Applications and limitations of radiomics

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.

Detection of epithelial growth factor receptor (EGFR) mutations on CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks

Background

We aim to analyze the ability to detect epithelial growth factor receptor (EGFR) mutations on chest CT images of patients with lung adenocarcinoma using radiomics and/or multi-level residual convolutionary neural networks (MCNNs).

Methods

We retrospectively collected 1,010 consecutive patients in Shanghai Chest Hospital from 2013 to 2017, among which 510 patients were EGFR-mutated and 500 patients were wild-type. The patients were randomly divided into a training set (810 patients) and a validation set (200 patients) according to a balanced distribution of clinical features. The CT images and the corresponding EGFR status measured by Amplification Refractory Mutation System (ARMS) method of the patients in the training set were utilized to construct both a radiomics-based model (M_Radiomics) and MCNNs-based model (M_MCNNs). The M_Radiomics and M_MCNNs were combined to build the Model_{Radiomics+MCNNs} (M_{Radiomics+MCNNs}). Clinical data of gender and smoking history constructed the clinical features-based model (M_Clinical). M_Clinical was then added into M_Radiomics, M_MCNNs, and M_{Radiomics+MCNNs} to establish the Model_{Radiomics+Clinical} (M_{Radiomics+Clinical}), the Model_{MCNNs+Clinical} (M_{MCNNs+Clinical}) and the Model_{Radiomics+MCNNs+Clinical} (M_{Radiomics+MCNNs+Clinical}). All the seven models were tested in the validation set to ascertain whether they were competent to detect EGFR mutations. The detection efficiency of each model was also compared in terms of area under the curve (AUC), sensitivity and specificity.

Results

The AUC of the M_Radiomics, M_MCNNs and M_{Radiomics+MCNNs} to predict EGFR mutations was 0.740, 0.810 and 0.811 respectively. The performance of M_MCNNs was better than that of M_Radiomics (P=0.0225). The addition of clinical features did not improve the AUC of the M_Radiomics (P=0.623), the M_MCNNs (P=0.114) and the M_{Radiomics+MCNNs} (P=0.058). The M_{Radiomics+MCNNs+Clinical} demonstrated the highest AUC value of 0.834. The M_MCNNs did not demonstrate any inferiority when compared with the M_{Radiomics+MCNNs} (P=0.742) and the M_{Radiomics+MCNNs+Clinical} (P=0.056).

Conclusions

Both of the M_Radiomics and the M_CNNs could predict EGFR mutations on CT images of patients with lung adenocarcinoma. The M_MCNNs outperformed the M_Radiomics in the detection of EGFR mutations. The combination of these two models, even added with clinical features, is not significantly more efficient than M_MCNNs alone.

Peritumoral tissue on preoperative imaging reveals microvascular invasion in hepatocellular carcinoma: a systematic review and meta-analysis

BACKGROUND

Histologic microvascular invasion (MVI) substantially worsens the prognosis of patients with hepatocellular carcinoma, and can only be diagnosed postoperatively. Preoperative assessment of MVI by imaging has been focused on tumor-related features, while peritumoral imaging features have been indicated elsewhere to be more accurate. The aim of the present study is to evaluate the association between peritumoral imaging features and MVI.

METHODS

Literature search was performed using the PubMed, Embase, and Cochrane Library databases. Summary results of the association between peritumoral imaging features and MVI were presented as the odds ratio (OR) and the 95% confidence interval. Meta-regression and subgroup analyses were performed when heterogeneity was detected. Diagnostic accuracy analysis was also conducted for identified features.

RESULTS

Ten studies were included in the analysis. Moderate and low heterogeneities were found among the seven studies on peritumoral enhancement and four studies on peritumoral hypointensity on HBP, respectively. Summary results revealed a significant association between MVI and peritumoral enhancement (OR 4.04 [2.23, 7.32], p < 0.05), and peritumoral hypointensity on HBP (OR 10.62 [5.31, 21.26], p < 0.05). Diagnostic accuracy analysis revealed high specificity (0.90-0.94) but low sensitivity (0.29-0.40) for both features to assess MVI.

CONCLUSION

The two peritumoral imaging features are significantly associated with MVI. The two features highly suggest MVI only when present with a high false negative rate. Promotion of their diagnostic efficiency can be a worthwhile task for future research.

A radiomics-based formula for the preoperative prediction of postoperative pancreatic fistula in patients with pancreaticoduodenectomy

Objective

The objective of the study was to develop and validate a radiomics-based formula for the preoperative prediction of postoperative pancreatic fistula (POPF) in patients undergoing pancreaticoduodenectomy (PD).

Materials and methods

A total of 117 consecutive patients who underwent PD were enrolled in this retrospective study. Radiomics features were extracted from portal venous phase computed tomography of the above patients. The least absolute shrinkage and selection operator logistic regression was used to construct a formula of Rad-score calculation. Then the performance of the formula was assessed with standard pancreatic Fistula Risk Score.

Results

The Rad-score could predict POPF with an area under the curve (AUC) of 0.8248 in the training cohort and of 0.7609 in the validation cohort. Patients who had experienced POPF generally had a statistically higher Rad-score than those who had not experienced POPF in both cohorts. The AUC of the Rad-score was statistically higher than the Fistula Risk Score for predicting POPF in both the training and validation cohort.

Conclusion

A novel radiomics-based formula was developed and validated for predicting POPF in patients who underwent PD, which provides a new method for identifying POPF risks and may help to improve informed decision-making in the prevention of POPF at low cost.

Predicting the Development of Normal-Appearing White Matter With Radiomics in the Aging Brain: A Longitudinal Clinical Study

Background: Normal-appearing white matter (NAWM) refers to the normal, yet diseased tissue around the white matter hyperintensities (WMH) on conventional MR images. Radiomics is an emerging quantitative imaging technique that provides more details than a traditional visual analysis. This study aims to explore whether WMH could be predicted during the early stages of NAWM, using a textural analysis in the general elderly population.

Methods: Imaging data were obtained from PACS between 2012 and 2017. The subjects (≥60 years) received two or more MRI exams on the same scanner with time intervals of more than 1 year. By comparing the baseline and follow-up images, patients with noted progression of WMH were included as the case group (n = 51), while age-matched subjects without WMH were included as the control group (n = 51). Segmentations of the regions of interest (ROIs) were done with the ITK software. Two ROIs of developing NAWM (dNAWM) and non-developing NAWM (non-dNAWM) were drawn separately on the FLAIR images of each patient. dNAWM appeared normal on the baseline images, yet evolved into WMH on the follow-up images. Non-dNAWM appeared normal on both the baseline and follow-up images. A third ROI of normal white matter (NWM) was extracted from the control group, which was normal on both baseline and follow-up images. Textural features were dimensionally reduced with ANOVA+MW, correlation analysis, and LASSO. Three models were built based on the optimal parameters of dimensional reduction, including Model 1 (NWM vs. dNAWM), Model 2 (non-dNAWM vs. dNAWM), and Model 3 (NWM vs. non-dNAWM). The ROC curve was adopted to evaluate the classification validity of these models.

Results: Basic characteristics of the patients and controls showed no significant differences. The AUC of Model 1 in training and test groups were 0.967 (95% CI: 0.831–0.999) and 0.954 (95% CI: 0.876–0.989), respectively. The AUC of Model 2 were 0.939 (95% CI: 0.856–0.982) and 0.846 (95% CI: 0.671–0.950). The AUC of Model 3 were 0.713 (95% CI: 0.593–0.814) and 0.667 (95% CI: 0.475–0.825).

Conclusion: Radiomics textural analysis can distinguish dNAWM from non-dNAWM on FLAIR images, which could be used for the early detection of NAWM lesions before they develop into visible WHM.

Classification of the glioma grading using radiomics analysis

Background

Grading of gliomas is critical information related to prognosis and survival. We aimed to apply a radiomics approach using various machine learning classifiers to determine the glioma grading.

Methods

We considered 285 (high grade n = 210, low grade n = 75) cases obtained from the Brain Tumor Segmentation 2017 Challenge. Manual annotations of enhancing tumors, non-enhancing tumors, necrosis, and edema were provided by the database. Each case was multi-modal with T1-weighted, T1-contrast enhanced, T2-weighted, and FLAIR images. A five-fold cross validation was adopted to separate the training and test data. A total of 468 radiomics features were calculated for three types of regions of interest. The minimum redundancy maximum relevance algorithm was used to select features useful for classifying glioma grades in the training cohort. The selected features were used to build three classifier models of logistics, support vector machines, and random forest classifiers. The classification performance of the models was measured in the training cohort using accuracy, sensitivity, specificity, and area under the curve (AUC) of the receiver operating characteristic curve. The trained classifier models were applied to the test cohort.

Results

Five significant features were selected for the machine learning classifiers and the three classifiers showed an average AUC of 0.9400 for training cohorts and 0.9030 (logistic regression 0.9010, support vector machine 0.8866, and random forest 0.9213) for test cohorts.

Discussion

Glioma grading could be accurately determined using machine learning and feature selection techniques in conjunction with a radiomics approach. The results of our study might contribute to high-throughput computer aided diagnosis system for gliomas.

Radiomics and liquid biopsy in oncology: the holons of systems medicine

Abstract

Radiomics is a process of extraction and analysis of quantitative features from diagnostic images. Liquid biopsy is a test done on a sample of blood to look for cancer cells or for pieces of tumourigenic DNA circulating in the blood. Radiomics and liquid biopsy have great potential in oncology, since both are minimally invasive, easy to perform, and can be repeated in patient follow-up visits, enabling the extraction of valuable information regarding tumour type, aggressiveness, progression, and response to treatment. Both methods are in their infancy, with major evidence of application in lung and gastrointestinal cancer, while still undergoing evaluation in other cancer types. In this paper, the main oncologic applications of radiomics and liquid biopsy are reviewed, and a synergistic approach incorporating both tests for cancer diagnosis and follow-up is discussed within the context of systems medicine.

Teaching Points

• Radiomics is a process of extraction and analysis of quantitative features from diagnostic images.

• Most clinical applications of radiomics are in the field of oncologic imaging.

• Radiomics applies to all imaging modalities.

• A cluster of radiomic features is a “radiomic signature”.

• Machine learning may improve the efficacy of radiomics analysis.

Differentiation between pilocytic astrocytoma and glioblastoma: a decision tree model using contrast-enhanced magnetic resonance imaging-derived quantitative radiomic features

OBJECTIVE

To differentiate brain pilocytic astrocytoma (PA) from glioblastoma (GBM) using contrast-enhanced magnetic resonance imaging (MRI) quantitative radiomic features by a decision tree model.

METHODS

Sixty-six patients from two centres (PA, n = 31; GBM, n = 35) were randomly divided into training and validation data sets (about 2:1). Quantitative radiomic features of the tumours were extracted from contrast-enhanced MR images. A subset of features was selected by feature stability and Boruta algorithm. The selected features were used to build a decision tree model. Predictive accuracy, sensitivity and specificity were used to assess model performance. The classification outcome of the model was combined with tumour location, age and gender features, and multivariable logistic regression analysis and permutation test using the entire data set were performed to further evaluate the decision tree model.

RESULTS

A total of 271 radiomic features were successfully extracted for each tumour. Twelve features were selected as input variables to build the decision tree model. Two features S(1, -1) Entropy and S(2, -2) SumAverg were finally included in the model. The model showed an accuracy, sensitivity and specificity of 0.87, 0.90 and 0.83 for the training data set and 0.86, 0.80 and 0.91 for the validation data set. The classification outcome of the model related to the actual tumour types and did not rely on the other three features (p < 0.001).

CONCLUSIONS

A decision tree model with two features derived from the contrast-enhanced MR images performed well in differentiating PA from GBM.

KEY POINTS

• MRI findings of PA and GBM are sometimes very similar. • Radiomics provides much more quantitative information about tumours. • Radiomic features can help to distinguish PA from GBM.

Data-driven, voxel-based analysis of brain PET images: Application of PCA and LASSO methods to visualize and quantify patterns of neurodegeneration

Spatial patterns of radiotracer binding in positron emission tomography (PET) images may convey information related to the disease topology. However, this information is not captured by the standard PET image analysis that quantifies the mean radiotracer uptake within a region of interest (ROI). On the other hand, spatial analyses that use more advanced radiomic features may be difficult to interpret. Here we propose an alternative data-driven, voxel-based approach to spatial pattern analysis in brain PET, which can be easily interpreted. We apply principal component analysis (PCA) to identify voxel covariance patterns, and optimally combine several patterns using the least absolute shrinkage and selection operator (LASSO). The resulting models predict clinical disease metrics from raw voxel values, allowing for inclusion of clinical covariates. The analysis is performed on high-resolution PET images from healthy controls and subjects affected by Parkinson’s disease (PD), acquired with a pre-synaptic and a post-synaptic dopaminergic PET tracer. We demonstrate that PCA identifies robust and tracer-specific binding patterns in sub-cortical brain structures; the patterns evolve as a function of disease progression. Principal component LASSO (PC-LASSO) models of clinical disease metrics achieve higher predictive accuracy compared to the mean tracer binding ratio (BR) alone: the cross-validated test mean squared error of adjusted disease duration (motor impairment score) was 16.3 ± 0.17 years² (9.7 ± 0.15) with mean BR, versus 14.4 ± 0.18 years² (8.9 ± 0.16) with PC-LASSO. We interpret the best-performing PC-LASSO models in the spatial sense and discuss them with reference to the PD pathology and somatotopic organization of the striatum. PC-LASSO is thus shown to be a useful method to analyze clinically-relevant tracer binding patterns, and to construct interpretable, imaging-based predictive models of clinical metrics.

Probabilistic classification of tumour habitats in soft tissue sarcoma

The purpose of this work is to propose a method to characterize tumour heterogeneity on MRI, using probabilistic classification based on a reference tissue. The method uses maps of the apparent diffusion coefficient (ADC), T₂ relaxation, and a calculated map representing high-b-value diffusion-weighted MRI (denoted simDWI) to identify up to five habitats (i.e. sub-regions) of tumours. In this classification method, the parameter values (ADC, T₂ , and simDWI) from each tumour voxel are compared against the corresponding parameter probability distributions in a reference tissue. The probability that a tumour voxel belongs to a specific habitat is the joint probability for all parameters. The classification can be visualized using a custom colour scheme. The proposed method was applied to data from seven patients with biopsy-confirmed soft tissue sarcoma, at three time-points over the course of pre-operative radiotherapy. Fast-spin-echo images with two different echo times and diffusion MRI with three b-values were obtained and used as inputs to the method. Imaging findings were compared with pathology reports from pre-radiotherapy biopsy and post-surgical resection. Regions of hypercellularity, high-T₂ proteinaceous fluid, necrosis, collagenous stroma, and fibrosis were identified within soft tissue sarcoma. The classifications were qualitatively consistent with pathological observations. The percentage of necrosis on imaging correlated strongly with necrosis estimated from FDG-PET before radiotherapy (R²  = 0.97) and after radiotherapy (R²  = 0.96). The probabilistic classification method identifies realistic habitats and reflects the complex microenvironment of tumours, as demonstrated in soft tissue sarcoma.

Morphological MRI-based features provide pretreatment survival prediction in glioblastoma

OBJECTIVES

We wished to determine whether tumor morphology descriptors obtained from pretreatment magnetic resonance images and clinical variables could predict survival for glioblastoma patients.

METHODS

A cohort of 404 glioblastoma patients (311 discoveries and 93 validations) was used in the study. Pretreatment volumetric postcontrast T1-weighted magnetic resonance images were segmented to obtain the relevant morphological measures. Kaplan-Meier, Cox proportional hazards, correlations, and Harrell's concordance indexes (c-indexes) were used for the statistical analysis.

RESULTS

A linear prognostic model based on the outstanding variables (age, contrast-enhanced (CE) rim width, and surface regularity) identified a group of patients with significantly better survival (p < 0.001, HR = 2.57) with high accuracy (discovery c-index = 0.74; validation c-index = 0.77). A similar model applied to totally resected patients was also able to predict survival (p < 0.001, HR = 3.43) with high predictive value (discovery c-index = 0.81; validation c-index = 0.92). Biopsied patients with better survival were well identified (p < 0.001, HR = 7.25) by a model including age and CE volume (c-index = 0.87).

CONCLUSIONS

Simple linear models based on small sets of meaningful MRI-based pretreatment morphological features and age predicted survival of glioblastoma patients to a high degree of accuracy. The partition of the population using the extent of resection improved the prognostic value of those measures.

KEY POINTS

• A combination of two MRI-based morphological features (CE rim width and surface regularity) and patients' age outperformed previous prognosis scores for glioblastoma. • Prognosis models for homogeneous surgical procedure groups led to even more accurate survival prediction based on Kaplan-Meier analysis and concordance indexes.

Radiomics based on multicontrast MRI can precisely differentiate among glioma subtypes and predict tumour-proliferative behaviour

PURPOSE

To explore the feasibility and diagnostic performance of radiomics based on anatomical, diffusion and perfusion MRI in differentiating among glioma subtypes and predicting tumour proliferation.

METHODS

220 pathology-confirmed gliomas and ten contrasts were included in the retrospective analysis. After being registered to T2FLAIR images and resampling to 1 mm³ isotropically, 431 radiomics features were extracted from each contrast map within a semi-automatic defined tumour volume. For single-contrast and the combination of all contrasts, correlations between the radiomics features and pathological biomarkers were revealed by partial correlation analysis, and multivariate models were built to identify the best predictive models with adjusted 0.632+ bootstrap AUC.

RESULTS

In univariate analysis, both non-wavelet and wavelet radiomics features were correlated significantly with tumour grade and the Ki-67 labelling index. The max R was 0.557 (p = 2.04E-14) in T₁C for tumour grade and 0.395 (p = 2.33E-07) in ADC for Ki-67. In the multivariate analysis, the combination of all-contrast radiomics features had the highest AUCs in both differentiating among glioma subtypes and predicting proliferation compared with those in single-contrast images. For low-/high-grade gliomas, the best AUC was 0.911. In differentiating among glioma subtypes, the best AUC was 0.896 for grades II-III, 0.997 for grades II-IV, and 0.881 for grades III-IV. In predicting proliferation levels, multicontrast features led to an AUC of 0.936.

CONCLUSION

Multicontrast radiomics supplies complementary information on both geometric characters and molecular biological traits, which correlated significantly with tumour grade and proliferation. Combining all-contrast radiomics models might precisely predict glioma biological behaviour, which may be attributed to presurgical personal diagnosis.

KEY POINTS

• Multicontrast MRI radiomics features are significantly correlated with tumour grade and Ki-67 LI. • Multimodality MRI provides independent but supplemental information in assessing glioma pathological behaviour. • Combined multicontrast MRI radiomics can precisely predict glioma subtypes and proliferation levels.

Lung tumor segmentation methods: Impact on the uncertainty of radiomics features for non-small cell lung cancer

Purpose

To evaluate the uncertainty of radiomics features from contrast-enhanced breath-hold helical CT scans of non-small cell lung cancer for both manual and semi-automatic segmentation due to intra-observer, inter-observer, and inter-software reliability.

Methods

Three radiation oncologists manually delineated lung tumors twice from 10 CT scans using two software tools (3D-Slicer and MIM Maestro). Additionally, three observers without formal clinical training were instructed to use two semi-automatic segmentation tools, Lesion Sizing Toolkit (LSTK) and GrowCut, to delineate the same tumor volumes. The accuracy of the semi-automatic contours was assessed by comparison with physician manual contours using Dice similarity coefficients and Hausdorff distances. Eighty-three radiomics features were calculated for each delineated tumor contour. Informative features were identified based on their dynamic range and correlation to other features. Feature reliability was then evaluated using intra-class correlation coefficients (ICC). Feature range was used to evaluate the uncertainty of the segmentation methods.

Results

From the initial set of 83 features, 40 radiomics features were found to be informative, and these 40 features were used in the subsequent analyses. For both intra-observer and inter-observer reliability, LSTK had higher reliability than GrowCut and the two manual segmentation tools. All observers achieved consistently high ICC values when using LSTK, but the ICC value varied greatly for each observer when using GrowCut and the manual segmentation tools. For inter-software reliability, features were not reproducible across the software tools for either manual or semi-automatic segmentation methods. Additionally, no feature category was found to be more reproducible than another feature category. Feature ranges of LSTK contours were smaller than those of manual contours for all features.

Conclusion

Radiomics features extracted from LSTK contours were highly reliable across and among observers. With semi-automatic segmentation tools, observers without formal clinical training were comparable to physicians in evaluating tumor segmentation.

Comparison of radiomics machine-learning classifiers and feature selection for differentiation of sacral chordoma and sacral giant cell tumour based on 3D computed tomography features

OBJECTIVE

We aimed to identify optimal machine-learning methods for preoperative differentiation of sacral chordoma (SC) and sacral giant cell tumour (SGCT) based on 3D non-enhanced computed tomography (CT) and CT-enhanced (CTE) features.

METHODS

A total of 95 patients were divided into a training set and a validation set. Three best feature selection methods (Relief, least absolute shrinkage and selection operator (LASSO) and Random Forest (RF)) and three classification methods, including generalised linear models (GLM), support vector machines (SVM) and RF, were compared for their performance in distinguishing SC and SGCT. The performance of the radiomics model was investigated via area under the receiver-operating characteristic curve (AUC) and accuracy (ACC) analysis.

RESULTS

The selection method LASSO + classifier GLM had the highest AUC of 0.984 and ACC of 0.897 in the validating set, followed by Relief + GLM (AUC = 0.909, ACC = 0.862) and LASSO + SVM (AUC = 0.900, ACC = 0.862) based on CTE features. For CT features, RF + GLM had the highest AUC of 0.889, while LASSO + GLM achieved a high ACC of 0.793 in the validating set. Regardless of the methods, CTE features significantly outperformed those from CT for the differentiation of SC and SGCT (Z_AUC = -3.029, Z_ACC = -4.553; p < 0.05).

CONCLUSIONS

Our study demonstrated CTE features performed better than CT features. The selection method LASSO + classifier GLM had the best performance in differentiation of SC and SGCT, which could enhance the application of radiomics methods in sacral tumours.

KEY POINTS

• Sacral chordoma and sacral giant cell tumour are the two most common primary tumours of the sacrum with many common clinical and imaging characteristics. • A radiomics model helps clinicians to identify the histology of a sacral tumour. • CTE features should be preferred.

Correlation between radiomic features based on contrast-enhanced computed tomography images and Ki-67 proliferation index in lung cancer: A preliminary study

BACKGROUND

The purpose of the study was to investigate the association between radiomic features based on contrast-enhanced multidetector computed tomography (CT) and the Ki-67 proliferation index (PI) in patients with lung cancer.

METHODS

One hundred and ten patients with lung cancer confirmed by surgical histology were retrospectively included. Radiomic features were extracted from preoperative contrast-enhanced chest multidetector CT images for each tumor using open-source three-dimensional Slicer software. Statistical analysis was performed to determine significant radiomic features serving as image predictors of Ki-67 status in lung cancer and to investigate the relationship between these features and Ki-67 PI.

RESULTS

Higher Ki-67 expression was more common in men (P = 0.02) and patients with a smoking history (P = 0.01). Twelve radiomic features were significantly associated with Ki-67 status. Multivariate logistic regression analysis identified inverse variance, minor axis, and elongation as independent predictors of Ki-67 PI. There was a positive correlation between inverse variance, minor axis, elongation (P = 0.00, P = 0.02, and P = 0.14, respectively) and Ki-67 PI. The area under the curve to identify high Ki-67 status for inverse variance was 0.77 with a cutoff value of 0.47, which was significantly higher than for minor axis and elongation (P = 0.02 and P = 0.03, respectively).

CONCLUSION

Radiomic features based on contrast CT images, including inverse variance, minor axis, and elongation, can serve as noninvasive predictors of Ki-67 status in patients with lung cancer. Inverse variance could be superior to the other radiomic features to identify high Ki-67 status.

Application of Radiomics and Decision Support Systems for Breast MR Differential Diagnosis

Over the years, MR systems have evolved from imaging modalities to advanced computational systems producing a variety of numerical parameters that can be used for the noninvasive preoperative assessment of breast pathology. Furthermore, the combination with state-of-the-art image analysis methods provides a plethora of quantifiable imaging features, termed radiomics that increases diagnostic accuracy towards individualized therapy planning. More importantly, radiomics can now be complemented by the emerging deep learning techniques for further process automation and correlation with other clinical data which facilitate the monitoring of treatment response, as well as the prediction of patient's outcome, by means of unravelling of the complex underlying pathophysiological mechanisms which are reflected in tissue phenotype. The scope of this review is to provide applications and limitations of radiomics towards the development of clinical decision support systems for breast cancer diagnosis and prognosis.

Novel Quantitative PET Techniques for Clinical Decision Support in Oncology

Quantitative image analysis has deep roots in the usage of positron emission tomography (PET) in clinical and research settings to address a wide variety of diseases. It has been extensively employed to assess molecular and physiological biomarkers in vivo in healthy and disease states, in oncology, cardiology, neurology, and psychiatry. Quantitative PET allows relating the time-varying activity concentration in tissues/organs of interest and the basic functional parameters governing the biological processes being studied. Yet, quantitative PET is challenged by a number of degrading physical factors related to the physics of PET imaging, the limitations of the instrumentation used, and the physiological status of the patient. Moreover, there is no consensus on the most reliable and robust image-derived PET metric(s) that can be used with confidence in clinical oncology owing to the discrepancies between the conclusions reported in the literature. There is also increasing interest in the use of artificial intelligence based techniques, particularly machine learning and deep learning techniques in a variety of applications to extract quantitative features (radiomics) from PET including image segmentation and outcome prediction in clinical oncology. These novel techniques are revolutionizing clinical practice and are now offering unique capabilities to the clinical molecular imaging community and biomedical researchers at large. In this report, we summarize recent developments and future tendencies in quantitative PET imaging and present example applications in clinical decision support to illustrate its potential in the context of clinical oncology.

Novel Nomogram for Preoperative Prediction of Early Recurrence in Intrahepatic Cholangiocarcinoma

Introduction: The emerging field of "radiomics" has considerable potential in disease diagnosis, pathologic grading, prognosis evaluation, and prediction of treatment response. We aimed to develop a novel radiomics nomogram based on radiomics features and clinical characteristics that could preoperatively predict early recurrence (ER) of intrahepatic cholangiocarcinoma (ICC) after partial hepatectomy. Methods: A predictive model was developed from a training cohort comprising 139 ICC patients diagnosed between January 2010 and June 2014. Radiomics features were extracted from arterial-phase image of contrast-enhanced magnetic resonance imaging. Feature selection and construction of a "radiomics signature" were through Spearman's rank correlation and least absolute shrinkage and selection operator (LASSO) logistic regression. Combined with clinical characteristics, a radiomics nomogram was developed with multivariable logistic regression. Performance of the nomogram was evaluated with regard to discrimination, calibration, and clinical utility. An independent validation cohort involving 70 patients recruited from July 2014 to March 2016 was used to evaluate the utility of the nomogram developed. Results: The radiomics signature, consisting of nine features, differed significantly between ER patients and non-ER patients in training and validation cohorts. The area under the curve (AUC) of the radiomics signature in training and validation cohorts was 0.82 (confidence interval [CI], 0.74-0.88) and 0.77 (95% CI, 0.65-0.86), respectively. The AUC of the radiomics nomogram combining the radiomics signature and clinical stage in the two cohorts was 0.90 (95%CI, 0.83-0.94) and 0.86 (95% CI, 0.76-0.93), respectively. Decision curve analysis confirmed the clinical usefulness of the radiomics nomogram. Conclusion: The non-invasive radiomics nomogram developed using the radiomics signature and clinical stage could be used to predict ER of ICC after partial hepatectomy.

A Deep Look Into the Future of Quantitative Imaging in Oncology: A Statement of Working Principles and Proposal for Change

The adoption of enterprise digital imaging, along with the development of quantitative imaging methods and the re-emergence of statistical learning, has opened the opportunity for more personalized cancer treatments through transformative data science research. In the last 5 years, accumulating evidence has indicated that noninvasive advanced imaging analytics (i.e., radiomics) can reveal key components of tumor phenotype for multiple lesions at multiple time points over the course of treatment. Many groups using homegrown software have extracted engineered and deep quantitative features on 3-dimensional medical images for better spatial and longitudinal understanding of tumor biology and for the prediction of diverse outcomes. These developments could augment patient stratification and prognostication, buttressing emerging targeted therapeutic approaches. Unfortunately, the rapid growth in popularity of this immature scientific discipline has resulted in many early publications that miss key information or use underpowered patient data sets, without production of generalizable results. Quantitative imaging research is complex, and key principles should be followed to realize its full potential. The fields of quantitative imaging and radiomics in particular require a renewed focus on optimal study design and reporting practices, standardization, interpretability, data sharing, and clinical trials. Standardization of image acquisition, feature calculation, and statistical analysis (i.e., machine learning) are required for the field to move forward. A new data-sharing paradigm enacted among open and diverse participants (medical institutions, vendors and associations) should be embraced for faster development and comprehensive clinical validation of imaging biomarkers. In this review and critique of the field, we propose working principles and fundamental changes to the current scientific approach, with the goal of high-impact research and development of actionable prediction models that will yield more meaningful applications of precision cancer medicine.

Biochemical recurrence prediction after radiotherapy for prostate cancer with T2w magnetic resonance imaging radiomic features

Background and purpose

High-risk prostate cancer patients are frequently treated with external-beam radiotherapy (EBRT). Of all patients receiving EBRT, 15–35% will experience biochemical recurrence (BCR) within five years. Magnetic resonance imaging (MRI) is commonly acquired as part of the diagnostic procedure and imaging-derived features have shown promise in tumour characterisation and biochemical recurrence prediction. We investigated the value of imaging features extracted from pre-treatment T2w anatomical MRI to predict five year biochemical recurrence in high-risk patients treated with EBRT.

Materials and methods

In a cohort of 120 high-risk patients, imaging features were extracted from the whole-prostate and a margin surrounding it. Intensity, shape and textural features were extracted from the original and filtered T2w-MRI scans. The minimum-redundancy maximum-relevance algorithm was used for feature selection. Random forest and logistic regression classifiers were used in our experiments. The performance of a logistic regression model using the patient’s clinical features was also investigated. To assess the prediction accuracy we used stratified 10-fold cross validation and receiver operating characteristic analysis, quantified by the area under the curve (AUC).

Results

A logistic regression model built using whole-prostate imaging features obtained an AUC of 0.63 in the prediction of BCR, outperforming a model solely based on clinical variables (AUC = 0.51). Combining imaging and clinical features did not outperform the accuracy of imaging alone.

Conclusions

These results illustrate the potential of imaging features alone to distinguish patients with an increased risk of recurrence, even in a clinically homogeneous cohort.

Correlation between DCE-MRI radiomics features and Ki-67 expression in invasive breast cancer

The aim of the present study was to investigate the association between Ki-67 expression and radiomics features of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in patients with invasive breast cancer. A total of 53 cases with low-Ki-67 expression (Ki-67 proliferation index <14%) and 106 cases with high-Ki-67 expression (Ki-67 proliferation index >14%) were investigated. A systematic approach was applied that focused on the automated segmentation of lesions and extraction of radiomics features. For each lesion 5 morphology, 4 gray-scale histogram and 6 texture features were obtained, and statistical analyzes were performed to assess the differences in these features between the low- and high-Ki-67 expressions. One morphology metric (area), 3 gray-scale histogram indexes (standard deviation, skewness and kurtosis) and 3 texture features (contrast, homogeneity and inverse differential moment) demonstrated a significant difference (P<0.05), with low-Ki-67 expression lesions tending to be smaller, clearer and heterogeneous when compared with the high-Ki-67 expressed cases. These results may provide a noninvasive means to better understand the proliferation of breast cancer.

Imaging response assessment of immunotherapy in patients with renal cell and urothelial carcinoma

PURPOSE OF REVIEW

Recent advances in anticancer immunotherapy have revolutionized the treatment of metastatic renal cell (RCC) and urothelial carcinoma. In this review, we discuss the mechanisms of action of these new therapeutic approaches, explicate the common adverse events, and highlight different imaging-based response criteria.

RECENT FINDINGS

The recent introduction of immune-checkpoint inhibitors led to substantial advances in therapy of metastatic RCC and urothelial carcinoma. Because of the distinct effector mechanisms of these new substances, atypical response patterns such as transient enlargements of tumor lesions, appearance of new lesions after therapy, no measurable decrease in tumor size, or delayed responses are observed in medical imaging studies. This indicates that the established imaging-based response assessment according to the Response Evaluation Criteria in Solid Tumors (RECIST) guidelines has shortcomings to comprehensively evaluate treatment effects.

SUMMARY

While monitoring response to immunotherapy still relies on RECIST criteria, immune-related response criteria have been established to better address the imaging changes occurring under immunotherapy. Further studies with long-term follow-up are needed to properly identify and predict response after treatment beyond progression. Because of the expanding clinical use of immune checkpoint inhibitors, radiologists, urologist, and oncologists should be familiar with common imaging findings under this respective therapy.

Gold nanoparticle-induced sonosensitization enhances the antitumor activity of ultrasound in colon tumor-bearing mice

PURPOSE

As a noninvasive and nonionizing radiation, ultrasound can be focused remotely, transferring acoustic energy deep in the body, thereby addressing the penetration depth barrier of the light-based therapies. In cancer therapy, the effectiveness of ultrasound can be enhanced by utilizing nanomaterials that exhibit sonosensitizing properties called as nanosonosensitizers. The gold nanoparticle (AuNP) has been recently presented as a potent nanosonosensitizer with the potential to simultaneously enhance both the thermal and mechanical interactions of ultrasound with the tissue of the human body. Accordingly, this paper attempts to evaluate the in vivo antitumor efficiency of ultrasound in combination with AuNP.

METHODS

BALB/c mice-bearing CT26 colorectal tumor model was intraperitoneally injected with AuNPs and then subjected to ultrasound irradiation (1 MHz; 2 W/cm2 ; 10 min) for three sessions. Furthermore, [18 F]FDG (2-deoxy-2-[18 F]fluoro-d-glucose) positron-emission tomography (PET) imaging was performed and the radiomic features from different feature categorizes were extracted to quantify the tumors' phenotype.

RESULTS

The tumors were dramatically shrunk and the mice appeared healthy over 21 days of study span without the evidence of relapse. The animals treated with AuNP + ultrasound exhibited an obvious decline in tumor metabolic parameters such as standard uptake value (SUV), total lesion glycolysis (TLG), and metabolic tumor volume (MTV) compared to other treatment groups.

CONCLUSION

These findings support the use of AuNP as a potent sonosensitizing agent with the potential to use the thermal and mechanical effects of ultrasound so as to cause damage to the focused tumor site, resulting in an improved antitumor efficacy.

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.

Unsupervised CT Lung Image Segmentation of a Mycobacterium Tuberculosis Infection Model

Tuberculosis (TB) is an infectious disease caused by Mycobacterium tuberculosis that produces pulmonary damage. Radiological imaging is the preferred technique for the assessment of TB longitudinal course. Computer-assisted identification of biomarkers eases the work of the radiologist by providing a quantitative assessment of disease. Lung segmentation is the step before biomarker extraction. In this study, we present an automatic procedure that enables robust segmentation of damaged lungs that have lesions attached to the parenchyma and are affected by respiratory movement artifacts in a Mycobacterium Tuberculosis infection model. Its main steps are the extraction of the healthy lung tissue and the airway tree followed by elimination of the fuzzy boundaries. Its performance was compared with respect to a segmentation obtained using: (1) a semi-automatic tool and (2) an approach based on fuzzy connectedness. A consensus segmentation resulting from the majority voting of three experts' annotations was considered our ground truth. The proposed approach improves the overlap indicators (Dice similarity coefficient, 94% ± 4%) and the surface similarity coefficients (Hausdorff distance, 8.64 mm ± 7.36 mm) in the majority of the most difficult-to-segment slices. Results indicate that the refined lung segmentations generated could facilitate the extraction of meaningful quantitative data on disease burden.

Cancer risk assessment in modern radiotherapy workflow with medical big data

Modern radiotherapy (RT) is being enriched by big digital data and intensive technology. Multimodality image registration, intelligence-guided planning, real-time tracking, image-guided RT (IGRT), and automatic follow-up surveys are the products of the digital era. Enormous digital data are created in the process of treatment, including benefits and risks. Generally, decision making in RT tries to balance these two aspects, which is based on the archival and retrieving of data from various platforms. However, modern risk-based analysis shows that many errors that occur in radiation oncology are due to failures in workflow. These errors can lead to imbalance between benefits and risks. In addition, the exact mechanism and dose-response relationship for radiation-induced malignancy are not well understood. The cancer risk in modern RT workflow continues to be a problem. Therefore, in this review, we develop risk assessments based on our current knowledge of IGRT and provide strategies for cancer risk reduction. Artificial intelligence (AI) such as machine learning is also discussed because big data are transforming RT via AI.

The Role of Advanced Brain Tumor Imaging in the Care of Patients with Central Nervous System Malignancies

OPINION STATEMENT

T1-weighted post-contrast and T2-weighted fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) constitute the gold standard for diagnosis and response assessment in neuro-oncologic patients but are limited in their ability to accurately reflect tumor biology and metabolism, particularly over the course of a patient's treatment. Advanced MR imaging methods are sensitized to different biophysical processes in tissue, including blood perfusion, tumor metabolism, and chemical composition of tissue, and provide more specific information on tissue physiology than standard MRI. This review provides an overview of the most common and emerging advanced imaging modalities in the field of brain tumor imaging and their applications in the care of neuro-oncologic patients.

[Radiomics: Big Data Instead of Biopsies in the Future?]

Precision medicine is increasingly pushed forward, also with respect to upcoming new targeted therapies. Individual characterization of diseases on the basis of biomarkers is a prerequisite for this development. So far, biomarkers are characterized clinically, histologically or on a molecular level. The implementation of broad screening methods (“Omics”) and the analysis of big data – in addition to single markers – allow to define biomarker signatures. Next to “Genomics”, “Proteomics”, and “Metabolicis”, “Radiomics” gained increasing interest during the last years. Based on radiologic imaging, multiple radiomic markers are extracted with the help of specific algorithms. These are correlated with clinical, (immuno-) histopathological, or genomic data. Underlying structural differences are based on the imaging metadata and are often not visible and therefore not detectable without specific software. Radiomics are depicted numerically or by graphs. The fact that radiomic information can be extracted from routinely performed imaging adds a specific appeal to this method. Radiomics could potentially replace biopsies and additional investigations. Alternatively, radiomics could complement other biomarkers and thus lead to a more precise, multimodal prediction. Until now, radiomics are primarily used to investigate solid tumors. Some promising studies in head and neck cancer have already been published.

A radiomics evaluation of 2D and 3D MRI texture features to classify brain metastases from lung cancer and melanoma

Brain metastases are occasionally detected before diagnosing their primary site of origin. In these cases, simple visual examination of medical images of the metastases is not enough to identify the primary cancer, so an extensive evaluation is needed. To avoid this procedure, a radiomics approach on magnetic resonance (MR) images of the metastatic lesions is proposed to classify two of the most frequent origins (lung cancer and melanoma). In this study, 50 T1-weighted MR images of brain metastases from 30 patients were analyzed: 27 of lung cancer and 23 of melanoma origin. A total of 43 statistical texture features were extracted from the segmented lesions in 2D and 3D. Five predictive models were evaluated using a nested cross-validation scheme. The best classification results were achieved using 3D texture features for all the models, obtaining an average AUC > 0.9 in all cases and an AUC = 0.947 ± 0.067 when using the best model (naïve Bayes).

Radiomics derived from amino-acid PET and conventional MRI in patients with high-grade gliomas

Radiomics is a technique that uses high-throughput computing to extract quantitative features from tomographic medical images such as MRI and PET that usually are beyond visual perception. Importantly, the radiomics approach can be performed using neuroimages that have already been acquired during the routine follow-up of the patients allowing an additional data evaluation at low cost. In Neuro-Oncology, these features can potentially be used for differential diagnosis of newly diagnosed cerebral lesions suggestive for brain tumors or for the prediction of response to a neurooncological treatment option. Furthermore, especially in the light of the recent update of the World Health Organization classification of brain tumors, radiomics also has the potential to non-invasively assess important prognostic and predictive molecular markers such as a mutation in the isocitrate dehydrogenase gene or a 1p/19q codeletion which are not accessible by conventional visual interpretation of MRI or PET findings. This review summarizes the current status of the rapidly evolving field of radiomics with a special focus on patients with high-grade gliomas.

Magnetic resonance imaging (MRI)-based radiomics for prostate cancer radiotherapy

In radiotherapy (RT) of prostate cancer, dose escalation has been shown to reduce biochemical failure. Dose escalation only to determinate prostate tumor habitats has the potential to improve tumor control with less toxicity than when the entire prostate is dose escalated. Other issues in the treatment of the RT patient include the choice of the RT technique (hypo- or standard fractionation) and the use and length of concurrent/adjuvant androgen deprivation therapy (ADT). Up to 50% of high-risk men demonstrate biochemical failure suggesting that additional strategies for defining and treating patients based on improved risk stratification are required. The use of multiparametric MRI (mpMRI) is rapidly gaining momentum in the management of prostate cancer because of its improved diagnostic potential and its ability to combine functional and anatomical information. Currently, the Prostate Imaging, Reporting and Diagnosis System (PIRADS) is the standard of care for region of interest (ROI) identification and risk classification. However, PIRADS was not designed for 3D tumor volume delineation; there is a large degree of subjectivity and PIRADS does not accurately and reproducibly elucidate inter- and intra-lesional spatial heterogeneity. "Radiomics", as it refers to the extraction and analysis of large number of advanced quantitative radiological features from medical images using high throughput methods, is perfectly suited as an engine to effectively sift through the multiple series of prostate mpMRI sequences and quantify regions of interest. The radiomic efforts can be summarized in two main areas: (I) detection/segmentation of the suspicious lesion; and (II) assessment of the aggressiveness of prostate cancer. As related to RT, the goal of the latter is in particular to identify patients at high risk for metastatic disease; and the aim of the former is to identify and segment cancerous lesions and thus provide targets for radiation boost. The article is structured as follows: first, we describe the radiomic approach; and second, we discuss the radiomic pipeline as tailored for RT of prostate cancer. In this process we summarize the current efforts and progress in integrating mpMRI radiomics into the radiotherapeutic management of prostate cancer with emphasis placed on its role in treatment target definition, treatment plan strategizing, and prognostic assessment. The described concepts, methods and tools are not currently applicable to the radiation oncology practice outside of the research setting. More data are required in the form of clinical trials to assess the robustness of radiomics-based predictive models, and to maximize the efficacy of these models.

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.

Spatial Bayesian modeling of GLCM with application to malignant lesion characterization

The emerging field of cancer radiomics endeavors to characterize intrinsic patterns of tumor phenotypes and surrogate markers of response by transforming medical images into objects that yield quantifiable summary statistics to which regression and machine learning algorithms may be applied for statistical interrogation. Recent literature has identified clinicopathological association based on textural features deriving from gray-level co-occurrence matrices (GLCM) which facilitate evaluations of gray-level spatial dependence within a delineated region of interest. GLCM-derived features, however, tend to contribute highly redundant information. Moreover, when reporting selected feature sets, investigators often fail to adjust for multiplicities and commonly fail to convey the predictive power of their findings. This article presents a Bayesian probabilistic modeling framework for the GLCM as a multivariate object as well as describes its application within a cancer detection context based on computed tomography. The methodology, which circumvents processing steps and avoids evaluations of reductive and highly correlated feature sets, uses latent Gaussian Markov random field structure to characterize spatial dependencies among GLCM cells and facilitates classification via predictive probability. Correctly predicting the underlying pathology of 81% of the adrenal lesions in our case study, the proposed method outperformed current practices which achieved a maximum accuracy of only 59%. Simulations and theory are presented to further elucidate this comparison as well as ascertain the utility of applying multivariate Gaussian spatial processes to GLCM objects.

Classifying brain metastases by their primary site of origin using a radiomics approach based on texture analysis: a feasibility study

OBJECTIVE

To examine the capability of MRI texture analysis to differentiate the primary site of origin of brain metastases following a radiomics approach.

METHODS

Sixty-seven untreated brain metastases (BM) were found in 3D T1-weighted MRI of 38 patients with cancer: 27 from lung cancer, 23 from melanoma and 17 from breast cancer. These lesions were segmented in 2D and 3D to compare the discriminative power of 2D and 3D texture features. The images were quantized using different number of gray-levels to test the influence of quantization. Forty-three rotation-invariant texture features were examined. Feature selection and random forest classification were implemented within a nested cross-validation structure. Classification was evaluated with the area under receiver operating characteristic curve (AUC) considering two strategies: multiclass and one-versus-one.

RESULTS

In the multiclass approach, 3D texture features were more discriminative than 2D features. The best results were achieved for images quantized with 32 gray-levels (AUC = 0.873 ± 0.064) using the top four features provided by the feature selection method based on the p-value. In the one-versus-one approach, high accuracy was obtained when differentiating lung cancer BM from breast cancer BM (four features, AUC = 0.963 ± 0.054) and melanoma BM (eight features, AUC = 0.936 ± 0.070) using the optimal dataset (3D features, 32 gray-levels). Classification of breast cancer and melanoma BM was unsatisfactory (AUC = 0.607 ± 0.180).

CONCLUSION

Volumetric MRI texture features can be useful to differentiate brain metastases from different primary cancers after quantizing the images with the proper number of gray-levels.

KEY POINTS

• Texture analysis is a promising source of biomarkers for classifying brain neoplasms. • MRI texture features of brain metastases could help identifying the primary cancer. • Volumetric texture features are more discriminative than traditional 2D texture features.

Exploring Applications of Radiomics in Magnetic Resonance Imaging of Head and Neck Cancer: A Systematic Review

Background

Radiomics has been widely investigated for non-invasive acquisition of quantitative textural information from anatomic structures. While the vast majority of radiomic analysis is performed on images obtained from computed tomography, magnetic resonance imaging (MRI)-based radiomics has generated increased attention. In head and neck cancer (HNC), however, attempts to perform consistent investigations are sparse, and it is unclear whether the resulting textural features can be reproduced. To address this unmet need, we systematically reviewed the quality of existing MRI radiomics research in HNC.

Methods

Literature search was conducted in accordance with guidelines established by Preferred Reporting Items for Systematic Reviews and Meta-Analyses. Electronic databases were examined from January 1990 through November 2017 for common radiomic keywords. Eligible completed studies were then scored using a standardized checklist that we developed from Enhancing the Quality and Transparency of Health Research guidelines for reporting machine-learning predictive model specifications and results in biomedical research, defined by Luo et al. (1). Descriptive statistics of checklist scores were populated, and a subgroup analysis of methodology items alone was conducted in comparison to overall scores.

Results

Sixteen completed studies and four ongoing trials were selected for inclusion. Of the completed studies, the nasopharynx was the most common site of study (37.5%). MRI modalities varied with only four of the completed studies (25%) extracting radiomic features from a single sequence. Study sample sizes ranged between 13 and 118 patients (median of 40), and final radiomic signatures ranged from 2 to 279 features. Analyzed endpoints included either segmentation or histopathological classification parameters (44%) or prognostic and predictive biomarkers (56%). Liu et al. (2) addressed the highest number of our checklist items (total score: 48), and a subgroup analysis of methodology checklist items alone did not demonstrate any difference in scoring trends between studies [Spearman’s ρ = 0.94 (p < 0.0001)].

Conclusion

Although MRI radiomic applications demonstrate predictive potential in analyzing diverse HNC outcomes, methodological variances preclude accurate and collective interpretation of data.

Building CT Radiomics Based Nomogram for Preoperative Esophageal Cancer Patients Lymph Node Metastasis Prediction

PURPOSE: To build and validate a radiomics-based nomogram for the prediction of pre-operation lymph node (LN) metastasis in esophageal cancer. PATIENTS AND METHODS: A total of 197 esophageal cancer patients were enrolled in this study, and their LN metastases have been pathologically confirmed. The data were collected from January 2016 to May 2016; patients in the first three months were set in the training cohort, and patients in April 2016 were set in the validation cohort. About 788 radiomics features were extracted from computed tomography (CT) images of the patients. The elastic-net approach was exploited for dimension reduction and selection of the feature space. The multivariable logistic regression analysis was adopted to build the radiomics signature and another predictive nomogram model. The predictive nomogram model was composed of three factors with the radiomics signature, where CT reported the LN number and position risk level. The performance and usefulness of the built model were assessed by the calibration and decision curve analysis. RESULTS: Thirteen radiomics features were selected to build the radiomics signature. The radiomics signature was significantly associated with the LN metastasis (P<0.001). The area under the curve (AUC) of the radiomics signature performance in the training cohort was 0.806 (95% CI: 0.732-0.881), and in the validation cohort it was 0.771 (95% CI: 0.632-0.910). The model showed good discrimination, with a Harrell’s Concordance Index of 0.768 (0.672 to 0.864, 95% CI) in the training cohort and 0.754 (0.603 to 0.895, 95% CI) in the validation cohort. Decision curve analysis showed our model will receive benefit when the threshold probability was larger than 0.15. CONCLUSION: The present study proposed a radiomics-based nomogram involving the radiomics signature, so the CT reported the status of the suspected LN and the dummy variable of the tumor position. It can be potentially applied in the individual preoperative prediction of the LN metastasis status in esophageal cancer patients.

Pharmacometabolomics Informs Quantitative Radiomics for Glioblastoma Diagnostic Innovation

Applications of omics systems biology technologies have enormous promise for radiology and diagnostics in surgical fields. In this context, the emerging fields of radiomics (a systems scale approach to radiology using a host of technologies, including omics) and pharmacometabolomics (use of metabolomics for patient and disease stratification and guiding precision medicine) offer much synergy for diagnostic innovation in surgery, particularly in neurosurgery. This synthesis of omics fields and applications is timely because diagnostic accuracy in central nervous system tumors still challenges decision-making. Considering the vast heterogeneity in brain tumors, disease phenotypes, and interindividual variability in surgical and chemotherapy outcomes, we believe that diagnostic accuracy can be markedly improved by quantitative radiomics coupled to pharmacometabolomics and related health information technologies while optimizing economic costs of traditional diagnostics. In this expert review, we present an innovation analysis on a systems-level multi-omics approach toward diagnostic accuracy in central nervous system tumors. For this, we suggest that glioblastomas serve as a useful application paradigm. We performed a literature search on PubMed for articles published in English between 2006 and 2016. We used the search terms "radiomics," "glioblastoma," "biomarkers," "pharmacogenomics," "pharmacometabolomics," "pharmacometabonomics/pharmacometabolomics," "collaborative informatics," and "precision medicine." A list of the top 4 insights we derived from this literature analysis is presented in this study. For example, we found that (i) tumor grading needs to be better refined, (ii) diagnostic precision should be improved, (iii) standardization in radiomics is lacking, and (iv) quantitative radiomics needs to prove clinical implementation. We conclude with an interdisciplinary call to the metabolomics, pharmacy/pharmacology, radiology, and surgery communities that pharmacometabolomics coupled to information technologies (chemoinformatics tools, databases, collaborative systems) can inform quantitative radiomics, thus translating Big Data and information growth to knowledge growth, rational drug development and diagnostics innovation for glioblastomas, and possibly in other brain tumors.

Radiomics of CT Features May Be Nonreproducible and Redundant: Influence of CT Acquisition Parameters

Purpose To identify the reproducible and nonredundant radiomics features (RFs) for computed tomography (CT). Materials and Methods Two phantoms were used to test RF reproducibility by using test-retest analysis, by changing the CT acquisition parameters (hereafter, intra-CT analysis), and by comparing five different scanners with the same CT parameters (hereafter, inter-CT analysis). Reproducible RFs were selected by using the concordance correlation coefficient (as a measure of the agreement between variables) and the coefficient of variation (defined as the ratio of the standard deviation to the mean). Redundant features were grouped by using hierarchical cluster analysis. Results A total of 177 RFs including intensity, shape, and texture features were evaluated. The test-retest analysis showed that 91% (161 of 177) of the RFs were reproducible according to concordance correlation coefficient. Reproducibility of intra-CT RFs, based on coefficient of variation, ranged from 89.3% (151 of 177) to 43.1% (76 of 177) where the pitch factor and the reconstruction kernel were modified, respectively. Reproducibility of inter-CT RFs, based on coefficient of variation, also showed large material differences, from 85.3% (151 of 177; wood) to only 15.8% (28 of 177; polyurethane). Ten clusters were identified after the hierarchical cluster analysis and one RF per cluster was chosen as representative. Conclusion Many RFs were redundant and nonreproducible. If all the CT parameters are fixed except field of view, tube voltage, and milliamperage, then the information provided by the analyzed RFs can be summarized in only 10 RFs (each representing a cluster) because of redundancy.

Evaluation of prognostic models developed using standardised image features from different PET automated segmentation methods

Background

Prognosis in oesophageal cancer (OC) is poor. The 5-year overall survival (OS) rate is approximately 15%. Personalised medicine is hoped to increase the 5- and 10-year OS rates. Quantitative analysis of PET is gaining substantial interest in prognostic research but requires the accurate definition of the metabolic tumour volume. This study compares prognostic models developed in the same patient cohort using individual PET segmentation algorithms and assesses the impact on patient risk stratification.

Consecutive patients (n = 427) with biopsy-proven OC were included in final analysis. All patients were staged with PET/CT between September 2010 and July 2016. Nine automatic PET segmentation methods were studied. All tumour contours were subjectively analysed for accuracy, and segmentation methods with < 90% accuracy were excluded. Standardised image features were calculated, and a series of prognostic models were developed using identical clinical data. The proportion of patients changing risk classification group were calculated.

Results

Out of nine PET segmentation methods studied, clustering means (KM2), general clustering means (GCM3), adaptive thresholding (AT) and watershed thresholding (WT) methods were included for analysis. Known clinical prognostic factors (age, treatment and staging) were significant in all of the developed prognostic models. AT and KM2 segmentation methods developed identical prognostic models. Patient risk stratification was dependent on the segmentation method used to develop the prognostic model with up to 73 patients (17.1%) changing risk stratification group.

Conclusion

Prognostic models incorporating quantitative image features are dependent on the method used to delineate the primary tumour. This has a subsequent effect on risk stratification, with patients changing groups depending on the image segmentation method used.

Electronic supplementary material

The online version of this article (10.1186/s13550-018-0379-3) contains supplementary material, which is available to authorized users.

Diagnostic Accuracy of Unenhanced CT Analysis to Differentiate Low-Grade From High-Grade Chromophobe Renal Cell Carcinoma

OBJECTIVE

The objective of our study was to evaluate tumor attenuation and texture on unenhanced CT for potential differentiation of low-grade from high-grade chromophobe renal cell carcinoma (RCC).

MATERIALS AND METHODS

A retrospective study of 37 consecutive patients with chromophobe RCC (high-grade, n = 13; low-grade, n = 24) who underwent preoperative unenhanced CT between 2011 and 2016 was performed. Two radiologists (readers 1 and 2) blinded to the histologic grade of the tumor and outcome of the patients subjectively evaluated tumor homogeneity (3-point scale: completely homogeneous, mildly heterogeneous, or mostly heterogeneous). A third radiologist, also blinded to tumor grade and patient outcome, measured attenuation and contoured tumors for quantitative texture analysis. Comparisons were performed between high-grade and low-grade tumors using the chi-square test for subjective variables and sex, independent t tests for patient age and tumor attenuation, and Mann-Whitney U tests for texture analysis. Logistic regression models and ROC curves were computed.

RESULTS

There were no differences in age or sex between the groups (p = 0.652 and 0.076). High-grade tumors were larger (mean ± SD, 62.6 ± 34.9 mm [range, 17.0-141.0 mm] vs 39.0 ± 17.9 mm [16.0-72.3 mm]; p = 0.009) and had higher attenuation (mean ± SD, 45.5 ± 8.2 HU [range, 29.0-55.0 HU] vs 35.3 ± 8.5 HU [14.0-51.0 HU]; p = 0.001) than low-grade tumors. CT size and attenuation achieved good accuracy to diagnose high-grade chromophobe RCC: The AUC ± standard error was 0.85 ± 0.08 (p < 0.0001) with a sensitivity of 69.0% and a specificity of 100%. Subjectively, high-grade tumors were more heterogeneous (mildly or markedly heterogeneous: 69.2% [9/13] for reader 1 and 76.9% [10/13] for reader 2; reader 1, p = 0.024; reader 2, p = 0.001) with moderate agreement (κ = 0.57). Combined texture features diagnosed high-grade tumors with a maximal AUC of 0.84 ± 0.06 (p < 0.0001).

CONCLUSION

Tumor attenuation and heterogeneity assessed on unenhanced CT are associated with high-grade chromophobe RCC and correlate well with the histopathologic chromophobe tumor grading system.

Energy-Specific Optimization of Attenuation Thresholds for Low-Energy Virtual Monoenergetic Images in Renal Lesion Evaluation

OBJECTIVE

The purpose of this study was to determine in vitro and in vivo the optimal threshold for renal lesion vascularity at low-energy (40-60 keV) virtual monoenergetic imaging.

MATERIALS AND METHODS

A rod simulating unenhanced renal parenchymal attenuation (35 HU) was fitted with a syringe containing water. Three iodinated solutions (0.38, 0.57, and 0.76 mg I/mL) were inserted into another rod that simulated enhanced renal parenchyma (180 HU). Rods were inserted into cylindric phantoms of three different body sizes and scanned with single- and dual-energy MDCT. In addition, 102 patients (32 men, 70 women; mean age, 66.8 ± 12.9 [SD] years) with 112 renal lesions (67 nonvascular, 45 vascular) measuring 1.1-8.9 cm underwent single-energy unenhanced and contrast-enhanced dual-energy CT. Optimal threshold attenuation values that differentiated vascular from nonvascular lesions at 40-60 keV were determined.

RESULTS

Mean optimal threshold values were 30.2 ± 3.6 (standard error), 20.9 ± 1.3, and 16.1 ± 1.0 HU in the phantom, and 35.9 ± 3.6, 25.4 ± 1.8, and 17.8 ± 1.8 HU in the patients at 40, 50, and 60 keV. Sensitivity and specificity for the thresholds did not change significantly between low-energy and 70-keV virtual monoenergetic imaging (sensitivity, 87-98%; specificity, 90-91%). The AUC from 40 to 70 keV was 0.96 (95% CI, 0.93-0.99) to 0.98 (95% CI, 0.95-1.00).

CONCLUSION

Low-energy virtual monoenergetic imaging at energy-specific optimized attenuation thresholds can be used for reliable characterization of renal lesions.

Emergence of Radiomics: Novel Methodology Identifying Imaging Biomarkers of Disease in Diagnosis, Response, and Progression

Radiomics is an emerging area within clinical radiology research. It seeks to take full advantage of all the information contained in multiple medical imaging modalities. With a radiomics approach, medical images are not limited to providing only a qualitative assessment but can also provide quantitative data by parameterizing image features. These parameters can be used to identify regions and volumes of interest and discriminate normal healthy tissue from abnormal or diseased tissue. Radiomics is an interlinked sequence of processes of vital importance that begins with the acquisition and selection of medical images that involve standardization of acquisition protocols and inter-equipment normalization. This is followed by the identification and segmentation of regions or volumes of interest by expert radiologists through the use of computational tools that offer speed while reducing variability and bias. The segmentation process is the most critical stage in radiomics. This sometimes requires the incorporation of a pre-processing stage consisting of advanced techniques (reconstruction processes, filtering, etc.). Thereafter, representative characteristics of the region or volume of interest are extracted by approaches based on statistics, morphological features, and transform-based variables. Next, a statistical selection of the parameters that provide a high association and correlation with the clinical condition of interest is performed. Finally, processes such as data integration, standardization, classification, and mining processes can be applied as needed for particular applications. Ongoing research in radiomics aims to reduce the time and costs involved in interpreting medical images while simultaneously increasing the quality of diagnoses and monitoring of as well as the selection of treatment strategies. The results of many studies combining radiomics with standard medical techniques are highly encouraging, and these new approaches are increasingly used. This review article details the components of radiomics and discusses its applications, challenges, and future directions for this exciting new field of study.