Radiomics Analysis Research Articles

Computed tomography radiomics study of invasion and instability of lung adenocarcinoma manifesting as ground glass nodule.

Ground-glass nodule (GGN) is the most common manifestation of lung adenocarcinoma on computed tomography (CT). Clinically, the success rate of preoperative diagnosis of GGN by puncture biopsy and other means is still low. The aim of this study is to investigate the clinical and radiomics characteristics of lung adenocarcinoma presenting as GGN on CT images using radiomics analysis methods, establish a radiomics model, and predict the classification of pathological tissue and instability of GGN type lung adenocarcinoma. This study retrospectively collected 249 patients with 298 GGN lesions who were pathologically confirmed of having lung adenocarcinoma. The images were imported into the Siemens scientific research prototype software to outline the region of interest and extract the radiomics features. Logistic model A (a radiomics model to identify the infiltration of lung adenocarcinoma manifesting as GGNs) was established using features after the dimensionality reduction process. The receiver operating characteristic (ROC) curve of the model on training set and the verification set was drawn, and the area under the curve (AUC) was calculated. Second, a total of 112 lesions were selected from 298 lesions originating from CT images of at least two occasions, and the time between the first CT and the preoperative CT was defined as not less than 90 days. The mass doubling time (MDT) of all lesions was calculated. According to the different MDT diagnostic thresholds instability was predicted. Finally, their AUCs were calculated and compared. There were statistically significant differences in age and lesion location distribution between the "noninvasive" lesion group and the invasive lesion group (P<0.05), but there were no statistically significant differences in sex (P>0.05). Model A had an AUC of 0.89, sensitivity of 0.75, and specificity of 0.86 in the training set and an AUC of 0.87, sensitivity of 0.63, and specificity of 0.90 in the validation set. There was no significant difference statistically in MDT between "noninvasive" lesions and invasive lesions (P>0.05). The AUCs of radiomics models B1, B2 and B3 were 0.89, 0.80, and 0.81, respectively; the sensitivities were 0.71, 0.54, and 0.76, respectively; the specificities were 0.83, 0.77, and 0.60, respectively; and the accuracies were 0.78, 0.65, and 0.69, respectively. There were statistically significant differences in age and location of lesions between the "noninvasive" lesion group and the invasive lesion group. The radiomics model can predict the invasiveness of lung adenocarcinoma manifesting as GGNs. There was no significant difference in MDT between "noninvasive" lesions and invasive lesions. The radiomics model can predict the instability of lung adenocarcinoma manifesting as GGN. When the threshold of MDT was set at 813 days, the model had higher specificity, accuracy, and diagnostic efficiency.

Predicting papillary thyroid microcarcinoma in American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) 3 nodules: radiomics analysis based on intratumoral and peritumoral ultrasound images.

A subset of patients undergoing thyroid surgery for presumed benign thyroid disease presented with papillary thyroid microcarcinoma (PTMC). A non-invasive and precise method for early recognition of PTMC are urgently needed. The aim of this study was to construct and validate a nomogram that combines intratumoral and peritumoral radiomics features as well as clinical features for predicting PTMC in the American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) 3 nodules using ultrasonography. A retrospective review was conducted on a cohort of 221 patients who presented with ACR TI-RADS 3 nodules. These patients were subsequently pathologically diagnosed with either PTMC or benign thyroid nodules. These patients were randomly divided into a training and test cohort with an 8:2 ratio for developing the clinical model, intratumor-region model, peritumor-region model and the combined-region model respectively. The radiomics features were extracted from ultrasound (US) images of each patient. We employed K-nearest neighbor (KNN) model as the base model for building the radiomics signature and clinical signature. Finally, a radiomics-clinical nomogram that combined intratumoral and peritumoral radiomics features as well as clinical features was developed. The prediction performance of each model was assessed by the area under the curve (AUC), sensitivity, specificity and calibration curve. A total of 23 radiomics features were selected to develop radiomics models. The combined-region radiomics model showed favorable prediction efficiency in both the training dataset (AUC: 0.955) and the test dataset (AUC: 0.923). A radiomics-clinical nomogram was constructed and achieved excellent calibration and discrimination, which yielded an AUC value of 0.950, a sensitivity of 0.950 and a specificity of 0.920. This study proposed the nomogram that contributes to the accurate and intuitive identification of PTMC in ACR TI-RADS 3 nodules.

Identification of Calculous Pyonephrosis by CT-Based Radiomics and Deep Learning.

Urgent detection of calculous pyonephrosis is crucial for surgical planning and preventing severe outcomes. This study aims to evaluate the performance of computed tomography (CT)-based radiomics and a three-dimensional convolutional neural network (3D-CNN) model, integrated with independent clinical factors, to identify patients with calculous pyonephrosis. We recruited 182 patients receiving either percutaneous nephrostomy tube placement or percutaneous nephrolithotomy for calculous hydronephrosis or pyonephrosis. The regions of interest were manually delineated on plain CT images and the CT attenuation value (HU) was measured. Radiomics analysis was performed using least absolute shrinkage and selection operator (LASSO). A 3D-CNN model was also developed. The better-performing machine-learning model was combined with independent clinical factors to build a comprehensive clinical machine-learning model. The performance of these models was assessed using receiver operating characteristic analysis and decision curve analysis. Fever, blood neutrophils, and urine leukocytes were independent risk factors for pyonephrosis. The radiomics model showed higher area under the curve (AUC) than the 3D-CNN model and HU (0.876 vs. 0.599, 0.578; p = 0.003, 0.002) in the testing cohort. The clinical machine-learning model surpassed the clinical model in both the training (0.975 vs. 0.904, p = 0.019) and testing (0.967 vs. 0.889, p = 0.045) cohorts.

Predictive value of m6A regulators in prognosis and immunotherapy response of clear cell renal cell carcinoma: a bioinformatics and radiomics analysis

Predictive value of m6A regulators in prognosis and immunotherapy response of clear cell renal cell carcinoma: a bioinformatics and radiomics analysis

Radiomics models to predict bone marrow metastasis of neuroblastoma using CT.

Bone marrow is the leading site for metastasis from neuroblastoma and affects the prognosis of patients with neuroblastoma. However, the accurate diagnosis of bone marrow metastasis is limited by the high spatial and temporal heterogeneity of neuroblastoma. Radiomics analysis has been applied in various cancers to build accurate diagnostic models but has not yet been applied to bone marrow metastasis of neuroblastoma. We retrospectively collected information from 187 patients pathologically diagnosed with neuroblastoma and divided them into training and validation sets in a ratio of 7:3. A total of 2632 radiomics features were retrieved from venous and arterial phases of contrast-enhanced computed tomography (CT), and nine machine learning approaches were used to build radiomics models, including multilayer perceptron (MLP), extreme gradient boosting, and random forest. We also constructed radiomics-clinical models that combined radiomics features with clinical predictors such as age, gender, ascites, and lymph gland metastasis. The performance of the models was evaluated with receiver operating characteristics (ROC) curves, calibration curves, and risk decile plots. The MLP radiomics model yielded an area under the ROC curve (AUC) of 0.97 (95% confidence interval [CI]: 0.95-0.99) on the training set and 0.90 (95% CI: 0.82-0.95) on the validation set. The radiomics-clinical model using an MLP yielded an AUC of 0.93 (95% CI: 0.89-0.96) on the training set and 0.91 (95% CI: 0.85-0.97) on the validation set. MLP-based radiomics and radiomics-clinical models can precisely predict bone marrow metastasis in patients with neuroblastoma.

Using T2-weighted magnetic resonance imaging-derived radiomics to classify cervical lymphadenopathy in children

BackgroundCervical lymphadenopathy is common in children and has diverse causes varying from benign to malignant, their similar manifestations making differential diagnosis difficult.ObjectiveThis study aimed to investigate whether radiomic models using conventional magnetic resonance imaging (MRI) could classify pediatric cervical lymphadenopathy.MethodsA total of 419 cervical lymph nodes from 146 patients, and encompassing four common etiologies (Kikuchi disease, reactive hyperplasia, suppurative lymphadenitis and malignancy), were randomly divided into training and testing sets in a ratio of 7:3. For each lymph node, 1,218 features were extracted from T2-weighted images. Then, the least absolute shrinkage and selection operator (LASSO) models were used to select the most relevant ones. Two models were built using a support vector machine classifier, one was to classify benign and malignant lymph nodes and the other further distinguished four different diseases. The performance was assessed by receiver operating characteristic curves and decision curve analysis.ResultsBy LASSO, 20 features were selected to construct a model to distinguish benign and malignant lymph nodes, which achieved an area under the curve (AUC) of 0.89 and 0.80 in the training and testing sets, respectively. Sixteen features were selected to construct a model to distinguish four different cervical lymphadenopathies. For each etiology, Kikuchi disease, reactive hyperplasia, suppurative lymphadenitis, and malignancy, an AUC of 0.97, 0.91, 0.88, and 0.87 was achieved in the training set, and an AUC of 0.96, 0.80, 0.82, and 0.82 was achieved in the testing set, respectively.ConclusionMRI-derived radiomic analysis provides a promising non-invasive approach for distinguishing causes of cervical lymphadenopathy in children.Graphical

Radiomic analysis of computed tomography myocardial texture: a new tool to identify sublclinical cardiac amyloidosis

Radiomic analysis of computed tomography myocardial texture: a new tool to identify sublclinical cardiac amyloidosis

Prediction of Histological Grade of Oral Squamous Cell Carcinoma Using Machine Learning Models Applied to 18F-FDG-PET Radiomics.

The histological grade of oral squamous cell carcinoma affects the prognosis. In the present study, we performed a radiomics analysis to extract features from 18F-FDG PET image data, created machine learning models from the features, and verified the accuracy of the prediction of the histological grade of oral squamous cell carcinoma. The subjects were 191 patients in whom an 18F-FDG-PET examination was performed preoperatively and a histopathological grade was confirmed after surgery, and their tumor sizes were sufficient for a radiomics analysis. These patients were split in a 70%/30% ratio for use as training data and testing data, respectively. We extracted 2993 radiomics features from the PET images of each patient. Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Naïve Bayes (NB), and K-Nearest Neighbor (KNN) machine learning models were created. The areas under the curve obtained from receiver operating characteristic curves for the prediction of the histological grade of oral squamous cell carcinoma were 0.72, 0.71, 0.84, 0.74, and 0.73 for LR, SVM, RF, NB, and KNN, respectively. We confirmed that a PET radiomics analysis is useful for the preoperative prediction of the histological grade of oral squamous cell carcinoma.

Radiomics analysis based on CT for predicting lymph node metastasis and prognosis in duodenal papillary carcinoma

ObjectivesRadiomics has been demonstrated to be strongly associated with TNM stage and patient prognosis. We aimed to develop a model for predicting lymph node metastasis (LNM) and survival.MethodsFor radiomics texture selection, 3D Slicer 5.0.3 software and the least absolute shrinkage and selection operator (LASSO) algorithm were used. Subsequently, the radiomics model, computed tomography (CT) image, and clinical risk model were compared. The performance of the three models was evaluated using receiver operating characteristic (ROC) curves, decision curve analysis (DCA), calibration plots, and clinical impact curves (CICs).ResultsFor the LNM prediction model, 224 patients with LNM information were used to construct a model that was applied to predict LNM. According to the CT data and clinical characteristics, we constructed a radiomics model, CT imaging model and clinical model. The radiomics model for evaluating LNM status showed excellent calibration and discrimination in the training cohort (AUC = 0.926, 95% CI = 0.869–0.982) and the validation cohort (AUC = 0.872, 95% CI = 0.802–0.941). DeLong’s test demonstrated that the difference among the three models was significant. Similarly, DCA and CIC showed that the radiomics model has better clinical utility than the CT imaging model and clinical model. Our model also exhibited good performance in predicting survival—in line with the findings of the model built with clinical risk factors.ConclusionsCT radiomics models exhibited better predictive performance for LNM than models built based on clinical risk characteristics and CT imaging and had comparative clinical utility for predicting patient prognosis.Critical relevance statementThe radiomics model showed excellent performance and discrimination for predicting LNM and survival of duodenal papillary carcinoma (DPC).Key PointsLNM status determines the most appropriate treatment for DPC.Our radiomics model for evaluating the LNM status of DPC performed excellently.The radiomics model had high sensitivity and specificity for predicting survival, exhibiting great clinical value.Graphical

Coronary CTA-based radiomic signature of pericoronary adipose tissue predict rapid plaque progression

ObjectivesTo explore the value of radiomic features derived from pericoronary adipose tissue (PCAT) obtained by coronary computed tomography angiography for prediction of coronary rapid plaque progression (RPP).MethodsA total of 1233 patients from two centers were included in this multicenter retrospective study. The participants were divided into training, internal validation, and external validation cohorts. Conventional plaque characteristics and radiomic features of PCAT were extracted and analyzed. Random Forest was used to construct five models. Model 1: clinical model. Model 2: plaque characteristics model. Model 3: PCAT radiomics model. Model 4: clinical + radiomics model. Model 5: plaque characteristics + radiomics model. The evaluation of the models encompassed identification accuracy, calibration precision, and clinical applicability. Delong’ test was employed to compare the area under the curve (AUC) of different models.ResultsSeven radiomic features, including two shape features, three first-order features, and two textural features, were selected to build the PCAT radiomics model. In contrast to the clinical model and plaque characteristics model, the PCAT radiomics model (AUC 0.85 for training, 0.84 for internal validation, and 0.81 for external validation; p < 0.05) achieved significantly higher diagnostic performance in predicting RPP. The separate combination of radiomics with clinical and plaque characteristics model did not further improve diagnostic efficacy statistically (p > 0.05).ConclusionRadiomic feature analysis derived from PCAT significantly improves the prediction of RPP as compared to clinical and plaque characteristics. Radiomic analysis of PCAT may improve monitoring RPP over time.Critical relevance statementOur findings demonstrate PCAT radiomics model exhibited good performance in the prediction of RPP, with potential clinical value.Key PointsRapid plaque progression may be predictable with radiomics from pericoronary adipose tissue.Fibrous plaque volume, diameter stenosis, and fat attenuation index were identified as risk factors for predicting rapid plaque progression.Radiomics features of pericoronary adipose tissue can improve the predictive ability of rapid plaque progression.Graphical

Radiomic analysis based on magnetic resonance imaging for the prediction of VEGF expression in hepatocellular carcinoma patients

ObjectiveThe purpose of this study was to investigate the ability of radiomic characteristics of magnetic resonance images to predict vascular endothelial growth factor (VEGF) expression in hepatocellular carcinoma (HCC) patients.MethodsOne hundred and twenty-four patients with HCC who underwent fat-suppressed T2-weighted imaging (FS-T2WI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) one week before surgical resection were enrolled in this retrospective study. Immunohistochemical analysis was used to evaluate the expression level of VEGF. Radiomic features were extracted from the axial FS-T2WI, DCE-MRI (arterial phase and portal venous phase) images of axial MRI. Least absolute shrinkage and selection operator (LASSO) and stepwise regression analyses were performed to select the best radiomic features. Multivariate logistic regression models were constructed and validated using tenfold cross-validation. Receiver operating characteristic (ROC) curve analysis, calibration curve analysis and decision curve analysis (DCA) were employed to evaluate these models.ResultsOur results show that there were 94 patients with high VEGF expression and 30 patients with low VEGF expression among the 124 HCC patients. The FS-T2WI, DCE-MRI and combined MRI radiomics models had AUCs of 0.8713, 0.7819, and 0.9191, respectively. There was no significant difference in the AUC between the FS-T2WI radiomics model and the DCE-MRI radiomics model (p > 0.05), but the AUC for the combined model was significantly greater than the AUCs for the other two models (p < 0.05) according to the DeLong test. The combined model had the greatest net benefit according to the DCA results.ConclusionThe radiomic model based on multisequence MR images has the potential to predict VEGF expression in HCC patients. The combined model showed the best performance.Graphical

NFS-23. ASSOCIATION OF LOCALIZED MAGNETIC RESONANCE IMAGING FEATURES IN ANTERIOR VISUAL PATHWAY WITH VISUAL ACUITY LOSS AMONG CHILDREN WITH NF1-OPG

Abstract BACKGROUND Half of children with optic pathway gliomas associated with neurofibromatosis type 1 (NF1-OPG) are at risk for vision acuity (VA) loss. NF1-OPGs manifest along the anterior visual pathway (AVP), where the overall shape and volume from MRI have been recently associated with VA loss. To investigate the added value of localized features within the optic nerves (ONs) and chiasm, we propose an automatic deep-learning framework for VA loss prediction from volumetric and radiomic analysis of ONs, chiasm, and AVP. METHODS We collected MRIs of 60 children with NF1-OPG from Children’s National Hospital (CNH, GE platform) and 75 from Children’s Hospital of Philadelphia (CHOP, Siemens platform). MRI sequences included T1-, T2-weighted and T2-FLAIR. Experts annotated the AVP to establish ground truth. Neuro-ophthalmic evaluation identified VA loss (LogMAR &amp;gt;= 0.2) in 52/135 children. Automatic AVP segmentation was performed using the Swin transformer network. ONs and chiasm were split through template-based registration. Brain volume, tumor location, child age, and 1,172 radiomic features were used to predict VA loss using support vector machines. Sequential feature selection associated with risk VA loss was done using sensitivity and univariate statistical tests (ANOVA). We compared predictions based on new ON and chiasm analysis with results based on AVP alone. RESULTS Balanced accuracy, sensitivity, specificity, and AUROC of VA loss prediction were 0.865, 0.882, 0.857, 0.899, respectively. Significant risk factors were maximal image intensity and gray level run length entropy in T2-FLAIR for ONs, and image intensity range in T2-FLAIR for chiasm. AUROC based on analysis of AVP alone dropped to 0.728 (p-value=0.028). CONCLUSIONS Deep learning-based analysis indicates an association of new localized MRI features in ONs and chiasm with VA loss, which show enhanced capability in predicting VA loss. This automated framework has the potential to accelerate and guide the treatment decisions for children with NF1-OPGs.

Correlation of CT-based radiomics analysis with pathological cellular infiltration in fibrosing interstitial lung diseases

PurposeWe aimed to identify computed tomography (CT) radiomics features that are associated with cellular infiltration and construct CT radiomics models predictive of cellular infiltration in patients with fibrotic ILD.Materials and methodsCT images of patients with ILD who underwent surgical lung biopsy (SLB) were analyzed. Radiomics features were extracted using artificial intelligence-based software and PyRadiomics. We constructed a model predicting cell counts in histological specimens, and another model predicting two classifications of higher or lower cellularity. We tested these models using external validation.ResultsOverall, 100 patients (mean age: 62 ± 8.9 [standard deviation] years; 61 men) were included. The CT radiomics model used to predict cell count in 140 histological specimens predicted the actual cell count in 59 external validation specimens (root-mean-square error: 0.797). The two-classification model’s accuracy was 70% and the F1 score was 0.73 in the external validation dataset including 30 patients.ConclusionThe CT radiomics-based model developed in this study provided useful information regarding the cellular infiltration in the ILD with good correlation with SLB specimens.

IMG-20. IDENTIFICATION OF PRE-OPERATIVE IMAGING FEATURES ASSOCIATED WITH COGNITIVE IMPAIRMENT IN CHILDREN WITH BRAIN TUMOURS USING A LANGUAGE MODELLING SYSTEM COMBINED WITH MACHINE LEARNING ANALYSIS

Abstract BACKGROUND Cognitive impairments caused by paediatric brain tumours negatively impact future quality-of-life. Impairments of processing speed and working memory are important contributors to challenges these children encounter in education. Inconsistent image acquisition and poor imaging quality can restrict advanced radiomic analysis in historical datasets. We developed and refined a language modelling human image-analysis support system, passing data to specific machine learning techniques that can utilise datasets with many data points in a limited population. METHODS Medical records of children treated for brain tumours with surgery and radiotherapy in Manchester between 2000-2020 were reviewed with imaging accessible from 2006-2020. Pre-operative MRI scans were reviewed for 52 patients along with their most recent processing speed and/or working memory test scores. 274 radiological variables analysing tumour features and damage to brain structures were reported per patient using a standardised reporting tool. Univariate analysis was performed using Linear Regression with time-to-test analysed as a covariate. Multivariate analysis used Elastic Net Regression. RESULTS Univariate analysis revealed risk factors for slower processing speed included midline shift, tumour in the corpus callosum, right temporal periventricular white matter oedema, diffusion changes in the left frontal lobe or amygdala and FLAIR changes in the left caudate nucleus (p&amp;lt;0.005). Risk factors for impaired working memory included midline shift, cortical tumour location, left periventricular white matter oedema and diffusion or FLAIR changes in the left frontal lobe (p&amp;lt;0.005). Despite the limited population size and review of only pre-operative features, multivariate analysis produced multiple predictive models able to outperform chance when predicting processing speed and working memory outcomes. CONCLUSIONS This study provides evidence that a radiological language modelling approach can be utilised for outcome prediction. It identified imaging features predictive of cognitive impairments prior to surgery and radiotherapy. Further work is needed to assess inter-observer variability and validate these findings.

IMG-21. HOW CAN WE BEST UTILISE HISTORICAL IMAGING DATASETS FOR RESEARCH IN PAEDIATRIC NEURO-ONCOLOGY? A REVIEW OF HISTORICAL PRE-OPERATIVE IMAGING PERFORMED FOR 92 PAEDIATRIC BRAIN TUMOUR PATIENTS WITH COMPARISON TO 2021 SIOPE GUIDELINES

Abstract BACKGROUND Utilisation of historical imaging is important for studying long-term outcomes in paediatric brain tumours often pooling data across multiple centres retrospectively over many years. Disparities in MRI sequence parameters, imaging planes, fields of view and resolution limit quantitative radiomic analysis and segmentation algorithms. We review historical imaging datasets, analysing the MRI scans performed assessing suitability for modern research techniques. METHODS Retrospective review of presenting MRI scans was performed in 92 paediatric brain tumour patients between 2006-2020. Whether imaging was performed internally at our paediatric neurosciences centre or externally at their referring hospitals was recorded. Image sequences, planes and slice thickness were compared with 2021 SIOPE guidelines for paediatric brain tumour imaging. T2 signal intensities for tumour, white and grey matter were compared for scans completed in different centres. Statistical testing was performed with T-Test and Paired T-Test. RESULTS 75% of patients had all SIOPE essential sequences and 60.9% were in the recommended planes. Only 1 patient had the desired slice thickness in all sequences recommended for 1.5T scanners. 38.5% had T1 and 27.8% had post-contrast T1 in the ≤1mm slice thickness recommended for 3T scanners. Comparison between 62 internal and 30 external scans revealed statistically significant differences in T2 signal intensities for regions of interest (ROI) within tumour, white and grey matter (p&amp;lt;0.01). Comparison of matched ROI for 12 patients imaged internally and externally demonstrated statistically significant differences in T2 signal intensities (p&amp;lt;0.05). Differences persisted despite normalisation of T2 signal intensity as ratios between tumour and the weighted mean of white and grey matter (p&amp;lt;0.01). CONCLUSIONS This study highlights the challenges of non-standardisation of imaging resolution and acquisition parameters in historical imaging. One approach to utilise these datasets is to use qualitative classifications where human experts label imaging features and encode data for future statistical or machine learning analysis.

Radiomics analysis of lesion-specific pericoronary adipose tissue to predict major adverse cardiovascular events in coronary artery disease

ObjectiveTo investigate the prognostic performance of radiomics analysis of lesion-specific pericoronary adipose tissue (PCAT) for major adverse cardiovascular events (MACE) with the guidance of CT derived fractional flow reserve (CT-FFR) in coronary artery disease (CAD).Materials and methodsThe study retrospectively analyzed 608 CAD patients who underwent coronary CT angiography. Lesion-specific PCAT was determined by the lowest CT-FFR value and 1691 radiomic features were extracted. MACE included cardiovascular death, nonfatal myocardial infarction, unplanned revascularization and hospitalization for unstable angina. Four models were generated, incorporating traditional risk factors (clinical model), radiomics score (Rad-score, radiomics model), traditional risk factors and Rad-score (clinical radiomics model) and all together (combined model). The model performances were evaluated and compared with Harrell concordance index (C-index), area under curve (AUC) of the receiver operator characteristic.ResultsLesion-specific Rad-score was associated with MACE (adjusted HR = 1.330, p = 0.009). The combined model yielded the highest C-index of 0.718, which was higher than clinical model (C-index = 0.639), radiomics model (C-index = 0.653) and clinical radiomics model (C-index = 0.698) (all p < 0.05). The clinical radiomics model had significant higher C-index than clinical model (p = 0.030). There were no significant differences in C-index between clinical or clinical radiomics model and radiomics model (p values were 0.796 and 0.147 respectively). The AUC increased from 0.674 for clinical model to 0.721 for radiomics model, 0.759 for clinical radiomics model and 0.773 for combined model.ConclusionRadiomics analysis of lesion-specific PCAT is useful in predicting MACE. Combination of lesion-specific Rad-score and CT-FFR shows incremental value over traditional risk factors.

Multivariate testing and effect size measures for batch effect evaluation in radiomic features

While precision medicine applications of radiomics analysis are promising, differences in image acquisition can cause “batch effects” that reduce reproducibility and affect downstream predictive analyses. Harmonization methods such as ComBat have been developed to correct these effects, but evaluation methods for quantifying batch effects are inconsistent. In this study, we propose the use of the multivariate statistical test PERMANOVA and the Robust Effect Size Index (RESI) to better quantify and characterize batch effects in radiomics data. We evaluate these methods in both simulated and real radiomics features extracted from full-field digital mammography (FFDM) data. PERMANOVA demonstrated higher power than standard univariate statistical testing, and RESI was able to interpretably quantify the effect size of site at extremely large sample sizes. These methods show promise as more powerful and interpretable methods for the detection and quantification of batch effects in radiomics studies.

Endoscopic ultrasonography-based intratumoral and peritumoral machine learning radiomics analyses for distinguishing insulinomas from non-functional pancreatic neuroendocrine tumors.

To develop and validate radiomics models utilizing endoscopic ultrasonography (EUS) images to distinguish insulinomas from non-functional pancreatic neuroendocrine tumors (NF-PNETs). A total of 106 patients, comprising 61 with insulinomas and 45 with NF-PNETs, were included in this study. The patients were randomly assigned to either the training or test cohort. Radiomics features were extracted from both the intratumoral and peritumoral regions, respectively. Six machine learning algorithms were utilized to train intratumoral prediction models, using only the nonzero coefficient features. The researchers identified the most effective intratumoral radiomics model and subsequently employed it to develop peritumoral and combined radiomics models. Finally, a predictive nomogram for insulinomas was constructed and assessed. A total of 107 radiomics features were extracted based on EUS, and only features with nonzero coefficients were retained. Among the six intratumoral radiomics models, the light gradient boosting machine (LightGBM) model demonstrated superior performance. Furthermore, a peritumoral radiomics model was established and evaluated. The combined model, integrating both the intratumoral and peritumoral radiomics features, exhibited a comparable performance in the training cohort (AUC=0.876) and achieved the highest accuracy in predicting outcomes in the test cohorts (AUC=0.835). The Delong test, calibration curves, and decision curve analysis (DCA) were employed to validate these findings. Insulinomas exhibited a significantly smaller diameter compared to NF-PNETs. Finally, the nomogram, incorporating diameter and radiomics signature, was constructed and assessed, which owned superior performance in both the training (AUC=0.929) and test (AUC=0.913) cohorts. A novel and impactful radiomics model and nomogram were developed and validated for the accurate differentiation of NF-PNETs and insulinomas utilizing EUS images.

Radiomic Parameters for the Evaluation of Response to Treatment in Metastatic Colorectal Cancer Patients with Liver Metastasis: Findings from the CAVE-GOIM mCRC Phase 2 Trial.

CAVE is a single arm, Phase 2 trial, that demonstrated anti-tumor activity of cetuximab rechallenge plus avelumab in patients with RAS wild type (wt) metastatic colorectal cancer (mCRC). We conducted a post hoc analysis to identify potential radiomic biomarkers for patients with CRC liver metastasis (LM). Patients with LM that could be measured by enhanced contrast phase computed tomography (CT) imaging at baseline and at first response evaluation were included. Multiple texture parameters were extracted with the LifeX Software. Delta-texture (D-TA) variations were calculated by comparing data at baseline and after treatment. Overall, 55/77 patients (71%) had LM; 39 met the inclusion criteria for the current analysis. The D-TA parameters that significantly correlated at univariate analysis with median progression-free survival (mPFS) were EntropyHistogram (p = 0.021), HomogeneityGLCM (p < 0.001) and Dissimilarity GLCM (p = 0.002). At multivariate analysis, only HomogeneityGLCM resulted significant for PFS (p = 0.001). Patients (19/39, 48.7%) with reduction of HomogeneityGLCM experienced better mPFS (4.6 vs 2.9 months; HR 0.45; 95% CI 0.23-0.88; p = 0.021) and median overall survival (mOS) (17.3 vs 6.8 months; HR 0.40, 95% CI 0.21-0.80; p = 0.010). A trend to better mPFS, was also observed in patients with RAS/BRAF wt circulating tumor DNA and reduction of HomogeneityGLCM. Overall survival was significantly better in this subgroup of patients with low HomogeneityGLCM: mOS was 17.8 (95% CI 15.5-20.2) versus 6.8 months (95% CI 3.6-10.0) (HR 0.34, 95% CI 0.14-0.81; p = 0.016). Reduction in the D-TA parameter HomogeneityGLCM by radiomic analysis correlates with improved outcomes in patients with LM receiving cetuximab rechallenge plus avelumab therapy. Larger prospective studies are needed to validate and confirm these findings.

Relating Macroscopic PET Radiomics Features to Microscopic Tumor Phenotypes Using a Stochastic Mathematical Model of Cellular Metabolism and Proliferation.

Cancers can manifest large variations in tumor phenotypes due to genetic and microenvironmental factors, which has motivated the development of quantitative radiomics-based image analysis with the aim to robustly classify tumor phenotypes in vivo. Positron emission tomography (PET) imaging can be particularly helpful in elucidating the metabolic profiles of tumors. However, the relatively low resolution, high noise, and limited PET data availability make it difficult to study the relationship between the microenvironment properties and metabolic tumor phenotype as seen on the images. Most of previously proposed digital PET phantoms of tumors are static, have an over-simplified morphology, and lack the link to cellular biology that ultimately governs the tumor evolution. In this work, we propose a novel method to investigate the relationship between microscopic tumor parameters and PET image characteristics based on the computational simulation of tumor growth. We use a hybrid, multiscale, stochastic mathematical model of cellular metabolism and proliferation to generate simulated cross-sections of tumors in vascularized normal tissue on a microscopic level. The generated longitudinal tumor growth sequences are converted to PET images with realistic resolution and noise. By changing the biological parameters of the model, such as the blood vessel density and conditions for necrosis, distinct tumor phenotypes can be obtained. The simulated cellular maps were compared to real histology slides of SiHa and WiDr xenografts imaged with Hoechst 33342 and pimonidazole. As an example application of the proposed method, we simulated six tumor phenotypes that contain various amounts of hypoxic and necrotic regions induced by a lack of oxygen and glucose, including phenotypes that are distinct on the microscopic level but visually similar in PET images. We computed 22 standardized Haralick texture features for each phenotype, and identified the features that could best discriminate the phenotypes with varying image noise levels. We demonstrated that "cluster shade" and "difference entropy" are the most effective and noise-resilient features for microscopic phenotype discrimination. Longitudinal analysis of the simulated tumor growth showed that radiomics analysis can be beneficial even in small lesions with a diameter of 3.5-4 resolution units, corresponding to 8.7-10.0 mm in modern PET scanners. Certain radiomics features were shown to change non-monotonically with tumor growth, which has implications for feature selection for tracking disease progression and therapy response.