Texture Radiomics Features Research Articles

Radiomics and quantitative multi-parametric MRI for predicting uterine fibroid growth.

Uterine fibroids (UFs) can pose a serious health risk to women. UFs are benign tumors that vary in clinical presentation from asymptomatic to causing debilitating symptoms. UF management is limited by our inability to predict UF growth rate and future morbidity. We aim to develop a predictive model to identify UFs with increased growth rates and possible resultant morbidity. We retrospectively analyzed 44 expertly outlined UFs from 20 patients who underwent two multi-parametric MR imaging exams as part of a prospective study over an average of 16 months. We identified 44 initial features by extracting quantitative magnetic resonance imaging (MRI) features plus morphological and textural radiomics features from DCE, T2, and apparent diffusion coefficient sequences. Principal component analysis reduced dimensionality, with the smallest number of components explaining over 97.5% of the variance selected. Employing a leave-one-fibroid-out scheme, a linear discriminant analysis classifier utilized these components to output a growth risk score. The classifier incorporated the first three principal components and achieved an area under the receiver operating characteristic curve of 0.80 (95% confidence interval [0.69; 0.91]), effectively distinguishing UFs growing faster than the median growth rate of from slower-growing ones within the cohort. Time-to-event analysis, dividing the cohort based on the median growth risk score, yielded a hazard ratio of 0.33 [0.15; 0.76], demonstrating potential clinical utility. We developed a promising predictive model utilizing quantitative MRI features and principal component analysis to identify UFs with increased growth rates. Furthermore, the model's discrimination ability supports its potential clinical utility in developing tailored patient and fibroid-specific management once validated on a larger cohort.

Radiomics-based machine learning approach for the prediction of grade and stage in upper urinary tract urothelial carcinoma-a step towards virtual biopsy.

Upper tract urothelial carcinoma is a rare, aggressive lesion, with early detection a key to its management. This study aimed to utilise computed tomographic urogram data to develop machine learning models for predicting tumour grading and staging in upper urothelial tract carcinoma patients and to compare these predictions with histopathological diagnosis used as reference standards. Protocol-based computed tomographic urogram data from 106 patients were obtained and visualised in 3D. Digital segmentation of the tumours was conducted by extracting textural radiomics features. They were further classified using 11 predictive models. The predicted grades and stages were compared to the histopathology of radical nephroureterectomy specimens. Classifier models worked well in mining the radiomics data and delivered satisfactory predictive machine learning models. The MultiLayer Panel showed 84% sensitivity and 93% specificity while predicting upper tract urothelial carcinoma grades. The Logistic Regression model showed a sensitivity of 83% and a specificity of 76% while staging. Similarly, other classifier algorithms (e.g., Support Vector classifier) provided a highly accurate prediction while grading upper tract urothelial carcinoma compared to clinical features alone or ureteroscopic biopsy histopathology. Data mining tools could handle medical imaging datasets from small (<2cm) tumours for upper tract urothelial carcinoma. The radiomics-based machine learning algorithms provide a potential tool to model tumour grading and staging with implications for clinical practice and the upgradation of current paradigms in cancer diagnostics. Machine learning based on radiomics features can predict upper tract urothelial cancer grading and staging with significant improvement over ureteroscopic histopathology. The study showcased the prowess of such emerging tools in the set objectives with implications towards virtual biopsy.

Abstract 7661: Integrating radiomic features from primary tumor and lymph node for survival prediction of HPV-associated oropharyngeal squamous cell carcinoma

Abstract Background: HPV-associated oropharyngeal squamous cell carcinoma (OPSCC) is a unique disease with improved prognosis which could benefit from treatment de-escalation. Current radiographic-based biomarkers mainly focus on primary tumor (PT) alone, neglecting the potential prognostic value from the lymph node (LN). In this work, we constructed a radiomic model integrating texture features from both PT and LN to predict overall survival (OS) in HPV-associated OPSCC patients. Methods: Pretreatment CT scans for 647 OPSCC patients treated with definitive radiation with or without chemotherapy were acquired from The Cancer Imaging Archive (TCIA) and the Cleveland Clinic (CCF), which were divided into training set D1 (n=294) and test set D2 (n=353). PT and LN contours were annotated by three. 2045 texture radiomic features were extracted from both PT and LN and were concatenated as input to a LASSO Cox model to predict OS. We used the harrell’s concordance index (C-index) and hazard ratio (HR) to evaluate the prognostic performance. We also performed the multivariable cox proportional hazard analysis adjusting for clinicopathological variables to validate the independent prognostic significance. Results: Among the 647 patients, median follow-up time was 70.23 months, with 155 patients (24%) experienced death. 22.6%, 38.8%, 23.2% and 15.4% had T stage I, II, III and IV tumors, respectively. 203 patients (31.4%) were treated with radiation alone. The model integrating PT and LN (M_PTLN) yielded higher C-index compared to PT alone (M_PT) on D1 (0.74 vs 0.66) and on D2 (0.64 vs 0.61). In univariate analysis on D2, M_PTLN is prognostic of OS (HR, 2.26; 95% CI, 1.32 - 3.87; log rank p = 0.003). In multivariable analysis, M_PTLN is still statistically significant (p = 0.04) accounting for clinicopathologic factors (age, gender, smoking pack-year [PY], T stage, N stage and treatment modality). Conclusions: We developed and validated a CT-based survival prediction model for HPV-associated OPSCC patients with significant prognostic values, accounting for both primary tumor and lymph node. If prospectively validated, the model has the potential to inform trial eligibility and treatment-related decision making. Citation Format: Bolin Song, Amaury Leroy, Kailin Yang, Qiyun Liu, Xiao Li, Jonathan Lee, Sarah Stock, Nabil F. Saba, James S. Lewis, Eric Deutsch, Mehmet Asim Bilen, Shlomo A. Koyfman, Anant Madabhushi. Integrating radiomic features from primary tumor and lymph node for survival prediction of HPV-associated oropharyngeal squamous cell carcinoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7661.

Radiomics using non-contrast CT to predict hemorrhagic transformation risk in stroke patients undergoing revascularization.

This study explores whether textural features from initial non-contrast CT scans of infarcted brain tissue are linked to hemorrhagic transformation susceptibility. Stroke patients undergoing thrombolysis or thrombectomy from Jan 2012 to Jan 2022 were analyzed retrospectively. Hemorrhagic transformation was defined using follow-up magnetic resonance imaging. A total of 94 radiomic features were extracted from the infarcted tissue on initial NCCT scans. Patients were divided into training and test sets (7:3 ratio). Two models were developed with fivefold cross-validation: one incorporating first-order and textural radiomic features, and another using only textural radiomic features. A clinical model was also constructed using logistic regression with clinical variables, and test set validation was performed. Among 362 patients, 218 had hemorrhagic transformations. The LightGBM model with all radiomics features had the best performance, with an area under the receiver operating characteristic curve (AUROC) of 0.986 (95% confidence interval [CI], 0.971-1.000) on the test dataset. The ExtraTrees model performed best when textural features were employed, with an AUROC of 0.845 (95% CI, 0.774-0.916). Minimum, maximum, and ten percentile values were significant predictors of hemorrhagic transformation. The clinical model showed an AUROC of 0.544 (95% CI, 0.431-0.658). The performance of the radiomics models was significantly better than that of the clinical model on the test dataset (p < 0.001). The radiomics model can predict hemorrhagic transformation using NCCT in stroke patients. Low Hounsfield unit was a strong predictor of hemorrhagic transformation, while textural features alone can predict hemorrhagic transformation. Using radiomic features extracted from initial non-contrast computed tomography, early prediction of hemorrhagic transformation has the potential to improve patient care and outcomes by aiding in personalized treatment decision-making and early identification of at-risk patients. • Predicting hemorrhagic transformation following thrombolysis in stroke is challenging since multiple factors are associated. • Radiomics features of infarcted tissue on initial non-contrast CT are associated with hemorrhagic transformation. • Textural features on non-contrast CT are associated with the frailty of the infarcted tissue.

Predicting rapid progression and overall survival in stage II-III pancreatic cancer using a CT-based radiomic signature.

706 Background: We sought to assess the ability of a computed tomography (CT)-derived radiomic signature (RS) feature set to predict rapid tumor progression and overall survival (OS) in patients diagnosed with stage II-III pancreatic cancer undergoing stereotactic body radiation therapy (SBRT) in sequence with chemotherapy. Methods: We conducted a retrospective study on a cohort of 240 stage II-III pancreatic cancer patients who underwent SBRT in sequence with chemotherapy at a single institution (Stanford Hospital and Clinics). Among them, 83 had pre-SBRT contrast-enhanced CT images with segmented tumors, forming the imaging set, from which we extracted 900 radiomic (quantitative pixel-level imaging characteristic) features from delineated tumor regions-of-interest. For RS development, we divided the imaging set into a training set (n = 53) for model development and a test set (n = 30) for evaluation. We built a binary prediction model on the training set to identify patients at risk of rapid tumor progression within three months after SBRT. We utilized logistic regression with the Least Absolute Shrinkage and Selection Operator algorithm for feature selection and classification. To fine-tune parameters, we performed five-fold cross-validation on the training set, repeating each set of parameters five times. Finally, we assessed the model's performance on the test set (n = 30) using the area under the curve (AUC) value. We selected the model with the best AUC, while also generating the predictive radiomic feature set. For OS prognostication, we conducted both univariate and multivariate Cox proportional-hazards analyses using RS and clinical features (age, sex, stage, vessel involvement, tumor location, performance status, body mass index, biological equivalent dose of radiation) as potential predictors. Results: The enrolled cohort consisted of 127 men (mean age, 69 years ± 12 [SD]) and 113 women (mean age, 69 years ± 13 [SD]). Seventeen textural radiomic features were identified as the RS, which demonstrated a high AUC in the test set for the prediction of rapid progression (AUC 0.850; 95% CI: 0.725, 0.975). Age was associated with OS in the multivariate model while high RS (&gt;0.5) was independently associated with OS in both univariate and multivariate models. Conclusions: CT-derived RS, combined with age, represent the most prognostic factors for stage II-III pancreatic cancer. RS, which likely reflects underlying biology or molecular alterations, contributes to the highest HR of poor OS and provide the strongest indicator of outcome following commonly used treatments in this patient population. [Table: see text]

Physics-Informed Discretization for Reproducible and Robust Radiomic Feature Extraction Using Quantitative MRI.

Given the limited repeatability and reproducibility of radiomic features derived from weighted magnetic resonance imaging (MRI), there may be significant advantages to using radiomics in conjunction with quantitative MRI. This study introduces a novel physics-informed discretization (PID) method for reproducible radiomic feature extraction and evaluates its performance using quantitative MRI sequences including magnetic resonance fingerprinting (MRF) and apparent diffusion coefficient (ADC) mapping. A multiscanner, scan-rescan dataset comprising whole-brain 3D quantitative (MRF T1, MRF T2, and ADC) and weighted MRI (T1w MPRAGE, T2w SPACE, and T2w FLAIR) from 5 healthy subjects was prospectively acquired. Subjects underwent 2 repeated acquisitions on 3 distinct 3 T scanners each, for a total of 6 scans per subject (30 total scans). First-order statistical (n = 23) and second-order texture (n = 74) radiomic features were extracted from 56 brain tissue regions of interest using the proposed PID method (for quantitative MRI) and conventional fixed bin number (FBN) discretization (for quantitative MRI and weighted MRI). Interscanner radiomic feature reproducibility was measured using the intraclass correlation coefficient (ICC), and the effect of image sequence (eg, MRF T1 vs T1w MPRAGE), as well as image discretization method (ie, PID vs FBN), on radiomic feature reproducibility was assessed using repeated measures analysis of variance. The robustness of PID and FBN discretization to segmentation error was evaluated by simulating segmentation differences in brainstem regions of interest. Radiomic features with ICCs greater than 0.75 following simulated segmentation were determined to be robust to segmentation. First-order features demonstrated higher reproducibility in quantitative MRI than weighted MRI sequences, with 30% (n = 7/23) features being more reproducible in MRF T1 and MRF T2 than weighted MRI. Gray level co-occurrence matrix (GLCM) texture features extracted from MRF T1 and MRF T2 were significantly more reproducible using PID compared with FBN discretization; for all quantitative MRI sequences, PID yielded the highest number of texture features with excellent reproducibility (ICC > 0.9). Comparing texture reproducibility of quantitative and weighted MRI, a greater proportion of MRF T1 (n = 225/370, 61%) and MRF T2 (n = 150/370, 41%) texture features had excellent reproducibility (ICC > 0.9) compared with T1w MPRAGE (n = 148/370, 40%), ADC (n = 115/370, 32%), T2w SPACE (n = 98/370, 27%), and FLAIR (n = 102/370, 28%). Physics-informed discretization was also more robust than FBN discretization to segmentation error, as 46% (n = 103/222, 46%) of texture features extracted from quantitative MRI using PID were robust to simulated 6 mm segmentation shift compared with 19% (n = 42/222, 19%) of weighted MRI texture features extracted using FBN discretization. The proposed PID method yields radiomic features extracted from quantitative MRI sequences that are more reproducible and robust than radiomic features extracted from weighted MRI using conventional (FBN) discretization approaches. Quantitative MRI sequences also demonstrated greater scan-rescan robustness and first-order feature reproducibility than weighted MRI.

A Radiomics Study: Classification of Breast Lesions by Textural Features from Mammography Images.

This study investigates the feasibility of using texture radiomics features extracted from mammography images to distinguish between benign and malignant breast lesions and to classify benign lesions into different categories and determine the best machine learning (ML) model to perform the tasks. Six hundred and twenty-two breast lesions from 200 retrospective patient data were segmented and analysed. Three hundred fifty radiomics features were extracted using the Standardized Environment for Radiomics Analysis (SERA) library, one of the radiomics implementations endorsed by the Image Biomarker Standardisation Initiative (IBSI). The radiomics features and selected patient characteristics were used to train selected machine learning models to classify the breast lesions. A fivefold cross-validation was used to evaluate the performance of the ML models and the top 10 most important features were identified. The random forest (RF) ensemble gave the highest accuracy (89.3%) and positive predictive value (66%) and likelihood ratio of 13.5 in categorising benign and malignant lesions. For the classification of benign lesions, the RF model again gave the highest likelihood ratio of 3.4 compared to the other models. Morphological and textural radiomics features were identified as the top 10 most important features from the random forest models. Patient age was also identified as one of the significant features in the RF model. We concluded that machine learning models trained against texture-based radiomics features and patient features give reasonable performance in differentiating benign versus malignant breast lesions. Our study also demonstrated that the radiomics-based machine learning models were able to emulate the visual assessment of mammography lesions, typically used by radiologists, leading to a better understanding of how the machine learning model arrive at their decision.

Radiogenomics Reveals Correlation between Quantitative Texture Radiomic Features of Biparametric MRI and Hypoxia-Related Gene Expression in Men with Localised Prostate Cancer

To perform multiscale correlation analysis between quantitative texture feature phenotypes of pre-biopsy biparametric MRI (bpMRI) and targeted sequence-based RNA expression for hypoxia-related genes. Images from pre-biopsy 3T bpMRI scans in clinically localised PCa patients of various risk categories (n = 15) were used to extract textural features. The genomic landscape of hypoxia-related gene expression was obtained using post-radical prostatectomy tissue for targeted RNA expression profiling using the TempO-sequence method. The nonparametric Games Howell test was used to correlate the differential expression of the important hypoxia-related genes with 28 radiomic texture features. Then, cBioportal was accessed, and a gene-specific query was executed to extract the Oncoprint genomic output graph of the selected hypoxia-related genes from The Cancer Genome Atlas (TCGA). Based on each selected gene profile, correlation analysis using Pearson's coefficients and survival analysis using Kaplan-Meier estimators were performed. The quantitative bpMR imaging textural features, including the histogram and grey level co-occurrence matrix (GLCM), correlated with three hypoxia-related genes (ANGPTL4, VEGFA, and P4HA1) based on RNA sequencing using the TempO-Seq method. Further radiogenomic analysis, including data accessed from the cBioportal genomic database, confirmed that overexpressed hypoxia-related genes significantly correlated with a poor survival outcomes, with a median survival ratio of 81.11:133.00 months in those with and without alterations in genes, respectively. This study found that there is a correlation between the radiomic texture features extracted from bpMRI in localised prostate cancer and the hypoxia-related genes that are differentially expressed. The analysis of expression data based on cBioportal revealed that these hypoxia-related genes, which were the focus of the study, are linked to an unfavourable survival outcomes in prostate cancer patients.

Correlation of Pre- and Post-radio-chemotherapy MRI Texture Features With Tumor Response in Rectal Cancer.

The present study aimed to investigate radiomics features derived from magnetic resonance imaging (MRI) in patients with locally advanced rectal cancer treated with neoadjuvant chemoradiotherapy (CRT). We retrospectively evaluated data of 53 patients (32 males, 21 females) with T3/T4 or N+ rectal cancer who underwent MRI before and after CRT. Twenty-seven texture radiomics features were extracted from regions of interest, delimiting the tumor on T2-weighted images. All 27 radiomics features extracted before CRT showed a statistically significant association with the tumor regression grade (TRG) (p<0.05), whereas, after CRT, only the Cluster Prominence value was the only variable to predict TRG (p=0.037, r=0.291). All 27 features extracted before CRT were able to predict response to CRT and Cluster Prominence continued to be statistically significant even after CRT. The impact of radiomics features derived from MRI could be further investigated in patients with locally advanced rectal cancer.

Single-timepoint low-dimensional characterization and classification of acute versus chronic multiple sclerosis lesions using machine learning

Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease characterized by the appearance of focal lesions across the central nervous system. The discrimination of acute from chronic MS lesions may yield novel biomarkers of inflammatory disease activity which may support patient management in the clinical setting and provide endpoints in clinical trials. On a single timepoint and in the absence of a prior reference scan, existing methods for acute lesion detection rely on the segmentation of hyperintense foci on post-gadolinium T1-weighted magnetic resonance imaging (MRI), which may underestimate recent acute lesion activity. In this paper, we aim to improve the sensitivity of acute MS lesion detection in the single-timepoint setting, by developing a novel machine learning approach for the automatic detection of acute MS lesions, using single-timepoint conventional non-contrast T1- and T2-weighted brain MRI. The MRI input data are supplemented via the use of a convolutional neural network generating “lesion-free” reconstructions from original “lesion-present” scans using image inpainting. A multi-objective statistical ranking module evaluates the relevance of textural radiomic features from the core and periphery of lesion sites, compared within “lesion-free” versus “lesion-present” image pairs. Then, an ensemble classifier is optimized through a recursive loop seeking consensus both in the feature space (via a greedy feature-pruning approach) and in the classifier space (via model selection repeated after each pruning operation). This leads to the identification of a compact textural signature characterizing lesion phenotype. On the patch-level task of acute versus chronic MS lesion classification, our method achieves a balanced accuracy in the range of 74.3–74.6% on fully external validation cohorts.

Intraoperative Evaluation of Brain-Tumor Microvascularization through MicroV IOUS: A Protocol for Image Acquisition and Analysis of Radiomic Features

Microvascular Doppler (MicroV) is a new-generation Doppler technique developed by Esaote (Esaote s.p.a., Genova, Italy), which is able to visualize small and low-flow vessels through a suppression of interfering signals. MicroV uses advanced filters that are able to differentiate tissue artifacts from low-speed blood flows; by exploiting the space-time coherence information, these filters can selectively suppress tissue components, preserving the signal coming from the microvascular flow. This technique is clinically applied to the study of the vascularization of parenchymatous lesions, often with better diagnostic accuracy than color/power Doppler techniques. The aim of this paper is to develop a reproducible protocol for the recording and collection of MicroV intraoperative ultrasound images by the use of a capable intraoperative ultrasound machine and post-processing aimed at evaluation of brain-tumor microvascularization through the analysis of radiomic features. The proposed protocol has been internally validated on eight patients and will be firstly applied to patients affected by WHO grade IV astrocytoma (glioblastoma-GBM) candidates for craniotomy and lesion removal. In a further stage, it will be generally applied to patients with primary or metastatic brain tumors. IOUS is performed before durotomy. Tumor microvascularization is evaluated using the MicroV Doppler technique and IOUS images are recorded, stored, and post-processed. IOUS images are remotely stored on the BraTIoUS database, which will promote international cooperation and multicentric analysis. Processed images and texture radiomic features are analyzed post-operatively using ImageJ, a free scientific image-analysis software based on the Sun-Java platform. Post-processing protocol is further described in-depth. The study of tumor microvascularization through advanced IOUS techniques such as MicroV could represent, in the future, a non-invasive and real-time method for intraoperative predictive evaluation of the tumor features. This evaluation could finally result in a deeper knowledge of brain-tumor behavior and in the on-going adaptation of the surgery with the improvement of surgical outcomes.

Reproducibility assessment of radiomics features in various ultrasound scan settings and different scanner vendors

IntroductionRadiomics in Ultrasound (US) imaging has been investigated for the prediction and prognosis of cancers. However, inter-scanner and intra-scanner variations may affect the reproducibility of radiomics results. This study aims to evaluate the reproducibility of US textural radiomics features across various scan settings and scanner vendors. Materials and methodsUS images in quality control (QC) phantom were obtained by three scanners (Philips, Samsung, and Siemens) with different scan settings and parameters. Circular regions of interest (ROIs) inside isoechoic, hypoechoic, and hyperechoic objects were manually delineated. Forty textural radiomics features were extracted from each ROI, and then the robust features that could distinguish different echogenic objects were obtained by the Mann-Whitney U test. Reproducibility of the robust radiomics features was assessed by the intraclass correlation coefficient (ICC) and coefficient of variation (CV); ICC>0.90 and %CV<20 were considered reproducible. ResultsAccording to the Mann-Whitney U test results, ten robust features could differentiate the hypoechoic, and 15 robust features could differentiate the hyperechoic objects from the isoechoic objects (P<0.001). The total ICC of the robust features for each echogenic object was >0.95 in different scanners and scan settings. Four and seven features were individually reproducible (%CV < 20, ICC>0.90) in hypoechoic and hyperechoic objects, respectively. Also, four features seem reproducible by changing the ROI location across the horizontal and vertical lines for both convex and linear array transducers. ConclusionsMost of the US textural radiomics features in this study were not reproducible. However, several features showed high reproducibility at different scan settings and scanners. These features may also be reproducible when ROI size and location change slightly.

Combination of pre-treatment dynamic [18F]FET PET radiomics and conventional clinical parameters for the survival stratification in patients with IDH-wildtype glioblastoma

PurposeThe aim of this study was to build and evaluate a prediction model which incorporates clinical parameters and radiomic features extracted from static as well as dynamic [18F]FET PET for the survival stratification in patients with newly diagnosed IDH-wildtype glioblastoma.MethodsA total of 141 patients with newly diagnosed IDH-wildtype glioblastoma and dynamic [18F]FET PET prior to surgical intervention were included. Patients with a survival time ≤ 12 months were classified as short-term survivors. First order, shape, and texture radiomic features were extracted from pre-treatment static (tumor-to-background ratio; TBR) and dynamic (time-to-peak; TTP) images, respectively, and randomly divided into a training (n = 99) and a testing cohort (n = 42). After feature normalization, recursive feature elimination was applied for feature selection using 5-fold cross-validation on the training cohort, and a machine learning model was constructed to compare radiomic models and combined clinical-radiomic models with selected radiomic features and clinical parameters. The area under the ROC curve (AUC), accuracy, sensitivity, specificity, and positive and negative predictive values were calculated to assess the predictive performance for identifying short-term survivors in both the training and testing cohort.ResultsA combined clinical-radiomic model comprising six clinical parameters and six selected dynamic radiomic features achieved highest predictability of short-term survival with an AUC of 0.74 (95% confidence interval, 0.60–0.88) in the independent testing cohort.ConclusionsThis study successfully built and evaluated prediction models using [18F]FET PET-based radiomic features and clinical parameters for the individualized assessment of short-term survival in patients with a newly diagnosed IDH-wildtype glioblastoma. The combination of both clinical parameters and dynamic [18F]FET PET–based radiomic features reached highest accuracy in identifying patients at risk. Although the achieved accuracy level remained moderate, our data shows that the integration of dynamic [18F]FET PET radiomic data into clinical prediction models may improve patient stratification beyond established prognostic markers.

Minimising multi-centre radiomics variability through image normalisation: a pilot study

Radiomics is an emerging technique for the quantification of imaging data that has recently shown great promise for deeper phenotyping of cardiovascular disease. Thus far, the technique has been mostly applied in single-centre studies. However, one of the main difficulties in multi-centre imaging studies is the inherent variability of image characteristics due to centre differences. In this paper, a comprehensive analysis of radiomics variability under several image- and feature-based normalisation techniques was conducted using a multi-centre cardiovascular magnetic resonance dataset. 218 subjects divided into healthy (n = 112) and hypertrophic cardiomyopathy (n = 106, HCM) groups from five different centres were considered. First and second order texture radiomic features were extracted from three regions of interest, namely the left and right ventricular cavities and the left ventricular myocardium. Two methods were used to assess features’ variability. First, feature distributions were compared across centres to obtain a distribution similarity index. Second, two classification tasks were proposed to assess: (1) the amount of centre-related information encoded in normalised features (centre identification) and (2) the generalisation ability for a classification model when trained on these features (healthy versus HCM classification). The results showed that the feature-based harmonisation technique ComBat is able to remove the variability introduced by centre information from radiomic features, at the expense of slightly degrading classification performance. Piecewise linear histogram matching normalisation gave features with greater generalisation ability for classification ( balanced accuracy in between 0.78 ± 0.08 and 0.79 ± 0.09). Models trained with features from images without normalisation showed the worst performance overall ( balanced accuracy in between 0.45 ± 0.28 and 0.60 ± 0.22). In conclusion, centre-related information removal did not imply good generalisation ability for classification.

Image resampling and discretization effect on the estimate of myocardial radiomic features from T1 and T2 mapping in hypertrophic cardiomyopathy

Radiomics is emerging as a promising and useful tool in cardiac magnetic resonance (CMR) imaging applications. Accordingly, the purpose of this study was to investigate, for the first time, the effect of image resampling/discretization and filtering on radiomic features estimation from quantitative CMR T1 and T2 mapping. Specifically, T1 and T2 maps of 26 patients with hypertrophic cardiomyopathy (HCM) were used to estimate 98 radiomic features for 7 different resampling voxel sizes (at fixed bin width), 9 different bin widths (at fixed resampling voxel size), and 7 different spatial filters (at fixed resampling voxel size/bin width). While we found a remarkable dependence of myocardial radiomic features from T1 and T2 mapping on image filters, many radiomic features showed a limited sensitivity to resampling voxel size/bin width, in terms of intraclass correlation coefficient (> 0.75) and coefficient of variation (< 30%). The estimate of most textural radiomic features showed a linear significant (p < 0.05) correlation with resampling voxel size/bin width. Overall, radiomic features from T2 maps have proven to be less sensitive to image preprocessing than those from T1 maps, especially when varying bin width. Our results might corroborate the potential of radiomics from T1/T2 mapping in HCM and hopefully in other myocardial diseases.

Radiomic features quantifying pixel-level characteristics of breast tumors from magnetic resonance imaging predict risk factors in triple-negative breast cancer.

e12612 Background: Computationally derived quantitative imaging (radiomic) features that describe tumor phenotypes at the pixel level have demonstrated associations with clinical characteristics in early investigations of other cancers. This implies that molecular differences among tumors may be reflected in their structure on the scales probed by 3D magnetic resonance imaging (MRI). We investigated whether radiomic features computed over tumor volumes from pre-treatment breast MRI could predict risk factors in triple-negative breast cancer (TNBC). Methods: We evaluated breast tumors on pre-treatment, post-contrast T1-weighted MRI from 156 patients with non-metastatic TNBC who underwent neoadjuvant chemotherapy. Tumor regions of interest were segmented by a convolutional neural network algorithm, with validation by breast radiologists. Features quantifying tumor shape and texture were extracted for the largest tumor present in each patient. We identified 23 principal components (PCs) describing these data within the original 364-dimensional feature space for further analysis. Tumor volume was also extracted for comparison with the shape and texture PCs, clinical variables and outcomes, but was kept separate from other radiomic features, since it directly correlates with clinical stage. We compiled for the cohort clinical variables including demographics, stage, grade, and, where available, absolute lymphocyte count (ALC) and Ki-67, a cellular proliferation index routinely used in clinical practice. We then performed a series of univariate and multivariate regression analyses to identify radiomic PCs and clinical variables that significantly predict patient outcomes, and radiomic PCs that predict established risk factors. Our multivariate analyses utilized 5-fold cross-validation and Monte-Carlo determination of p-values (based on 3000 random samplings from the null hypothesis), to ensure statistical rigor in identifying predictive relationships while correcting for multiple hypothesis testing. Results: Our univariate analyses confirmed expected correlations between: overall survival and pre-treatment tumor volume (p = 0.010); survival and ALC (p = 0.002); and clinical stage and tumor volume (p = 1.2⨉10-7). From our multivariate analysis, shape and texture radiomic features were predictive of: tumor volume (p &lt; 0.001); clinical stage (p &lt; 0.001); and Ki-67 (p = 0.02). We confirmed that Ki-67 was predictive of post-treatment residual cancer (p = 0.014), as has been previously reported. Conclusions: Radiomic features predict breast cancer risk factors that are significant for determining outcomes for TNBC patients. Combinations of radiomic shape and texture features track closely with tumor volumes, stage, and proliferative activity, potentially reflecting underlying molecular evolution.

CT-based morphologic and radiomics features for the classification of MYCN gene amplification status in pediatric neuroblastoma.

MYCN onco-gene amplification in neuroblastoma confers patients to the high-risk disease category for which prognosis is poor and more aggressive multimodal treatment is indicated. This retrospective study leverages machine learning techniques to develop a computed tomography (CT)-based model incorporating semantic and non-semantic features for non-invasive prediction of MYCN amplification status in pediatric neuroblastoma. From 2009 to 2020, 54 pediatric patients treated for neuroblastoma at a specialized children's hospital with pre-treatment contrast-enhanced CT and MYCN status were identified (training cohort, n = 44; testing cohort, n = 10). Six morphologic features and 107 quantitative gray-level texture radiomics features extracted from manually drawn volume-of-interest were analyzed. Following feature selection and class balancing, the final predictive model was developed with eXtreme Gradient Boosting (XGBoost) algorithm. Accumulated local effects (ALE) plots were used to explore main effects of the predictive features. Tumor texture maps were also generated for visualization of radiomics features. One morphologic and 2 radiomics features were selected for model building. The XGBoost model from the training cohort yielded an area under the receiver operating characteristics curve (AUC-ROC) of 0.930 (95% CI, 0.85-1.00), optimized F1-score of 0.878, and Matthews correlation coefficient (MCC) of 0.773. Evaluation on the testing cohort returned AUC-ROC of 0.880 (95% CI, 0.64-1.00), optimized F1-score of 0.933, and MCC of 0.764. ALE plots and texture maps showed higher "GreyLevelNonUniformity" values, lower "Strength" values, and higher number of image-defined risk factors contribute to higher predicted probability of MYCN amplification. The machine learning model reliably classified MYCN amplification in pediatric neuroblastoma and shows potential as a surrogate imaging biomarker.

COVID-19 prognostic modeling using CT radiomic features and machine learning algorithms: Analysis of a multi-institutional dataset of 14,339 patients

BackgroundWe aimed to analyze the prognostic power of CT-based radiomics models using data of 14,339 COVID-19 patients. MethodsWhole lung segmentations were performed automatically using a deep learning-based model to extract 107 intensity and texture radiomics features. We used four feature selection algorithms and seven classifiers. We evaluated the models using ten different splitting and cross-validation strategies, including non-harmonized and ComBat-harmonized datasets. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were reported. ResultsIn the test dataset (4,301) consisting of CT and/or RT-PCR positive cases, AUC, sensitivity, and specificity of 0.83 ± 0.01 (CI95%: 0.81–0.85), 0.81, and 0.72, respectively, were obtained by ANOVA feature selector + Random Forest (RF) classifier. Similar results were achieved in RT-PCR-only positive test sets (3,644). In ComBat harmonized dataset, Relief feature selector + RF classifier resulted in the highest performance of AUC, reaching 0.83 ± 0.01 (CI95%: 0.81–0.85), with a sensitivity and specificity of 0.77 and 0.74, respectively. ComBat harmonization did not depict statistically significant improvement compared to a non-harmonized dataset. In leave-one-center-out, the combination of ANOVA feature selector and RF classifier resulted in the highest performance. ConclusionLung CT radiomics features can be used for robust prognostic modeling of COVID-19. The predictive power of the proposed CT radiomics model is more reliable when using a large multicentric heterogeneous dataset, and may be used prospectively in clinical setting to manage COVID-19 patients.

Changes in Magnetic Resonance Imaging Radiomic Features in Response to Androgen Deprivation Therapy in Patients with Intermediate- and High-risk Prostate Cancer

AimsThe benefits of neoadjuvant androgen deprivation therapy (nADT) in the management of intermediate- and high-risk prostate cancer patients have been well-established. The aim of this study was to identify radiomic prognostic features derived from routine anatomic magnetic resonance imaging (MRI) sequences that can predict the response of the prostate cancer to nADT. Materials and methodsPatients with intermediate- and high-risk prostate cancer (with one of clinical stage ≥ T2c, Gleason score ≥7 or presenting prostate-specific antigen ≥10) who received 3 months of ADT prior to radical external beam radiotherapy were enrolled into this study. The relative blood volume and the relative blood flow were used as dynamic MRI kinetic parameters to quantify vascular changes as responses to nADT. For all pre- and post-nADT data sets, a combination of T2-weighted and contrast-enhanced T1-weighted anatomic images were used to define regions of interest (ROI) as the dominant malignant nodules (DMNs) and the benign prostate (the entire prostate with the summed DMNs being subtracted). MRI textural radiomic features associated with prostate cancer response in the literature of energy and homogeneity were selected. Pyradiomics was used to extract textural features of the ROIs. A Wilcoxon signed-rank test was carried out to investigate if there were statistically significant differences in values of radiomic features between: (i) benign prostate ROI and DMN pre-nADT; (ii) pre- and post-nADT of benign prostate ROI; (iii) pre- and post-nADT of DMN. Changes in radiomic features and dynamic MRI kinetic parameters were correlated using the Spearman correlation test. ResultsTwenty prostate cancer patients were recruited into the study. The median time between the first baseline scan and the first on-treatment scan was 91.5 days (range 82–105). One patient had no discernible tumour visible, leaving 19 patients with evaluable data for the analysis. Baseline homogeneity and energy values differed significantly between benign and malignant tissue (P < 0.01). In response to nADT, homogeneity and energy showed reciprocal changes, significantly increased in benign prostate while decreasing in the DMN. The reduction in tumour homogeneity and energy feature values showed a positive association with the decline in tumour blood flow and tumour blood volume induced by androgen deprivation as derived from dynamic MRI parameters. ConclusionEnergy and homogeneity radiomic features derived from MRI of benign and malignant prostate showed significant reciprocal changes in response to nADT. This study confirms the potential of these radiomic features to act as surrogate markers of tumour androgen sensitivity due to their strong association with ADT-induced physiological effects in prostate tumours.

Delta radiomics analysis of Magnetic Resonance guided radiotherapy imaging data can enable treatment response prediction in pancreatic cancer

BackgroundMagnetic Resonance Image guided Stereotactic body radiotherapy (MRgRT) is an emerging technology that is increasingly used in treatment of visceral cancers, such as pancreatic adenocarcinoma (PDAC). Given the variable response rates and short progression times of PDAC, there is an unmet clinical need for a method to assess early RT response that may allow better prescription personalization. We hypothesize that quantitative image feature analysis (radiomics) of the longitudinal MR scans acquired before and during MRgRT may be used to extract information related to early treatment response.MethodsHistogram and texture radiomic features (n = 73) were extracted from the Gross Tumor Volume (GTV) in 0.35T MRgRT scans of 26 locally advanced and borderline resectable PDAC patients treated with 50 Gy RT in 5 fractions. Feature ratios between first (F1) and last (F5) fraction scan were correlated with progression free survival (PFS). Feature stability was assessed through region of interest (ROI) perturbation.ResultsLinear normalization of image intensity to median kidney value showed improved reproducibility of feature quantification. Histogram skewness change during treatment showed significant association with PFS (p = 0.005, HR = 2.75), offering a potential predictive biomarker of RT response. Stability analyses revealed a wide distribution of feature sensitivities to ROI delineation and was able to identify features that were robust to variability in contouring.ConclusionsThis study presents a proof-of-concept for the use of quantitative image analysis in MRgRT for treatment response prediction and providing an analysis pipeline that can be utilized in future MRgRT radiomic studies.