERIALS

A texture phantom composed of 10 different cartridges of different materials was scanned on eight different CT scanners from three different manufacturers. The images were reconstructed for various slice thicknesses. For each slice thickness, the reconstruction Field Of View (FOV) was varied to render pixel sizes ranging from 0.39 to 0.98 mm. A fixed spherical region of interest (ROI) was contoured on the images of the shredded rubber cartridge and the 3D printed, 20% fill, acrylonitrile butadiene styrene plastic cartridge (ABS20) for all phantom imaging sets. Radiomic features were extracted from the ROIs using an in-house program. Features categories were: shape (10), intensity (16), GLCM (24), GLZSM (11), GLRLM (11), and NGTDM (5), fractal dimensions (8) and first-order wavelets (128), for a total of 213 features. Voxel-size resampling was performed to investigate the usefulness of extracting features using a suitably chosen voxel size. Acquired phantom image sets were resampled to a voxel size of 1 × 1 × 2 mm³ using linear interpolation. Image features were therefore extracted from resampled and original datasets and the absolute value of the percent coefficient of variation (%COV) for each feature was calculated. Based on the %COV values, features were classified in 3 groups: (1) features with large variations before and after resampling (%COV >50); (2) features with diminished variation (%COV <30) after resampling; and (3) features that had originally moderate variation (%COV <50%) and were negligibly affected by resampling. Group 2 features were further studied by modifying feature definitions to include voxel size. Original and voxel-size normalized features were used for interscanner comparisons. A subsequent analysis investigated feature dependency on gray-level discretization by extracting 51 texture features from ROIs from each of the 10 different phantom cartridges using 16, 32, 64, 128, and 256 gray levels.

RESULTS

Out of the 213 features extracted, 150 were reproducible across voxel sizes, 42 improved significantly (%COV <30, Group 2) after resampling, and 21 had large variations before and after resampling (Group 1). Ten features improved significantly after definition modification effectively removed their voxel-size dependency. Interscanner comparison indicated that feature variability among scanners nearly vanished for 8 of these 10 features. Furthermore, 17 out of 51 texture features were found to be dependent on the number of gray levels. These features were redefined to include the number of gray levels which greatly reduced this dependency.

CONCLUSION

Voxel-size resampling is an appropriate pre-processing step for image datasets acquired with variable voxel sizes to obtain more reproducible CT features. We found that some of the radiomics features were voxel size and gray-level discretization-dependent. The introduction of normalizing factors in their definitions greatly reduced or removed these dependencies.

Variability in CT lung-nodule quantification: Effects of dose reduction and reconstruction methods on density and texture based features

Purpose:

To investigate the effects of dose level and reconstruction method on density and texture based features computed from CT lung nodules.

Methods:

This study had two major components. In the first component, a uniform water phantom was scanned at three dose levels and images were reconstructed using four conventional filtered backprojection (FBP) and four iterative reconstruction (IR) methods for a total of 24 different combinations of acquisition and reconstruction conditions. In the second component, raw projection (sinogram) data were obtained for 33 lung nodules from patients scanned as a part of their clinical practice, where low dose acquisitions were simulated by adding noise to sinograms acquired at clinical dose levels (a total of four dose levels) and reconstructed using one FBP kernel and two IR kernels for a total of 12 conditions. For the water phantom, spherical regions of interest (ROIs) were created at multiple locations within the water phantom on one reference image obtained at a reference condition. For the lung nodule cases, the ROI of each nodule was contoured semiautomatically (with manual editing) from images obtained at a reference condition. All ROIs were applied to their corresponding images reconstructed at different conditions. For 17 of the nodule cases, repeat contours were performed to assess repeatability. Histogram (eight features) and gray level co-occurrence matrix (GLCM) based texture features (34 features) were computed for all ROIs. For the lung nodule cases, the reference condition was selected to be 100% of clinical dose with FBP reconstruction using the B45f kernel; feature values calculated from other conditions were compared to this reference condition. A measure was introduced, which the authors refer to as Q, to assess the stability of features across different conditions, which is defined as the ratio of reproducibility (across conditions) to repeatability (across repeat contours) of each feature.

Results:

The water phantom results demonstrated substantial variability among feature values calculated across conditions, with the exception of histogram mean. Features calculated from lung nodules demonstrated similar results with histogram mean as the most robust feature (Q ≤ 1), having a mean and standard deviation Q of 0.37 and 0.22, respectively. Surprisingly, histogram standard deviation and variance features were also quite robust. Some GLCM features were also quite robust across conditions, namely, diff. variance, sum variance, sum average, variance, and mean. Except for histogram mean, all features have a Q of larger than one in at least one of the 3% dose level conditions.

Conclusions:

As expected, the histogram mean is the most robust feature in their study. The effects of acquisition and reconstruction conditions on GLCM features vary widely, though trending toward features involving summation of product between intensities and probabilities being more robust, barring a few exceptions. Overall, care should be taken into account for variation in density and texture features if a variety of dose and reconstruction conditions are used for the quantification of lung nodules in CT, otherwise changes in quantification results may be more reflective of changes due to acquisition and reconstruction conditions than in the nodule itself.

Texture analysis of medical images for radiotherapy applications

The high-throughput extraction of quantitative information from medical images, known as radiomics, has grown in interest due to the current necessity to quantitatively characterize tumour heterogeneity. In this context, texture analysis, consisting of a variety of mathematical techniques that can describe the grey-level patterns of an image, plays an important role in assessing the spatial organization of different tissues and organs. For these reasons, the potentiality of texture analysis in the context of radiotherapy has been widely investigated in several studies, especially for the prediction of the treatment response of tumour and normal tissues. Nonetheless, many different factors can affect the robustness, reproducibility and reliability of textural features, thus limiting the impact of this technique. In this review, an overview of the most recent works that have applied texture analysis in the context of radiotherapy is presented, with particular focus on the assessment of tumour and tissue response to radiations. Preliminary, the main factors that have an influence on features estimation are discussed, highlighting the need of more standardized image acquisition and reconstruction protocols and more accurate methods for region of interest identification. Despite all these limitations, texture analysis is increasingly demonstrating its ability to improve the characterization of intratumour heterogeneity and the prediction of clinical outcome, although prospective studies and clinical trials are required to draw a more complete picture of the full potential of this technique.

Characterization of PET/CT images using texture analysis: the past, the present… any future?

After seminal papers over the period 2009 - 2011, the use of texture analysis of PET/CT images for quantification of intratumour uptake heterogeneity has received increasing attention in the last 4 years. Results are difficult to compare due to the heterogeneity of studies and lack of standardization. There are also numerous challenges to address. In this review we provide critical insights into the recent development of texture analysis for quantifying the heterogeneity in PET/CT images, identify issues and challenges, and offer recommendations for the use of texture analysis in clinical research. Numerous potentially confounding issues have been identified, related to the complex workflow for the calculation of textural features, and the dependency of features on various factors such as acquisition, image reconstruction, preprocessing, functional volume segmentation, and methods of establishing and quantifying correspondences with genomic and clinical metrics of interest. A lack of understanding of what the features may represent in terms of the underlying pathophysiological processes and the variability of technical implementation practices makes comparing results in the literature challenging, if not impossible. Since progress as a field requires pooling results, there is an urgent need for standardization and recommendations/guidelines to enable the field to move forward. We provide a list of correct formulae for usual features and recommendations regarding implementation. Studies on larger cohorts with robust statistical analysis and machine learning approaches are promising directions to evaluate the potential of this approach.

Lung nodule malignancy classification using only radiologist-quantified image features as inputs to statistical learning algorithms: probing the Lung Image Database Consortium dataset with two statistical learning methods

In the assessment of nodules in CT scans of the lungs, a number of image-derived features are diagnostically relevant. Currently, many of these features are defined only qualitatively, so they are difficult to quantify from first principles. Nevertheless, these features (through their qualitative definitions and interpretations thereof) are often quantified via a variety of mathematical methods for the purpose of computer-aided diagnosis (CAD). To determine the potential usefulness of quantified diagnostic image features as inputs to a CAD system, we investigate the predictive capability of statistical learning methods for classifying nodule malignancy. We utilize the Lung Image Database Consortium dataset and only employ the radiologist-assigned diagnostic feature values for the lung nodules therein, as well as our derived estimates of the diameter and volume of the nodules from the radiologists' annotations. We calculate theoretical upper bounds on the classification accuracy that are achievable by an ideal classifier that only uses the radiologist-assigned feature values, and we obtain an accuracy of 85.74 [Formula: see text], which is, on average, 4.43% below the theoretical maximum of 90.17%. The corresponding area-under-the-curve (AUC) score is 0.932 ([Formula: see text]), which increases to 0.949 ([Formula: see text]) when diameter and volume features are included and has an accuracy of 88.08 [Formula: see text]. Our results are comparable to those in the literature that use algorithmically derived image-based features, which supports our hypothesis that lung nodules can be classified as malignant or benign using only quantified, diagnostic image features, and indicates the competitiveness of this approach. We also analyze how the classification accuracy depends on specific features and feature subsets, and we rank the features according to their predictive power, statistically demonstrating the top four to be spiculation, lobulation, subtlety, and calcification.

Evaluating stability of histomorphometric features across scanner and staining variations: prostate cancer diagnosis from whole slide images

Quantitative histomorphometry (QH) is the process of computerized feature extraction from digitized tissue slide images to predict disease presence, behavior, and outcome. Feature stability between sites may be compromised by laboratory-specific variables including dye batch, slice thickness, and the whole slide scanner used. We present two new measures, preparation-induced instability score and latent instability score, to quantify feature instability across and within datasets. In a use case involving prostate cancer, we examined QH features which may detect cancer on whole slide images. Using our method, we found that five feature families (graph, shape, co-occurring gland tensor, sub-graph, and texture) were different between datasets in 19.7% to 48.6% of comparisons while the values expected without site variation were 4.2% to 4.6%. Color normalizing all images to a template did not reduce instability. Scanning the same 34 slides on three scanners demonstrated that Haralick features were most substantively affected by scanner variation, being unstable in 62% of comparisons. We found that unstable feature families performed significantly worse in inter- than intrasite classification. Our results appear to suggest QH features should be evaluated across sites to assess robustness, and class discriminability alone should not represent the benchmark for digital pathology feature selection.

DiFranco M, Opposits G, K. Krizsan A, Garai I, Czernin J, Varga J, Tron L, Balkay L. A Study on the Basic Criteria for Selecting Heterogeneity Parameters of F18-FDG PET Images

Textural analysis might give new insights into the quantitative characterization of metabolically active tumors. More than thirty textural parameters have been investigated in former F18-FDG studies already. The purpose of the paper is to declare basic requirements as a selection strategy to identify the most appropriate heterogeneity parameters to measure textural features. Our predefined requirements were: a reliable heterogeneity parameter has to be volume independent, reproducible, and suitable for expressing quantitatively the degree of heterogeneity. Based on this criteria, we compared various suggested measures of homogeneity. A homogeneous cylindrical phantom was measured on three different PET/CT scanners using the commonly used protocol. In addition, a custom-made inhomogeneous tumor insert placed into the NEMA image quality phantom was imaged with a set of acquisition times and several different reconstruction protocols. PET data of 65 patients with proven lung lesions were retrospectively analyzed as well. Four heterogeneity parameters out of 27 were found as the most attractive ones to characterize the textural properties of metabolically active tumors in FDG PET images. These four parameters included Entropy, Contrast, Correlation, and Coefficient of Variation. These parameters were independent of delineated tumor volume (bigger than 25-30 ml), provided reproducible values (relative standard deviation< 10%), and showed high sensitivity to changes in heterogeneity. Phantom measurements are a viable way to test the reliability of heterogeneity parameters that would be of interest to nuclear imaging clinicians.

Understanding Changes in Tumor Texture Indices in PET: A Comparison Between Visual Assessment and Index Values in Simulated and Patient Data

The use of texture indices to characterize tumor heterogeneity from PET images is being increasingly investigated in retrospective studies, yet the interpretation of PET-derived texture index values has not been thoroughly reported. Furthermore, the calculation of texture indices lacks a standardized methodology, making it difficult to compare published results. To allow for texture index value interpretation, we investigated the changes in value of 6 texture indices computed from simulated and real patient data. Methods: Ten sphere models mimicking different activity distribution patterns and the ¹⁸F-FDG PET images from 54 patients with breast cancer were used. For each volume of interest, 6 texture indices were measured. The values of texture indices and how they changed as a function of the activity distribution were assessed and compared with the visual assessment of tumor heterogeneity. Results: Using the sphere models and real tumors, we identified 2 sets of texture indices reflecting different types of uptake heterogeneity. Set 1 included homogeneity, entropy, short-run emphasis, and long-run emphasis, all of which were sensitive to the presence of uptake heterogeneity but did not distinguish between hyper- and hyposignal within an otherwise uniform activity distribution. Set 2 comprised high-gray-level-zone emphasis and low-gray-level-zone emphasis, which were mostly sensitive to the average uptake rather than to the uptake local heterogeneity. Four of 6 texture indices significantly differed between homogeneous and heterogeneous lesions as defined by 2 nuclear medicine physicians (P < 0.05). All texture index values were sensitive to voxel size (variations up to 85.8% for the most homogeneous sphere models) and edge effects (variations up to 29.1%). Conclusion: Unlike a previous report, our study found that variations in texture indices were intuitive in the sphere models and real tumors: the most homogeneous uptake distribution exhibited the highest homogeneity and lowest entropy. Two families of texture index reflecting different types of uptake patterns were identified. Variability in texture index values as a function of voxel size and inclusion of tumor edges was demonstrated, calling for a standardized calculation methodology. This study provides guidance for nuclear medicine physicians in interpreting texture indices in future studies and clinical practice.

Radiomics and its emerging role in lung cancer research, imaging biomarkers and clinical management: State of the art

With the development of functional imaging modalities we now have the ability to study the microenvironment of lung cancer and its genomic instability. Radiomics is defined as the use of automated or semi-automated post-processing and analysis of large amounts of quantitative imaging features that can be derived from medical images. The automated generation of these analytical features helps to quantify a number of variables in the imaging assessment of lung malignancy. These imaging features include: tumor spatial complexity, elucidation of the tumor genomic heterogeneity and composition, subregional identification in terms of tumor viability or aggressiveness, and response to chemotherapy and/or radiation. Therefore, a radiomic approach can help to reveal unique information about tumor behavior. Currently available radiomic features can be divided into four major classes: (a) morphological, (b) statistical, (c) regional, and (d) model-based. Each category yields quantitative parameters that reflect specific aspects of a tumor. The major challenge is to integrate radiomic data with clinical, pathological, and genomic information to decode the different types of tissue biology. There are many currently available radiomic studies on lung cancer for which there is a need to summarize the current state of the art.

Sensitivity of Image Features to Noise in Conventional and Respiratory-Gated PET/CT Images of Lung Cancer: Uncorrelated Noise Effects

The effect of noise on image features has yet to be studied in depth. Our objective was to explore how significantly image features are affected by the addition of uncorrelated noise to an image. The signal-to-noise ratio and noise power spectrum were calculated for a positron emission tomography/computed tomography scanner using a Ge-68 phantom. The conventional and respiratory-gated positron emission tomography/computed tomography images of 31 patients with lung cancer were retrospectively examined. Multiple sets of noise images were created for each original image by adding Gaussian noise of varying standard deviation equal to 2.5%, 4.0%, and 6.0% of the maximum intensity for positron emission tomography images and 10, 20, 50, 80, and 120 Hounsfield units for computed tomography images. Image features were extracted from all images, and percentage differences between the original image and the noise image feature values were calculated. These features were then categorized according to the noise sensitivity. The contour-dependent shape descriptors averaged below 4% difference in positron emission tomography and below 13% difference in computed tomography between noise and original images. Gray level size zone matrix features were the most sensitive to uncorrelated noise exhibiting average differences >200% for conventional and respiratory-gated images in computed tomography and 90% in positron emission tomography. Image feature differences increased as the noise level increased for shape, intensity, and gray-level co-occurrence matrix features in positron emission tomography and for gray-level co-occurrence matrix and gray-level size zone matrix features in conventional computed tomography. Investigators should be aware of the noise effects on image features.

Measuring total liver function on sulfur colloid SPECT/CT for improved risk stratification and outcome prediction of hepatocellular carcinoma patients

Background

Assessment of liver function is critical in hepatocellular carcinoma (HCC) patient management. We evaluated parameters of [^99mTc] sulfur colloid (SC) SPECT/CT liver uptake for association with clinical measures of liver function and outcome in HCC patients.

Methods

Thirty patients with HCC and variable Child-Turcotte-Pugh scores (CTP A5-C10) underwent [^99mTc]SC SPECT/CT scans for radiotherapy planning. Gross tumor volume (GTV), anatomic liver volume (ALV), and spleen were contoured on CT. SC SPECT image parameters include threshold-based functional liver volumes (FLV) relative to ALV, mean liver-to-spleen uptake ratio (L/S_mean), and total liver function (TLF) ratio derived from the product of FLV and L/S_mean. Optimal SC uptake thresholds were determined by ROC analysis for maximizing CTP classification accuracy. Image metrics were tested for rank correlation to composite scores and clinical liver function parameters. Image parameters of liver function were tested for association to overall survival with Cox proportional hazard regression.

Results

Optimized thresholds on SC SPECT were 58 % of maximum uptake for FLV, 38 % for L/S_mean, and 58 % for TLF. TLF produced the highest CTP classification accuracy (AUC = 0.93) at threshold of 0.35 (sensitivity = 0.88, specificity = 0.86). Higher TLF was associated with lower CTP score: TLF_A = 0.6 (0.4–0.8) versus TLF_B = 0.2 (0.1–0.3), p < 10⁻⁴. TLF was rank correlated to albumin and bilirubin (|R| > 0.63). Only TLF >0.30 was independently associated with overall survival when adjusting for CTP class (HR = 0.12, 95 % CI = 0.02–0.58, p = 0.008).

Conclusions

SC SPECT/CT liver uptake correlated with differential liver function. TLF was associated with improved overall survival and may aid in personalized oncologic management of HCC patients.