Radiomics-based Machine Learning Model Research Articles

Deep Neural Networks and Machine Learning Radiomics Modelling for Prediction of Relapse in Mantle Cell Lymphoma.

Simple SummaryMantle cell lymphoma (MCL) is an aggressive lymphoid tumour with a poor prognosis. There exist no routine biomarkers for the early prediction of relapse. Our study compared the potential of radiomics-based machine learning and 3D deep learning models as non-invasive biomarkers to risk-stratify MCL patients, thus promoting precision imaging in clinical oncology. Mantle cell lymphoma (MCL) is a rare lymphoid malignancy with a poor prognosis characterised by frequent relapse and short durations of treatment response. Most patients present with aggressive disease, but there exist indolent subtypes without the need for immediate intervention. The very heterogeneous behaviour of MCL is genetically characterised by the translocation t(11;14)(q13;q32), leading to Cyclin D1 overexpression with distinct clinical and biological characteristics and outcomes. There is still an unfulfilled need for precise MCL prognostication in real-time. Machine learning and deep learning neural networks are rapidly advancing technologies with promising results in numerous fields of application. This study develops and compares the performance of deep learning (DL) algorithms and radiomics-based machine learning (ML) models to predict MCL relapse on baseline CT scans. Five classification algorithms were used, including three deep learning models (3D SEResNet50, 3D DenseNet, and an optimised 3D CNN) and two machine learning models based on K-nearest Neighbor (KNN) and Random Forest (RF). The best performing method, our optimised 3D CNN, predicted MCL relapse with a 70% accuracy, better than the 3D SEResNet50 (62%) and the 3D DenseNet (59%). The second-best performing method was the KNN-based machine learning model (64%) after principal component analysis for improved accuracy. Our optimised CNN developed by ourselves correctly predicted MCL relapse in 70% of the patients on baseline CT imaging. Once prospectively tested in clinical trials with a larger sample size, our proposed 3D deep learning model could facilitate clinical management by precision imaging in MCL.

Radiomics versus Conventional Assessment to Identify Symptomatic Participants at Carotid Computed Tomography Angiography

Introduction: Carotid computed tomography angiography (CTA) is routinely used for evaluating the atherosclerotic process. Radiomics allows the extraction of imaging markers of lesion heterogeneity and spatial complexity. These quantitative features can be used as the input for machine learning (ML). Therefore, in this study, we aimed to evaluate the diagnostic performance of radiomics-based ML assessment of carotid CTA data to identify symptomatic patients with carotid artery atherosclerosis. Methods: In this retrospective study, participants with carotid artery atherosclerosis who underwent carotid CTA and brain magnetic resonance imaging from May 2010 to December 2017 were studied. The participants were grouped into symptomatic and asymptomatic groups according to their recent symptoms (determination of ipsilateral ischemic stroke). Eight conventional plaque features and 2,107 radiomics parameters were extracted from carotid CTA images. A radiomics-based ML model was fitted on the training set, and the radiomics-based ML model and conventional assessment were compared using the area under the curve (AUC) to identify symptomatic participants. Results: After excluding participants with other stroke sources, 120 patients with 148 carotid arteries were analyzed. Of these 148 carotid arteries, 34 (22.97%) were classified into the symptomatic group. Plaque ulceration (odds ratio [OR] = 0.257; 95% confidence interval [CI], 0.094–0.698) and plaque enhancement (OR = 0.305; 95% CI, 0.094–0.988) were associated with the symptomatic status. Twenty radiomics parameters were chosen to be inputs in the radiomics-based ML model. In the identification of symptomatic participants, the discriminatory value of the radiomics-based ML model was significantly higher than that of the conventional assessment (AUC = 0.858 vs. AUC = 0.706, p = 0.021). Conclusion: Radiomics-based ML analysis improves the discriminatory power of carotid CTA in the identification of recent ischemic symptoms in patients with carotid artery atherosclerosis.

Identification of pathology-confirmed vulnerable atherosclerotic lesions by coronary computed tomography angiography using radiomics analysis.

To explore whether radiomics-based machine learning (ML) models could outperform conventional diagnostic methods at identifying vulnerable lesions on coronary computed tomographic angiography (CCTA). In this retrospective study, 36 heart transplant recipients with coronary heart disease (CAD) and end-stage heart failure were included. Pathological cross-section samples of 350 plaques were collected and coregistered to patients' preoperative CCTA images. A total of 1184 radiomic features were extracted from CCTA images. Through feature selection and stratified fivefold cross-validation, we derived eight radiomics-based ML models for lesion vulnerability prediction. An independent set of 196 plaques from another 8 CAD patients who underwent heart transplants was collected to validate radiomics-based ML models' diagnostic accuracy against conventional CCTA feature-based diagnosis (presence of at least 2 high-risk plaque features). The performance of the prediction models was assessed by the area under the receiver operating characteristic curve (AUC) with 95% confidence intervals (CI). The training group used to develop radiomics-based ML models contained 200/350 (57.1%) vulnerable plaques and the external validation group was composed of 67.3% (132/196) vulnerable plaques. The radiomics-based ML model based on eight radiomic features showed excellent cross-validation diagnostic accuracy (AUC: 0.900 ± 0.033). In the validation group, diagnosis based on conventional CCTA features demonstrated moderate performance (AUC: 0.656 [95% CI: 0.593 -0.718]), while the radiomics-based ML model showed higher diagnostic ability (0.782 [95% CI: 0.710 -0.846]). Radiomics-based ML models showed better diagnostic ability than the conventional CCTA features at assessing coronary plaque vulnerability. • CCTA has great potential in the diagnosis of vulnerable coronary artery lesions. • Radiomics model built through CCTA could discriminate coronary vulnerable lesions in good diagnostic ability. • Radiomics model could improve the ability of vulnerability diagnosis against traditional CCTA method, sensitivity especially.

Abstract 39: Ability Of Radiomics Versus Humans In Predicting First-pass Effect After Endovascular Treatment In The Escape-na1 Trial

Introduction: First-pass effect (FPE), i.e., achieving reperfusion with a single thrombectomy device pass, is associated with better clinical outcomes in patients with acute stroke. FPE is therefore increasingly being used as a marker of device and procedural efficacy. We evaluated the ability of thrombus-based radiomics models to predict FPE in patients undergoing endovascular thrombectomy (EVT) and compare performance to experts and non-radiomics thrombus characteristics. Methods: Patients with thin-slice non-contrast CT and CT angiography from The Efficacy and Safety of Nerinetide for the Treatment of Acute Ischemic Stroke (ESCAPE-NA1) trial were included. Thrombi were manually segmented on all images. Data was randomly split into a derivation set that included a training and a validation subset and an independent test set. Radiomics features were extracted from the derivation data set. Three expert stroke physicians reviewed baseline imaging and clinical data for the test set. The machine learning (ML) models were compared to the three experts in predicting the primary outcome (FPE) in the test set using area under the receiver operating characteristic curves (AUC-ROC). Results: A total of 554 patients with available thin-slice images comprised of a derivation set (training subset [n=388, 70%]), validation subset [n=55, 10%]), and a test set (n=111, 20%). FPE was seen in 31.8% in the derivation set and 31.5 % in the test set. AUC of the best radiomics model was 0.74 (95% CI: 0.64, 0.84), which was higher than the mean AUC of the three experts 0.60 (95% CI: 0.50, 0.71) ( P =0.009). Specificity of radiomics was better than the mean specificity of the three experts, 46 of 76 (60%) vs. 35 of 76 (46.4%), P =0.004, whereas sensitivity was not significantly different between radiomics (28 of 35 [79%]) and experts (27 of 35 [77%]). Moreover, radiomics features performed better than non-radiomics features such as thrombus volume and permeability measurements in predicting FPE ( P <0.05). Conclusion: A radiomics-based ML model of thrombus characteristics on non-contrast CT and CT angiography performs better than experts and non-radiomics image characteristics in predicting FPE in patients with acute stroke treated with EVT.

Prediction of Clinical Outcome for High-Intensity Focused Ultrasound Ablation of Uterine Leiomyomas Using Multiparametric MRI Radiomics-Based Machine Leaning Model.

ObjectivesThis study sought to develop a multiparametric MRI radiomics-based machine learning model for the preoperative prediction of clinical success for high-intensity-focused ultrasound (HIFU) ablation of uterine leiomyomas.MethodsOne hundred and thirty patients who received HIFU ablation therapy for uterine leiomyomas were enrolled in this retrospective study. Radiomics features were extracted from T2-weighted (T2WI) image and ADC map derived from diffusion-weighted imaging (DWI). Three feature selection algorithms including least absolute shrinkage and selection operator (LASSO), recursive feature elimination (RFE), and ReliefF algorithm were used to select radiomics features, respectively, which were fed into four machine learning classifiers including k-nearest neighbors (KNN), logistic regression (LR), random forest (RF), and support vector machine (SVM) for the construction of outcome prediction models before HIFU treatment. The performance, predication ability, and clinical usefulness of these models were verified and evaluated using receiver operating characteristics (ROC), calibration, and decision curve analyses.ResultsThe radiomics analysis provided an effective preoperative prediction for HIFU ablation of uterine leiomyomas. Using SVM with ReliefF algorithm, the multiparametric MRI radiomics model showed the favorable performance with average accuracy of 0.849, sensitivity of 0.814, specificity of 0.896, positive predictive value (PPV) of 0.903, negative predictive value (NPV) of 0.823, and the area under the ROC curve (AUC) of 0.887 (95% CI = 0.848–0.939) in fivefold cross-validation, followed by RF with ReliefF. Calibration and decision curve analyses confirmed the potential of model in predication ability and clinical usefulness.ConclusionsThe radiomics-based machine learning model can predict preoperatively HIFU ablation response for the patients with uterine leiomyomas and contribute to determining individual treatment strategies.

Radiation Versus Immune Checkpoint Inhibitor Associated Pneumonitis: Distinct Radiologic Morphologies.

BackgroundPatients with non‐small cell lung cancer may develop pneumonitis after thoracic radiotherapy (RT) and immune checkpoint inhibitors (ICIs). We hypothesized that distinct morphologic features are associated with different pneumonitis etiologies.Materials and MethodsWe systematically compared computed tomography (CT) features of RT‐ versus ICI‐pneumonitis. Clinical and imaging features were tested for association with pneumonitis severity. Lastly, we constructed an exploratory radiomics‐based machine learning (ML) model to discern pneumonitis etiology.ResultsBetween 2009 and 2019, 82 patients developed pneumonitis: 29 after thoracic RT, 23 after ICI, and 30 after RT + ICI. Fifty patients had grade 2 pneumonitis, 22 grade 3, and 7 grade 4. ICI‐pneumonitis was more likely bilateral (65% vs. 28%; p = .01) and involved more lobes (66% vs. 45% involving at least three lobes) and was less likely to have sharp border (17% vs. 59%; p = .004) compared with RT‐pneumonitis. Pneumonitis morphology after RT + ICI was heterogeneous, with 47% bilateral, 37% involving at least three lobes, and 40% sharp borders. Among all patients, risk factors for severe pneumonitis included poor performance status, smoking history, worse lung function, and bilateral and multifocal involvement on CT. An ML model based on seven radiomic features alone could distinguish ICI‐ from RT‐pneumonitis with an area under the receiver‐operating curve of 0.76 and identified the predominant etiology after RT + ICI concordant with multidisciplinary consensus.ConclusionRT‐ and ICI‐pneumonitis exhibit distinct spatial features on CT. Bilateral and multifocal lung involvement is associated with severe pneumonitis. Integrating these morphologic features in the clinical management of patients who develop pneumonitis after RT and ICIs may improve treatment decision‐making.Implications for PracticePatients with non‐small cell lung cancer often receive thoracic radiation and immune checkpoint inhibitors (ICIs), both of which can cause pneumonitis. This study identified similarities and differences in pneumonitis morphology on computed tomography (CT) scans among pneumonitis due to radiotherapy (RT) alone, ICI alone, and the combination of both. Patients who have bilateral CT changes involving at least three lobes are more likely to have ICI‐pneumonitis, whereas those with unilateral CT changes with sharp borders are more likely to have radiation pneumonitis. After RT and/or ICI, severe pneumonitis is associated with bilateral and multifocal CT changes. These results can help guide clinicians in triaging patients who develop pneumonitis after radiation and during ICI treatment.

Radiomics side experiments and DAFIT approach in identifying pulmonary hypertension using Cardiac MRI derived radiomics based machine learning models

Side experiments are performed on radiomics models to improve their reproducibility. We measure the impact of myocardial masks, radiomic side experiments and data augmentation for information transfer (DAFIT) approach to differentiate patients with and without pulmonary hypertension (PH) using cardiac MRI (CMRI) derived radiomics. Feature extraction was performed from the left ventricle (LV) and right ventricle (RV) myocardial masks using CMRI in 82 patients (42 PH and 40 controls). Various side study experiments were evaluated: Original data without and with intraclass correlation (ICC) feature-filtering and DAFIT approach (without and with ICC feature-filtering). Multiple machine learning and feature selection strategies were evaluated. Primary analysis included all PH patients with subgroup analysis including PH patients with preserved LVEF (≥ 50%). For both primary and subgroup analysis, DAFIT approach without feature-filtering was the highest performer (AUC 0.957–0.958). ICC approaches showed poor performance compared to DAFIT approach. The performance of combined LV and RV masks was superior to individual masks alone. There was variation in top performing models across all approaches (AUC 0.862–0.958). DAFIT approach with features from combined LV and RV masks provide superior performance with poor performance of feature filtering approaches. Model performance varies based upon the feature selection and model combination.

Differentiation of Low-Grade Astrocytoma From Anaplastic Astrocytoma Using Radiomics-Based Machine Learning Techniques.

PurposeTo investigate the diagnostic ability of radiomics-based machine learning in differentiating atypical low-grade astrocytoma (LGA) from anaplastic astrocytoma (AA).MethodsThe current study involved 175 patients diagnosed with LGA (n = 95) or AA (n = 80) and treated in the Neurosurgery Department of West China Hospital from April 2010 to December 2019. Radiomics features were extracted from pre-treatment contrast-enhanced T1 weighted imaging (T1C). Nine diagnostic models were established with three selection methods [Distance Correlation, least absolute shrinkage, and selection operator (LASSO), and Gradient Boosting Decision Tree (GBDT)] and three classification algorithms [Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and random forest (RF)]. The sensitivity, specificity, accuracy, and areas under receiver operating characteristic curve (AUC) of each model were calculated. Diagnostic ability of each model was evaluated based on these indexes.ResultsNine radiomics-based machine learning models with promising diagnostic performances were established. For LDA-based models, the optimal one was the combination of LASSO + LDA with AUC of 0.825. For SVM-based modes, Distance Correlation + SVM represented the most promising diagnostic performance with AUC of 0.808. And for RF-based models, Distance Correlation + RF were observed to be the optimal model with AUC of 0.821.ConclusionRadiomic-based machine-learning has the potential to be utilized in differentiating atypical LGA from AA with reliable diagnostic performance.

Prediction of unenhanced lesion evolution in multiple sclerosis using radiomics-based models: a machine learning approach

Background: The volume change of multiple sclerosis (MS) lesion is related to its activity and can be used to assess disease progression. Therefore, the purpose of this study was to develop radiomics models for predicting the evolution of unenhanced MS lesions by using different kinds of machine learning algorithms and explore the optimal model.Methods: In this prospective observation, 45 follow-up MR images obtained in 36 patients with MS (mean age 32.53±10.91; 23 women, 13 men) were evaluated. The lesions will be defined as interval activity and interval inactivity, respectively, based on the percentage of enlargement or reduction of the lesion >20% in the follow-up MR images. We extracted radiomic features of lesions on FLAIR images, and used recursive feature elimination (RFE), ReliefF algorithm and least absolute shrinkage and selection operator (LASSO) for feature selection, then three classification models including logistic regression, random forest and support vector machine (SVM) were used to build predictive models. The performance of the models were evaluated based on the sensitivity, specificity, precision, negative predictive value (NPV) and receiver operating characteristic curve (ROC) curves analyses.Results: 135 interval inactivity lesions and 110 interval activity lesions were registered in our study. A total of 972 radiomics features were extracted, of which 265 were robust. The consistency and effectiveness of model performance were compared and verified by different combinations of feature selection and machine learning methods in different K-fold cross-validation strategies where K ranges from 5 to 10, thus demonstrating the stability and robustness. SVM classifier with ReliefF algorithm had the best prediction performance with an average accuracy of 0.827, sensitivity of 0.809, specificity of 0.841, precision of 0.921, NPV of 0.948 and the areas under the ROC curves (AUC) of 0.857 (95% CI: 0.812–0.902) in the cohorts.Conclusion: The results demonstrated that the radiomics-based machine learning model has potential in predicting the evolution of MS lesions.

Development and Validation of a Radiomics Model for Differentiating Bone Islands and Osteoblastic Bone Metastases at Abdominal CT.

Background It is important to diagnose sclerotic bone lesions in order to determine treatment strategy. Purpose To evaluate the diagnostic performance of a CT radiomics-based machine learning model for differentiating bone islands and osteoblastic bone metastases. Materials and Methods In this retrospective study, patients who underwent contrast-enhanced abdominal CT and were diagnosed with a bone island or osteoblastic metastasis between 2015 to 2019 at either of two different institutions were included: institution 1 for the training set and institution 2 for the external test set. Radiomics features were extracted. The random forest (RF) model was built using 10 selected features, and subsequent 10-fold cross-validation was performed. In the test phase, the RF model was tested with an external test set. Three radiologists reviewed the CT images for the test set. The sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC) were calculated for the models and each of the three radiologists. The AUCs of the radiomics model and radiologists were compared. Results A total of 177 patients (89 with a bone island and 88 with metastasis; mean age, 66 years ± 12 [standard deviation]; 111 men) were in the training set, and 64 (23 with a bone island and 41 with metastasis; mean age, 69 years ± 14; 59 men) were in the test set. Radiomics features (n = 1218) were extracted. The average AUC of the RF model from 10-fold cross-validation was 0.89 (sensitivity, 85% [75 of 88 patients]; specificity, 82% [73 of 89 patients]; and accuracy, 84% [148 of 177 patients]). In the test set, the AUC of the trained RF model was 0.96 (sensitivity, 80% [33 of 41 patients]; specificity, 96% [22 of 23 patients]; and accuracy, 86% [55 of 64 patients]). The AUCs for the three readers were 0.95 (95% CI: 0.90, 1.00), 0.96 (95% CI: 0.90, 1.00), and 0.88 (95% CI: 0.80, 0.96). The AUC of radiomics model was higher than that of only reader 3 (0.96 vs 0.88, respectively; P = .03). Conclusion A CT radiomics-based random forest model was proven useful for differentiating bone islands from osteoblastic metastases and showed better diagnostic performance compared with an inexperienced radiologist. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Vannier in this issue.

Radiomics-based machine learning model for efficiently classifying transcriptome subtypes in glioblastoma patients from MRI

BackgroundIn the field of glioma, transcriptome subtypes have been considered as an important diagnostic and prognostic biomarker that may help improve the treatment efficacy. However, existing identification methods of transcriptome subtypes are limited due to the relatively long detection period, the unattainability of tumor specimens via biopsy or surgery, and the fleeting nature of intralesional heterogeneity. In search of a superior model over previous ones, this study evaluated the efficiency of eXtreme Gradient Boosting (XGBoost)-based radiomics model to classify transcriptome subtypes in glioblastoma patients. MethodsThis retrospective study retrieved patients from TCGA-GBM and IvyGAP cohorts with pathologically diagnosed glioblastoma, and separated them into different transcriptome subtypes groups. GBM patients were then segmented into three different regions of MRI: enhancement of the tumor core (ET), non-enhancing portion of the tumor core (NET), and peritumoral edema (ED). We subsequently used handcrafted radiomics features (n = 704) from multimodality MRI and two-level feature selection techniques (Spearman correlation and F-score tests) in order to find the features that could be relevant. ResultsAfter the feature selection approach, we identified 13 radiomics features that were the most meaningful ones that can be used to reach the optimal results. With these features, our XGBoost model reached the predictive accuracies of 70.9%, 73.3%, 88.4%, and 88.4% for classical, mesenchymal, neural, and proneural subtypes, respectively. Our model performance has been improved in comparison with the other models as well as previous works on the same dataset. ConclusionThe use of XGBoost and two-level feature selection analysis (Spearman correlation and F-score) could be expected as a potential combination for classifying transcriptome subtypes with high performance and might raise public attention for further research on radiomics-based GBM models.

A Machine Learning Model Based on PET/CT Radiomics and Clinical Characteristics Predicts ALK Rearrangement Status in Lung Adenocarcinoma.

ObjectivesAnaplastic lymphoma kinase (ALK) rearrangement status examination has been widely used in clinic for non-small cell lung cancer (NSCLC) patients in order to find patients that can be treated with targeted ALK inhibitors. This study intended to non-invasively predict the ALK rearrangement status in lung adenocarcinomas by developing a machine learning model that combines PET/CT radiomic features and clinical characteristics.MethodsFive hundred twenty-six patients of lung adenocarcinoma with PET/CT scan examination were enrolled, including 109 positive and 417 negative patients for ALK rearrangements from February 2016 to March 2019. The Artificial Intelligence Kit software was used to extract radiomic features of PET/CT images. The maximum relevance minimum redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) logistic regression were further employed to select the most distinguishable radiomic features to construct predictive models. The mRMR is a feature selection method, which selects the features with high correlation to the pathological results (maximum correlation), meanwhile retain the features with minimum correlation between them (minimum redundancy). LASSO is a statistical formula whose main purpose is the feature selection and regularization of data model. LASSO method regularizes model parameters by shrinking the regression coefficients, reducing some of them to zero. The feature selection phase occurs after the shrinkage, where every non-zero value is selected to be used in the model. Receiver operating characteristic (ROC) analysis was used to evaluate the performance of the models, and the performance of different models was compared by the DeLong test.ResultsA total of 22 radiomic features were extracted from PET/CT images for constructing the PET/CT radiomic model, and majority of these features used were based on CT features (20 out of 22), only 2 PET features were included (PET percentile 10 and PET difference entropy). Moreover, three clinical features associated with ALK mutation (age, burr and pleural effusion) were also employed to construct a combined model of PET/CT and clinical model. We found that this combined model PET/CT-clinical model has a significant advantage to predict the ALK mutation status in the training group (AUC = 0.87) and the testing group (AUC = 0.88) compared with the clinical model alone in the training group (AUC = 0.76) and the testing group (AUC = 0.74) respectively. However, there is no significant difference between the combined model and PET/CT radiomic model.ConclusionsThis study demonstrated that PET/CT radiomics-based machine learning model has potential to be used as a non-invasive diagnostic method to help diagnose ALK mutation status for lung adenocarcinoma patients in the clinic.

Detecting MLC modeling errors using radiomics-based machine learning in patient-specific QA with an EPID for intensity-modulated radiation therapy.

We sought to develop machine learning models to detect multileaf collimator (MLC) modeling errors with the use of radiomic features of fluence maps measured in patient-specific quality assurance (QA) for intensity-modulated radiation therapy (IMRT) with an electric portal imaging device (EPID). Fluence maps measured with EPID for 38 beams from 19 clinical IMRT plans were assessed. Plans with various degrees of error in MLC modeling parameters [i.e., MLC transmission factor (TF) and dosimetric leaf gap (DLG)] and plans with an MLC positional error for comparison were created. For a total of 152 error plans for each type of error, we calculated fluence difference maps for each beam by subtracting the calculated maps from the measured maps. A total of 837 radiomic features were extracted from each fluence difference map, and we determined the number of features used for the training dataset in the machine learning models by using random forest regression. Machine learning models using the five typical algorithms [decision tree, k-nearest neighbor (kNN), support vector machine (SVM), logistic regression, and random forest] for binary classification between the error-free plan and the plan with the corresponding error for each type of error were developed. We used part of the total dataset to perform fourfold cross-validation to tune the models, and we used the remaining test dataset to evaluate the performance of the developed models. A gamma analysis was also performed between the measured and calculated fluence maps with the criteria of 3%/2 and 2%/2mm for all of the types of error. The radiomic features and its optimal number were similar for the models for the TF and the DLG error detection, which was different from the MLC positional error. The highest sensitivity was obtained as 0.913 for the TF error with SVM and logistic regression, 0.978 for the DLG error with kNN and SVM, and 1.000 for the MLC positional error with kNN, SVM, and random forest. The highest specificity was obtained as 1.000 for the TF error with a decision tree, SVM, and logistic regression, 1.000 for the DLG error with a decision tree, logistic regression, and random forest, and 0.909 for the MLC positional error with a decision tree and logistic regression. The gamma analysis showed the poorest performance in which sensitivities were 0.737 for the TF error and the DLG error and 0.882 for the MLC positional error for 3%/2mm. The addition of another type of error to fluence maps significantly reduced the sensitivity for the TF and the DLG error, whereas no effect was observed for the MLC positional error detection. Compared to the conventional gamma analysis, the radiomics-based machine learning models showed higher sensitivity and specificity in detecting a single type of the MLC modeling error and the MLC positional error. Although the developed models need further improvement for detecting multiple types of error, radiomics-based IMRT QA was shown to be a promising approach for detecting the MLC modeling error.

Applying a random projection algorithm to optimize machine learning model for predicting peritoneal metastasis in gastric cancer patients using CT images

Background and ObjectiveNon-invasively predicting the risk of cancer metastasis before surgery can play an essential role in determining which patients can benefit from neoadjuvant chemotherapy. This study aims to investigate and test the advantages of applying a random projection algorithm to develop and optimize a radiomics-based machine learning model to predict peritoneal metastasis in gastric cancer patients using a small and imbalanced computed tomography (CT) image dataset. MethodsA retrospective dataset involving CT images acquired from 159 patients is assembled, including 121 and 38 cases with and without peritoneal metastasis, respectively. A computer-aided detection scheme is first applied to segment primary gastric tumor volumes and initially compute 315 image features. Then, five gradients boosting machine (GBM) models embedded with five feature selection methods (including random projection algorithm, principal component analysis, least absolute shrinkage, and selection operator, maximum relevance and minimum redundancy, and recursive feature elimination) along with a synthetic minority oversampling technique, are built to predict the risk of peritoneal metastasis. All GBM models are trained and tested using a leave-one-case-out cross-validation method. ResultsResults show that the GBM model embedded with a random projection algorithm yields a significantly higher prediction accuracy (71.2%) than the other four GBM models (p<0.05). The precision, sensitivity, and specificity of this optimal GBM model are 65.78%, 43.10%, and 87.12%, respectively. ConclusionsThis study demonstrates that CT images of the primary gastric tumors contain discriminatory information to predict the risk of peritoneal metastasis, and a random projection algorithm is a promising method to generate optimal feature vector, improving the performance of machine learning based prediction models.

Radiomics-based machine learning model to predict risk of death within 5-years in clear cell renal cell carcinoma patients

PurposeThe aim of this study was to develop radiomics–based machine learning models based on extracted radiomic features and clinical information to predict the risk of death within 5 years for prognosis of clear cell renal cell carcinoma (ccRCC) patients. MethodsAccording to image quality and clinical data availability, we eventually selected 70 ccRCC patients that underwent CT scans. Manual volume-of-interest (VOI) segmentation of each image was performed by an experienced radiologist using the 3D slicer software package. Prior to feature extraction, image pre-processing was performed on CT images to extract different image features, including wavelet, Laplacian of Gaussian, and resampling of the intensity values to 32, 64 and 128 bin levels. Overall, 2544 3D radiomics features were extracted from each VOI for each patient. Minimum Redundancy Maximum Relevance (MRMR) algorithm was used as feature selector. Four classification algorithms were used, including Generalized Linear Model (GLM), Support Vector Machine (SVM), K-nearest Neighbor (KNN) and XGBoost. We used the Bootstrap resampling method to create validation sets. Area under the receiver operating characteristic (ROC) curve (AUROC), accuracy, sensitivity, and specificity were used to assess the performance of the classification models. ResultsThe best single performance among 8 different models was achieved by the XGBoost model using a combination of radiomic features and clinical information (AUROC, accuracy, sensitivity, and specificity with 95% confidence interval were 0.95–0.98, 0.93–0.98, 0.93–0.96 and ~1.0, respectively). ConclusionsWe developed a robust radiomics-based classifier that is capable of accurately predicting overall survival of RCC patients for prognosis of ccRCC patients. This signature may help identifying high-risk patients who require additional treatment and follow up regimens.

Application of MRI Radiomics-Based Machine Learning Model to Improve Contralateral BI-RADS 4 Lesion Assessment.

ObjectiveThis study aimed to explore the potential of magnetic resonance imaging (MRI) radiomics-based machine learning to improve assessment and diagnosis of contralateral Breast Imaging Reporting and Data System (BI-RADS) category 4 lesions in women with primary breast cancer.Materials and MethodsA total of 178 contralateral BI-RADS 4 lesions (97 malignant and 81 benign) collected from 178 breast cancer patients were involved in our retrospective dataset. T1 + C and T2 weighted images were used for radiomics analysis. These lesions were randomly assigned to the training (n = 124) dataset and an independent testing dataset (n = 54). A three-dimensional semi-automatic segmentation method was performed to segment lesions depicted on T2 and T1 + C images, 1,046 radiomic features were extracted from each segmented region, and a least absolute shrinkage and operator feature selection method reduced feature dimensionality. Three support vector machine (SVM) classifiers were trained to build classification models based on the T2, T1 + C, and fusion image features, respectively. The diagnostic performance of each model was evaluated and tested using the independent testing dataset. The area under the receiver operating characteristic curve (AUC) was used as a performance metric.ResultsThe T1+C image feature-based model and T2 image feature-based model yielded AUCs of 0.71 ± 0.07 and 0.69 ± 0.07 respectively, and the difference between them was not significant (P > 0.05). After fusing T1 + C and T2 imaging features, the proposed model’s AUC significantly improved to 0.77 ± 0.06 (P < 0.001). The fusion model yielded an accuracy of 74.1%, which was higher than that of the T1 + C (66.7%) and T2 (59.3%) image feature-based models.ConclusionThe MRI radiomics-based machine learning model is a feasible method to assess contralateral BI-RADS 4 lesions. T2 and T1 + C image features provide complementary information in discriminating benign and malignant contralateral BI-RADS 4 lesions.

External Validation Of A Radiomics-Based Machine Learning Model For Distinguishing Radiation Necrosis From Progression Of Brain Metastases Treated With Stereotactic Radiosurgery

Radiation necrosis (RN) is common and potentially debilitating after stereotactic radiosurgery (SRS) for brain metastases (BM). The goal of this study is to validate a previously reported radiomics signature for distinguishing RN from true progression (TP) using an independent image set. Patients with BM were treated with Gammaknife SRS at Wake Forest University between 2004 and 2012 (WF dataset). Those who developed radiographic evidence of progression in the treated lesion and subsequently had pathologic confirmation were included. The lesions were identified on T1 post-contrast (T1c) and T2 fluid attenuated inversion recovery (T2 FLAIR) MRI. 51 T1c and 51 T2 FLAIR radiomics features were extracted using a software developed in house. An IsoSVM machine learning classifier model trained using our institutional data (JH dataset) was tested using WF dataset. Fifty-three lesions from 43 patients were included in this study, of which all had T1c and 33 had T2 FLAIR images available. Seven lesions were pure RN pathologically, while the TP group included 37 pure tumor and 9 mixed tumor and necrosis. Twenty-two patients (51%) underwent whole brain radiotherapy with a median dose of 35Gy. Median SRS marginal dose was 18Gy (range 10-22Gy). There was no statistically significant difference among baseline characteristics between RN and TP. Of the 6 T1c radiomic features from the JH dataset, 4 demonstrated the same trend in the WF dataset (neighborhood grey tone difference matrix (NGTDM) texture strength, NGTDM coarseness, grey level run length matrix (GLRLM) grey level nonuniformity, and GLRLM run percentage). NGTDM texture strength (mean RN 161.13 vs TP 70.86, p = 0.06), and NGTDM coarseness (mean RN 0.03 vs TP 0.02, p = 0.07) approached significance. Of the 4 FLAIR features, 3 showed similar trend (kurtosis, minimum, NGTDM texture strength). The original IsoSVM model trained with all 6 T1c and 4 FLAIR features showed sensitivity of 65%, specificity of 87%, and AUC of 0.81. Due to the limited number of cases with FLAIR available, we re-trained the model with JH dataset using only the 4 T1c features showing the same trend between the two datasets, resulting in sensitivity and specificity of 80% and 73%, with AUC of 0.8. This model when tested on the WF dataset showed sensitivity and specificity of 71% and 52%, with AUC of 0.62. When the model was re-trained with the 4 T1c and 3 FLAIR features trending similarly between the two datasets, the sensitivity and specificity were 50% and 100%, and the AUC improved to 0.84. Radiomics profiling of post-SRS T1c and T2 FLAIR MRI showed promising trends between RN and TP in 4 T1c features across two independent datasets. Radiomics-based IsoSVM model can distinguish RN vs TP in individual patients. Statistical significance and model performance were inferior in the validation (WF) dataset due to smaller sample size and class imbalance with fewer RN cases. Further studies with larger datasets are needed to optimize and validate the model.

A Preliminary Study for Distinguish Hormone-Secreting Functional Adrenocortical Adenoma Subtypes Using Multiparametric CT Radiomics-Based Machine Learning Model and Nomogram.

Purpose: To explore the application value of multiparametric computed tomography (CT) radiomics in non-invasive differentiation between aldosterone-producing and cortisol-producing functional adrenocortical adenomas.Methods: This retrospective review analyzed 83 patients including 41 patients with aldosterone-producing adenoma and 42 patients with cortisol-producing adenoma. The quantitative radiomics features were extracted from the complete unenhanced, arterial, and venous phase CT images. A comparative study of several frequently used machine learning models (linear discriminant analysis, logistic regression, random forest, and support vector machine) combined with different feature selection methods was implemented in order to determine which was most advantageous for differential diagnosis using radiomics features. Then, the integrated model using the combination of radiomic signature and clinic–radiological features was built, and the associated calibration curve was also presented. The diagnostic performance of these models was estimated and compared using the area under the receiver operating characteristic (ROC) curve (AUC).Result: In the radiomics-based machine learning model, logistic regression model with LASSO (least absolute shrinkage and selection operator) outperformed the other models, which yielded a sensitivity of 0.935, a specificity of 0.823, and an accuracy of 0.887 [AUC = 0.882, 95% confidence interval (CI) = 0.819–0.945]. Moreover, the nomogram representing the integrated model achieved good discrimination performances, which yielded a sensitivity of 0.915, a specificity of 0.928, and an accuracy of 0.922 (AUC = 0.902, 95% CI = 0.822–0.982), and it was better than that of the radiomics model alone.Conclusion: This study found that the combination of multiparametric radiomics signature and clinic–radiological features can non-invasively differentiate the subtypes of hormone-secreting functional adrenocortical adenomas, which may have good potential for facilitating the diagnosis and treatment in clinical practice.

Machine learning-based analysis of [18F]DCFPyL PET radiomics for risk stratification in primary prostate cancer

PurposeQuantitative prostate-specific membrane antigen (PSMA) PET analysis may provide for non-invasive and objective risk stratification of primary prostate cancer (PCa) patients. We determined the ability of machine learning-based analysis of quantitative [18F]DCFPyL PET metrics to predict metastatic disease or high-risk pathological tumor features.MethodsIn a prospective cohort study, 76 patients with intermediate- to high-risk PCa scheduled for robot-assisted radical prostatectomy with extended pelvic lymph node dissection underwent pre-operative [18F]DCFPyL PET-CT. Primary tumors were delineated using 50–70% peak isocontour thresholds on images with and without partial-volume correction (PVC). Four hundred and eighty standardized radiomic features were extracted per tumor. Random forest models were trained to predict lymph node involvement (LNI), presence of any metastasis, Gleason score ≥ 8, and presence of extracapsular extension (ECE). For comparison, models were also trained using standard PET features (SUVs, volume, total PSMA uptake). Model performance was validated using 50 times repeated 5-fold cross-validation yielding the mean receiver-operator characteristic curve AUC.ResultsThe radiomics-based machine learning models predicted LNI (AUC 0.86 ± 0.15, p < 0.01), nodal or distant metastasis (AUC 0.86 ± 0.14, p < 0.01), Gleason score (0.81 ± 0.16, p < 0.01), and ECE (0.76 ± 0.12, p < 0.01). The highest AUCs reached using standard PET metrics were lower than those of radiomics-based models. For LNI and metastasis prediction, PVC and a higher delineation threshold improved model stability. Machine learning pre-processing methods had a minor impact on model performance.ConclusionMachine learning-based analysis of quantitative [18F]DCFPyL PET metrics can predict LNI and high-risk pathological tumor features in primary PCa patients. These findings indicate that PSMA expression detected on PET is related to both primary tumor histopathology and metastatic tendency. Multicenter external validation is needed to determine the benefits of using radiomics versus standard PET metrics in clinical practice.

Electron Density and Biologically Effective Dose (BED) Radiomics-Based Machine Learning Models to Predict Late Radiation-Induced Subcutaneous Fibrosis.

Purpose: to predict the occurrence of late subcutaneous radiation induced fibrosis (RIF) after partial breast irradiation (PBI) for breast carcinoma by using machine learning (ML) models and radiomic features from 3D Biologically Effective Dose (3D-BED) and Relative Electron Density (3D-RED).Methods: 165 patients underwent external PBI following a hypo-fractionation protocol consisting of 40 Gy/10 fractions, 35 Gy/7 fractions, and 28 Gy/4 fractions, for 73, 60, and 32 patients, respectively. Physicians evaluated toxicity at regular intervals by the Common Terminology Adverse Events (CTAE) version 4.0. RIF was assessed every 3 months after the completion of radiation course and scored prospectively. RIF was experienced by 41 (24.8%) patients after average 5 years of follow up.The Hounsfield Units (HU) of the CT-images were converted into relative electron density (3D-RED) and Dose maps into Biologically Effective Dose (3D-BED), respectively. Shape, first-order and textural features of 3D-RED and 3D-BED were calculated in the planning target volume (PTV) and breast. Clinical and demographic variables were also considered (954 features in total). Imbalance of the dataset was addressed by data augmentation using ADASYN technique. A subset of non-redundant features that best predict the data was identified by sequential feature selection. Support Vector Machines (SVM), ensemble machine learning (EML) using various aggregation algorithms and Naive Bayes (NB) classifiers were trained on patient dataset to predict RIF occurrence. Models were assessed using sensitivity and specificity of the ML classifiers and the area under the receiver operator characteristic curve (AUC) of the score functions in repeated 5-fold cross validation on the augmented dataset.Results: The SVM model with seven features was preferred for RIF prediction and scored sensitivity 0.83 (95% CI 0.80–0.86), specificity 0.75 (95% CI 0.71–0.77) and AUC of the score function 0.86 (0.85–0.88) on cross-validation. The selected features included cluster shade and Run Length Non-uniformity of breast 3D-BED, kurtosis and cluster shade from PTV 3D-RED, and 10th percentile of PTV 3D-BED.Conclusion: Textures extracted from 3D-BED and 3D-RED in the breast and PTV can predict late RIF and may help better select patient candidates to exclusive PBI.