PET Features Research Articles

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.

Abstract PS13-50: Relationship of dedicated breast PET and MRI features in breast cancer patients receiving neoadjuvant chemotherapy

Abstract Introduction: Dedicated breast positron emission tomography (dbPET) is an emerging imaging technique with the spatial resolution needed to assess functionality and intra-tumor heterogeneity in primary breast lesions. Breast cancer patients may benefit from dbPET imaging combined with molecularly targeted agents to non-invasively assess and predict response to targeted therapy in the neoadjuvant treatment setting. We have previously observed that [18F]-fluorodeoxyglucose (FDG) PET provides tumor metabolic information complementary to the angiogenic properties reflected by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) for characterizing triple-negative breast cancers (TNBC) (1). In this study, we examined the relationship between FDG-dbPET and MRI features in a cohort of breast cancer patients receiving neoadjuvant chemotherapy (NAC). Methods: With institutional review board approval, patients with biopsy-proven locally-advanced breast cancer were imaged with breast MRI and dbPET before (T0) and after three weeks (T1) of NAC. Standard DCE-MRI was obtained using a dedicated breast coil. Patients also underwent dbPET with 5 mCi of FDG at 45 minutes post-injection. Functional tumor volumes (FTV) were calculated from DCE-MRI by summing all voxels with an early percent enhancement (PE) exceeding 70% within a manually defined volume of interest (VOI). Maximum and mean PE (PEMax, PEMean) values within the VOI were also computed for analyses. Tumors were segmented in dbPET images using semi-automated threshold-driven methods. Body weight-corrected maximum and mean standardized uptake values (SUVMax, SUVMean), total lesion glycolysis (TLG), and metabolic tumor volume (MTV) were calculated for FDG-dbPET. Percent change relative to T0 (Δ = 100*(T1 - T0)/T0) was calculated for each feature. Spearman’s correlation coefficient was used to evaluate the relationship between MRI and dbPET features. Results: Of the 16 patients enrolled in this study, 13 patients (N = 15 unique tumors) with MRI and dbPET at T0 and T1 were included in the analysis. 46% (6/13) of the patients had TNBC. Our initial findings indicated that ΔPEMax and ΔSUVMax had the highest correlation (ρ = 0.59, p = 0.022). FTV and TLG at T1 were also correlated (ρ = 0.56, p = 0.032). Among all imaging features, ΔMTV showed the largest post-treatment difference between TNBC (-54.5%) and non-TNBC (-6.06%) groups. Among MRI features, ΔFTV exhibited the largest difference between the groups: -70.4% in TNBC and -43.1% in non-TNBC. ΔSUVMax and ΔTLG were additional dbPET features with large differences between TNBC and non-TNBC patients (Table 1). Conclusion: This exploratory study suggests that post-treatment ΔSUVMax and TLG provide complementary metabolic information to angiogenic properties (ΔPEMax and FTV, respectively) reflected by MRI. Other dbPET features may provide independent information adjunct to MRI for describing primary breast tumors. Patients with TNBC exhibited larger reductions in FDG uptake values and metabolic volume than non-TNBC patients. These observed reductions may improve early treatment response in patients with TNBC, enabling more precise treatment guidance. Further studies in larger cohorts are needed to validate these initial observations. 1. Bolouri MS, et al. Triple-Negative and Non-Triple-Negative Invasive Breast Cancer: Association between MR and Fluorine 18 Fluorodeoxyglucose PET Imaging. Radiology 2013;269:354-61 Comparison of ΔMRI and ΔFDG-dbPET in TNBC vs non-TNBC patientsAll Tumors (N = 15 tumors) Median(IQR)TNBC (N = 6 tumors) Median(IQR)non-TNBC (N = 9 tumors) Median(IQR)DCE-MRIΔFTV (%)-66.1 (-77.6, -19.1)-70.4 (-79.0, -62.1)-43.1 (-72.5, -2.95)ΔPEMax (%)-9.91 (-31.5, 19.5)-10.3 (-26.3, 24.2)-9.91 (-31.8, 8.42)ΔPEMean (%)-10.7 (-22.4, 2.77)-9.57 (-12.8, 0.29)-17.0 (-26.0, 2.33)FDG-dbPETΔSUVMax (%)-31.6 (-53.9, -20.6)-47.3 (-55.7, -41.1)-23.1 (-31.6, 1.13)ΔSUVMean (%)-34.1 (-65.4, -14.2)-48.5 (-74.6, -9.74)-34.1 (-44.5, -14.8)ΔMTV (%)-6.18 (-57.0, 38.5)-54.5 (-75.4, 15.3)-6.06 (-47.2, 38.9)ΔTLG (%)-64.9 (-75.2, 23.3)-75.2 (-84.0, -60.0)-47.9 (-65.2, 31.1) Citation Format: Deep K Hathi, Ella F Jones, Wen Li, David C Newitt, Ruby Guo, Youngho Seo, Robert R Flavell, Bonnie N Joe, Diane Heditsian, Susie Brain, ISPY-2 Imaging Working Group, ISPY-2 Consortium, Laura J Esserman, Nola M Hylton. Relationship of dedicated breast PET and MRI features in breast cancer patients receiving neoadjuvant chemotherapy [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS13-50.

Radiomics analysis of 18F-Choline PET/CT in the prediction of disease outcome in high-risk prostate cancer: an explorative study on machine learning feature classification in 94 patients.

The aim of this study was (1) to investigate the application of texture analysis of choline PET/CT images in prostate cancer (PCa) patients and (2) to propose a machine-learning radiomics model able to select PET features predictive of disease progression in PCa patients with a same high-risk class at restaging. Ninety-four high-risk PCa patients who underwent restaging Cho-PET/CT were analyzed. Follow-up data were recorded for a minimum of 13 months after the PET/CT scan. PET images were imported in LIFEx toolbox to extract 51 features from each lesion. A statistical system based on correlation matrix and point-biserial-correlation coefficient has been implemented for features reduction and selection, while Discriminant analysis (DA) was used as a method for features classification in a whole sample and sub-groups for primary tumor or local relapse (T), nodal disease (N), and metastatic disease (M). In the whole group, 2 feature (HISTO_Entropy_log10; HISTO_Energy_Uniformity) results were able to discriminate the occurrence of disease progression at follow-up, obtaining the best performance in DA classification (sensitivity 47.1%, specificity 76.5%, positive predictive value (PPV) 46.7%, and accuracy 67.6%). In the sub-group analysis, the best performance in DA classification for T was obtained by selecting 3 features (SUVmin; SHAPE_Sphericity; GLCM_Correlation) with a sensitivity of 91.6%, specificity 84.1%, PPV 79.1%, and accuracy 87%; for N by selecting 2 features (HISTO = _Energy Uniformity; GLZLM_SZLGE) with a sensitivity of 68.1%, specificity 91.4%, PPV 83%, and accuracy 82.6%; and for M by selecting 2 features (HISTO_Entropy_log10 - HISTO_Entropy_log2) with a sensitivity 64.4%, specificity 74.6%, PPV 40.6%, and accuracy 72.5%. This machine learning model demonstrated to be feasible and useful to select Cho-PET features for T, N, and M with valuable association with high-risk PCa patients' outcomes. • Artificial intelligence applications are feasible and useful to select Cho-PET features. • Our model demonstrated the presence of specific features for T, N, and M with valuable association with high-risk PCa patients' outcomes. • Further prospective studies are necessary to confirm our results and to develop the application of artificial intelligence in PET imaging of PCa.

Adaptive feature selection in PET scans based on shared information and multi-label learning

Feature selection for tumor treatment outcome prediction in PET scans amounts to determine the best predictors in order to classify treatment outcomes for unseen data. Several challenges have to be addressed, notably the dissensus about the most predictive radiomic features and the relatively small-sized and imbalanced training samples. To overcome these issues, we propose a multi-label learning (MLL) problem formulation that we associated to the technique of feature selection with shared information among multiple tasks. In fact, we opt to simultaneously mining the shared knowledge across similar tasks from MRI and fused PET/MRI images in order to sort PET features according to their relevancy. Then, by repeatedly training an SVM model, different feature sets are elected to be informative for each iteration. Thus, by associating these sets with their correspondents in other iterations, the MLL formulation is performed in order to adequately adapt both feature sets and training instances according to their fit with unseen data. The adaptive synthetic sampling technique is adapted to the framework of MLL data balance. After that, the global and local label correlation technique is used for multi-labeling, while taking into consideration the capacity of MLL on recommending the best sets of features and training instances. Finally, an SVM classifier is applied with both selected features and training sub-samples in order to output prediction results. Conducted experiments show that attributing each feature set to a particular set of instances and training the model with balanced subset of the whole training data allows to outperform relevant methods from the state of the art.

Pleomorphic carcinoma of the lung: Prognostic models of semantic, radiomics and combined features from CT and PET/CT in 85 patients

IntroductionTo demonstrate semantic, radiomics, and the combined risk models related to the prognoses of pulmonary pleomorphic carcinomas (PCs). MethodsWe included 85 patients (M:F = 71:14; age, 35–88 [mean, 63 years]) whose imaging features were divided into training (n = 60) and test (n = 25) sets. Nineteen semantic and 142 radiomics features related to tumors were computed. Semantic risk score (SRS) model was built using the Cox-least absolute shrinkage and selection operator (LASSO) approach. Radiomics risk score (RRS) from CT and PET features and combined risk score (CRS) adopting both semantic and radiomics features were also constructed. Risk groups were stratified by the median of the risk scores of the training set. Survival analysis was conducted with the Kaplan-Meier plots. ResultsOf 85 PCs, adenocarcinoma was the most common epithelial component found in 63 (73 %) tumors. In SRS model, four features were stratified into high- and low-risk groups (HR, 4.119; concordance index ([C-index], 0.664) in the test set. In RRS model, five features helped improve the stratification (HR, 3.716; C-index, 0.591) and in CRS model, three features helped perform the best stratification (HR, 4.795; C-index, 0.617). The two significant features of CRS models were the SUVmax and the histogram feature of energy ([CT Firstorder Energy]). ConclusionIn PCs of the lungs, the combined model leveraging semantic and radiomics features provides a better prognosis compared to using semantic and radiomics features separately. The high SUVmax of solid portion (CT Firstorder Energy) of tumors is associated with poor prognosis in lung PCs.

Incorporating radiomic feature of pretreatment 18F-FDG PET improves survival stratification in patients with EGFR-mutated lung adenocarcinoma.

BackgroundTo investigate the survival prognostic value of the radiomic features of 18F-FDG PET in patients who had EGFR (epidermal growth factor receptor) mutated lung adenocarcinoma and received targeted TKI (tyrosine kinase inhibitor) treatment.MethodsFifty-one patients with stage III-IV lung adenocarcinoma and actionable EGFR mutation who received first-line TKI were retrospectively analyzed. All patients underwent pretreatment 18F-FDG PET/CT, and we calculated the PET-derived radiomic features. Cox proportional hazard model was used to examine the association between the radiomic features and the survival outcomes, including progression-free survival (PFS) and overall survival (OS). A score model was established according to the independent prognostic predictors and we compared this model to the TNM staging system using Harrell's concordance index (c-index).ResultsForty-eight patients (94.1%) experienced disease progression and 41 patients (80.4%) died. Primary tumor SUV entropy > 5.36, and presence of pleural effusion were independently associated with worse OS (both p < 0.001) and PFS (p = 0.001, and 0.003, respectively). We used these two survival predictors to devise a scoring system (score 0–2). Patients with a score of 1 or 2 had a worse survival than those with a score of 0 (HR for OS: 3.6, p = 0.006 for score 1, and HR: 21.8, p < 0.001 for score 2; HR for PFS: 2.2, p = 0.027 for score 1 and HR: 8.8, p < 0.001 for score 2). Our scoring system surpassed the TNM staging system (c-index = 0.691 versus 0.574, p = 0.013 for OS, and c-index = 0.649 versus 0.517, p = 0.004 for PFS).ConclusionsIn this preliminary study, combining PET radiomics with clinical risk factors may improve survival stratification in stage III-IV lung adenocarcinoma with actionable EFGR mutation. Our proposed scoring system may assist with optimization of individualized treatment strategies in these patients.

NIMG-38. NON-INVASIVE PREDICTION OF MGMT PROMOTER METHYLATION USING COMBINED FET PET/MRI RADIOMICS

Abstract BACKGROUND Recently, the Response Assessment in Neuro-Oncology (RANO) Working Group emphasized the additional diagnostic value of amino acid PET in addition to MRI. However, the number of studies using amino acid PET/MRI radiomics is still low. We investigated the potential of combined O-(2-[18F]fluoroethyl)-L-tyrosine (FET) PET/MRI radiomics for the non-invasive prediction of the O6-methylguanine-DNA methyl-transferase (MGMT) promoter methylation status in glioma patients. METHODS Seventy-one patients with newly diagnosed glioma (predominantly WHO grade III and IV glioma, 82%) underwent a hybrid FET PET/MRI scan. Forty-six patients (65%) had a methylated MGMT promoter. The tumor and tumor subregions were manually segmented on conventional MRI. In total, 199 standardized features were obtained from FET PET, contrast-enhanced T1-weighted, T2-weighted, and fluid attenuated inversion recovery (FLAIR) MRI. After feature extraction and data normalization, patients were randomly assigned to a training and a test dataset for final model evaluation in a ratio of 70/30, with a balanced distribution of the MGMT promoter methylation status. Feature selection was performed by recursive feature elimination using random forest regressors. For the final model generation, the number of features was limited to seven to avoid data overfitting. Different algorithms for model generation were compared, and the model performance in the training data was assessed by 5-fold cross-validation. Finally, the best performing models were applied to the test dataset to evaluate the robustness of the models. RESULTS In the test dataset, the best radiomics signatures obtained from MRI or FET PET alone achieved diagnostic accuracies for the prediction of the MGMT promoter methylation of 64% and 70%, respectively. In contrast, the highest diagnostic accuracy of 83% was obtained by combining FET PET and MRI features. CONCLUSION Combined FET PET/MRI radiomics allows the non-invasive prediction of the MGMT promoter methylation status in patients with gliomas, providing more diagnostic information than either modality alone.

A Pre-Treatment PET And CT-Derived Model for Progression-Free Survival in Early Stage Non-Small Cell Lung Cancer

The maximum value of the standardized uptake value (SUV) and tumor diameter have been associated with progression after stereotactic body radiotherapy (SBRT) for early stage non-small cell lung cancer (NSCLC). In this work, we explored whether other tumor parameters based on pre-SBRT PETs and CTs predict progression-free survival (PFS) in this patient population. 412 patients treated between 2006 and 2017 were included. Patients had to have PETs and CTs available within three months prior to SBRT start. The median prescription dose was 50Gy in 5 fractions. The planning-CT gross tumor volumes (GTVs) were propagated onto the pre-treatment PETs and CTs using b-spline deformable image registration. PET intensity features (90th percentile, entropy, maximum, mean, peak, robust mean absolute deviation, SD, valley) and CT shape features (compactness, diameter, elliptic axes, elongation, flatness, number of lobules/peaks/spicules, sphericity, surface area, surface to volume ratio, volume) were extracted. Data were split into training and hold-out validation subsets (n = 283, 123; 70%, 30%). In the training subset, the imaging features and six patient characteristics (age, gender, histology, performance status, prior surgery, tumor location) were tested for association with PFS using Cox Proportional Hazards regression with re-sampling (bootstrapping with 1000 samples). Significance was denoted at p≤0.0019 (corrected for 26 tests). A bootstrapped forward-stepwise multivariate analysis was undertaken including only non-strongly correlated predictors (Spearman’s rank, |Rs|<0.70). The most frequently selected model was explored in the validation subset in which model performance was assessed using the c-index and the prediction-stratified high and low risk tertiles (HR, LR) of the observed PFS were compared. Nineteen of the 20 identified candidate predictors were either PET or CT features (p-value range: 3E-9, 1.2E-3). The intra-imaging modality correlation between features was strong (median |Rs|: PET: 0.93; CT: 0.76) and only four features were passed on to multivariate analysis: PET entropy, CT number of peaks, CT major axis, and gender. The most frequently selected model included PET entropy and CT number of peaks; the c-index in the validation subset was 0.77 and the prediction-stratified survival indicated a clear separation between the observed HR and LR: e.g. a PFS of 60% was observed at 12 months in HR vs. 22 months in LR. This PET and CT-based model identified the SUV distribution randomness (entropy) and spiculated tumor pattern on CTs as the most important features in predicting PFS in early stage NSCLC. The associated performance on the hold-out validation subset was good and its use has the potential to further improve the prediction of response to SBRT for this patient population. This model will be used to identify high-risk patients based on the predicted PFS in an upcoming phase II study on adjuvant immunotherapy.

Roles of Clinical and Image-Based Features for Outcomes in Larynx Cancer Patients

In advanced larynx cancer, the use of induction chemotherapy (IC) to select patients for organ preservation offers outcomes comparable to upfront laryngectomy. Assessment of tumor response to IC is performed with direct laryngoscopy (DL), and barriers to this approach exist. In this study, we retrospectively investigate radiomic and clinical features as predictors of OS and laryngectomy free survival (LFS). Larynx cancer patients from a single institution with usable CT and PET scans were included. Clinical variables investigated were tumor subsite, T-stage, N-stage, neutrophil-lymphocyte ratio (NLR), and lymphocyte-monocyte ratio (LMR). CT-based variables were pre-IC GTV, GTV change between pre and post IC (ΔGTV), and 43 radiomic texture features. Cox proportional hazards was utilized for OS and LFS; logistic regression was used for IC response. Feature selection for each model was performed by identifying variables with a significant (p ≤ 0.05) Spearman rank. Model accuracy was assessed using the average concordance index from 100 iterations of stratified, 5-fold cross validation. Of 84 patents with CT, and clinical features, 5 features were found to have significant correlations with surgeon-assessed IC response: 3 texture-based radiomic features (coarseness, busyness, grey level nonuniformity (GLN)), ΔGTV, and N-stage. The latter 3 features contributed significantly to the final model (see table). For OS, only N-stage had significant correlation. For LFS, 8 features were found to have significant correlations, including 5 texture features (coarseness, busyness, strength, grey level variance (GLV), large zone emphasis (LZE)), NLR, LMR, and ΔGTV. Of these 8 features, 2 contributed significantly to the final model (see table). When including PET features for 54 patients, only LFS was found to correlate significantly with PET-based features. Of 12 features found to have significant correlation with LFS, 5 contributed significantly to the final model: PET contrast (hazard ratio (HR) = 1.70, p = 0.02), PET correlation (HR = 3790, p = 0.03), PET dissimilarity (HR = 1.20E-05, p = 0.048), ΔGTV (HR = 0.959, p = 0.001), and NLR (HR = 3.67, p = 0.04). The average concordance index for this model was 0.628 ± 0.149. In this study, we investigate the role of CT-based, PET-based, and clinical features as predictors for OS, LFS, and surgeon-assessed response to IC in a bioselection approach to larynx cancer. Initial results indicate that several features may have value as predictors for OS, LFS, and IC response. Due to limited data size, further studies with larger datasets are necessary to validate these results and further explore the potential of these variables.Tabled 1Abstract 2698; TableModelFeatureHazard/Odds ratiop-valueMean C-idxICGLN0.4750.0280.808 ± 0.099ΔGTV5.23< 0.001N-stage1.860.003OSN-stage1.8740.010.643 ± 0.102LFSΔGTV0.578< 0.001NLR2.740.0050.664 ± 0.101 Open table in a new tab

Machine learning based on clinico-biological features integrated 18F-FDG PET/CT radiomics for distinguishing squamous cell carcinoma from adenocarcinoma of lung

PurposeTo develop and validate a clinico-biological features and 18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) radiomic-based nomogram via machine learning for the pretherapy prediction of discriminating between adenocarcinoma (ADC) and squamous cell carcinoma (SCC) in non-small cell lung cancer (NSCLC).MethodsA total of 315 NSCLC patients confirmed by postoperative pathology between January 2017 and June 2019 were retrospectively analyzed and randomly divided into the training (n = 220) and validation (n = 95) sets. Preoperative clinical factors, serum tumor markers, and PET, and CT radiomic features were analyzed. Prediction models were developed using the least absolute shrinkage and selection operator (LASSO) regression analysis. The performance of the models was evaluated and compared by the area under receiver-operator characteristic (ROC) curve (AUC) and DeLong test. The clinical utility of the models was determined via decision curve analysis (DCA). Then, a nomogram was developed based on the model with the best predictive efficiency and clinical utility and was validated using the calibration plots.ResultsIn total, 122 SCC and 193 ADC patients were enrolled in this study. Four independent prediction models were separately developed to differentiate SCC from ADC using clinical factors-tumor markers, PET radiomics, CT radiomics, and their combination. The DeLong test and DCA showed that the Combined Model, consisting of 2 clinical factors, 2 tumor markers, 7 PET radiomics, and 3 CT radiomic parameters, held the highest predictive efficiency and clinical utility in predicting the NSCLC subtypes compared with the use of these parameters alone in both the training and validation sets (AUCs (95% CIs) = 0.932 (0.900–0.964), 0.901 (0.840–0.957), respectively) (p < 0.05). A quantitative nomogram was subsequently constructed using the independently risk factors from the Combined Model. The calibration curves indicated a good consistency between the actual observations and nomogram predictions.ConclusionThis study presents an integrated clinico-biologico-radiological nomogram that can be accurately and noninvasively used for the individualized differentiation SCC from ADC in NSCLC, thereby assisting in clinical decision making for precision treatment.

The Role of Texture Analysis and Radiomics in Prediction of Bone Marrow Infiltration in Patients with Lymphoma

Purpose. To explore the ability of image-based parameters with texture parameters in the differentiation of bone marrow infiltration by positron emission tomography-computed tomography (PET/CT). Procedures. We retrospectively evaluated the baseline PET/CT scan of 44 patients with histologically proven lymphoma. Seventy-three of features were extracted using CGITA software and statistical analysis were carried out on SPSS program. Results. Spearman correlation analysis revealed a strong positive correlation between conventional PET metrics and texture features (19 significant association features) and inverse correlation was found (one parameter significant association feature). Area under the curve and p-value of receive operating characteristics showed that (HILRE (4-bin), HILZE (64-bin), LRE (64-bin), LZE (64-bin), max spectrum (8-bin), busyness (64-bin),code similarity (32-bin & 64-bin)) were significant discriminator of bone marrow infiltration among other features (AUC>0.682, p < 0.05). Univariate analyses of texture features showed that code similarity and LRE (both 64 bin) has significant prediction to BMI. Multivariate analyses revealed that LRE (64 bin) p= 0.031; odds ratio: 1.022; 95% CI, 1.002-1.043) were independent variables for bone marrow infiltration. Conclusions. Significant associations emerged between PET features and bone marrow infiltration in lymphoma. Texture analysis on PET/CT shows potential to differentiate between bone marrow infiltration in patients with Lymphoma.

Prognostic value of FDG-PET radiomics with machine learning in pancreatic cancer

Patients with pancreatic cancer have a poor prognosis, therefore identifying particular tumor characteristics associated with prognosis is important. This study aims to investigate the utility of radiomics with machine learning using 18F-fluorodeoxyglucose (FDG)-PET in patients with pancreatic cancer. We enrolled 161 patients with pancreatic cancer underwent pretreatment FDG-PET/CT. The area of the primary tumor was semi-automatically contoured with a threshold of 40% of the maximum standardized uptake value, and 42 PET features were extracted. To identify relevant PET parameters for predicting 1-year survival, Gini index was measured using random forest (RF) classifier. Twenty-three patients were censored within 1 year of follow-up, and the remaining 138 patients were used for the analysis. Among the PET parameters, 10 features showed statistical significance for predicting overall survival. Multivariate analysis using Cox HR regression revealed gray-level zone length matrix (GLZLM) gray-level non-uniformity (GLNU) as the only PET parameter showing statistical significance. In RF model, GLZLM GLNU was the most relevant factor for predicting 1-year survival, followed by total lesion glycolysis (TLG). The combination of GLZLM GLNU and TLG stratified patients into three groups according to risk of poor prognosis. Radiomics with machine learning using FDG-PET in patients with pancreatic cancer provided useful prognostic information.

Repeatability of Radiomic Features of Brown Adipose Tissue.

The aim of this study was to assess the repeatability of activated brown adipose tissue (BAT) radiomic features. To decipher radiomic features that may provide useful information on BAT, the impact of reconstruction methods and imaging modality choice was also evaluated. Methods: Twenty-seven healthy adults enrolled in this study. After a cooling procedure to activate BAT, volunteers underwent 18F-FDG imaging. Participants underwent repeat imaging using the same imaging protocols and a similar 18F-FDG dose within 14 d. Active BAT was segmented using the BARCIST 1.0 methods. Radiomic features were extracted from each region of interest on high-definition PET (HD PET), non-HD PET, and CT images. Lin's concordance correlation coefficient was used to estimate the repeatability of the extracted radiomic features. To determine whether BAT radiomic feature repeatability correlates with BAT SUVmax repeatability, participants were stratified based on the relative difference in SUVmax between sessions. Non-HD PET repeatable features were clustered together using hierarchical clustering, and the normalized dynamic range of each feature was computed to identify the most informative feature within each cluster. Results: Eighteen of the 27 volunteers had sufficient BAT activity for radiomic analysis. Sixty-six HD PET, 66 non-HD PET, and 6 CT features showed high repeatability (concordance correlation coefficient ≥ 0.80). Feature repeatability was significantly higher for PET than for CT, but there was no statistically significant difference between HD and non-HD PET in radiomic feature repeatability. The repeatability of radiomic features extracted from each modality and reconstruction method type followed the trend in SUVmax, as participants with lower relative differences in SUVmax between initial and repeated imaging sessions had higher radiomic feature repeatability. Hierarchical clustering of the high-repeatability PET features resulted in 10 highly correlated clusters (R2 ≥ 0.95). Seven features, including SUVmax, did not cluster with any other features. Conclusion: Several clusters of highly repeatable BAT radiomic features derived from 18F-FDG PET/CT appear to provide information regarding BAT activity distinct from SUVmax These features might be explored as quantitative imaging biomarkers of BAT activity in future studies.

A preliminary PET radiomics study of brain metastases using a fully automatic segmentation method

BackgroundPositron Emission Tomography (PET) is increasingly utilized in radiomics studies for treatment evaluation purposes. Nevertheless, lesion volume identification in PET images is a critical and still challenging step in the process of radiomics, due to the low spatial resolution and high noise level of PET images. Currently, the biological target volume (BTV) is manually contoured by nuclear physicians, with a time expensive and operator-dependent procedure.This study aims to obtain BTVs from cerebral metastases in patients who underwent L-[11C]methionine (11C-MET) PET, using a fully automatic procedure and to use these BTVs to extract radiomics features to stratify between patients who respond to treatment or not. For these purposes, 31 brain metastases, for predictive evaluation, and 25 ones, for follow-up evaluation after treatment, were delineated using the proposed method. Successively, 11C-MET PET studies and related volumetric segmentations were used to extract 108 features to investigate the potential application of radiomics analysis in patients with brain metastases. A novel statistical system has been implemented for feature reduction and selection, while discriminant analysis was used as a method for feature classification.ResultsFor predictive evaluation, 3 features (asphericity, low-intensity run emphasis, and complexity) were able to discriminate between responder and non-responder patients, after feature reduction and selection. Best performance in patient discrimination was obtained using the combination of the three selected features (sensitivity 81.23%, specificity 73.97%, and accuracy 78.27%) compared to the use of all features. Secondly, for follow-up evaluation, 8 features (SUVmean, SULpeak, SUVmin, SULpeak prod-surface-area, SUVmean prod-sphericity, surface mean SUV 3, SULpeak prod-sphericity, and second angular moment) were selected with optimal performance in discriminant analysis classification (sensitivity 86.28%, specificity 87.75%, and accuracy 86.57%) outperforming the use of all features.ConclusionsThe proposed system is able i) to extract 108 features for each automatically segmented lesion and ii) to select a sub-panel of 11C-MET PET features (3 and 8 in the case of predictive and follow-up evaluation), with valuable association with patient outcome. We believe that our model can be useful to improve treatment response and prognosis evaluation, potentially allowing the personalization of cancer treatment plans.

Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy.

Gallium-68 prostate-specific membrane antigen positron emission tomography (68Ga-PSMA-PET) is a highly sensitive method to detect prostate cancer (PC) metastases. Visual discrimination between malignant and physiologic/unspecific tracer accumulation by a nuclear medicine (NM) specialist is essential for image interpretation. In the future, automated machine learning (ML)-based tools will assist physicians in image analysis. The aim of this work was to develop a tool for analysis of 68Ga-PSMA-PET images and to compare its efficacy to that of human readers. Five different ML methods were compared and tested on multiple positron emission tomography/computed tomography (PET/CT) data-sets. Forty textural features extracted from both PET- and low-dose CT data were analyzed. In total, 2419 hotspots from 72 patients were included. Comparing results from human readers to those of ML-based analyses, up to 98% area under the curve (AUC), 94% sensitivity (SE), and 89% specificity (SP) were achieved. Interestingly, textural features assessed in native low-dose CT increased the accuracy significantly. Thus, ML based on 68Ga-PSMA-PET/CT radiomics features can classify hotspots with high precision, comparable to that of experienced NM physicians. Additionally, the superiority of multimodal ML-based analysis considering all PET and low-dose CT features was shown. Morphological features seemed to be of special additional importance even though they were extracted from native low-dose CTs.

Abstract 868: Prediction of clinical benefit to checkpoint blockade in advanced NSCLC patients using radiomics of PET/CT images

Abstract Immunotherapy has emerged as a one of the most successful treatment approach in advanced Non-small Cell Lung Cancer (NSCLC) patients. However, only 25-50% patients obtain durable clinical benefit (DCB), and 7-43% patient may suffer from immune-related severe adverse events. Therefore robust biomarkers that are predictive of response immune-checkpoint blockades at baseline are needed. Though PD-L1 expression by immunohistochemistry (IHC) is the only clinically-approved biomarker to trigger treatment decisions, sampling bias and the dynamic nature of PD-L1 expression can affect the final results. In this study, we investigated the value of PET/CT images in the predicting progression-free survival (PFS) and overall survival (OS) of immunotherapy in advanced NSCLC patients. The retrospective arm of this study included 400 patients with histologically-confirmed NSCLC, and these were split into training (N=284) and test (N=116) cohorts. A residual-convolutional-network model was developed to generate a deep learning score (PDL1-DLS) to predict the PD-L1 expression status with PET/CT images. Another 146 retrospective patients with histologically confirmed stage IIIB-IV NSCLC treated with immunotherapy (IO) were split into training (N=99) and test (N=47) cohorts, and a multiparametric radiomics signature (mpRS) generated from conventional PET and CT radiomics features was developed to predict DCB using an improved least absolute shrinkage and selection operator (LASSO) method. Subsequently, multivariable Cox regression models for PFS and OS prediction were trained with the mpRS, PDL1-DLS, and clinical characteristics, and these were further validated with 48 prospective IO-treated NSCLC patients. The PDL1-DLS significantly discriminated PD-L1 positive and negative patients with AUCs of 0.89 (95%CI:0.84-0.94) and 0.84 (95%CI:0.76-0.92) for training and test cohorts. The mpRS could predict DCB with AUCs of 0.86 (95%CI:0.79-0.94) and 0.83 (95%CI:0.71-0.94) in the training and test cohorts. In both retrospective and prospective IO-treated cohorts, the combination of binarized PDL1-DLS and binarized mpRS significantly improved the prediction of PFS and OS compared to single binarized PDL1-DLS (PFS: p&amp;lt;0.001, OS: p&amp;lt;0.001) or single binarized mpRS (PFS: p&amp;lt;0.040, OS: p&amp;lt;0.10). Subsequent multivariable Cox regression models identified binarized mpRS, binarized PDL1-DLS, ECOG, and histology as significant independent variables, and could achieve significant higher C-indices of 0.75 (95%CI:0.70-0.79) and 0.79 (95%CI:0.73-0.86) in prediction of PFS, 0.79 (95%CI:0.73-0.85) and 0.76 (95%CI:0.61-0.92) in prediction of OS in the retrospective and prospective IO-treated cohorts. We have therefore observed that, prior to therapy initiation, PET/CT fusion images can be used independently to predict PD-L1 status or DCB of immunotherapy in advanced NSCLC patients. Both of these models can be leveraged to improve more precise and individualized decision support in the treatment of patients with advanced NSCLC. Citation Format: Wei Mu, Ilke Tunali, Jhanelle E. Gray, Jin Qi, Matthew B. Schabath, Robert J. Gillies. Prediction of clinical benefit to checkpoint blockade in advanced NSCLC patients using radiomics of PET/CT images [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 868.

Metabolic radiogenomics in lung cancer: associations between FDG PET image features and oncogenic signaling pathway alterations

This study investigated the associations between image features extracted from tumor 18F-fluorodeoxyglucose (FDG) uptake and genetic alterations in patients with lung cancer. A total of 137 patients (age, 62.7 ± 10.2 years) who underwent FDG positron emission tomography/computed tomography (PET/CT) and targeted deep sequencing analysis for a tumor lesion, comprising 61 adenocarcinoma (ADC), 31 squamous cell carcinoma (SQCC), and 45 small cell lung cancer (SCLC) patients, were enrolled in this study. From the tumor lesions, 86 image features were extracted, and 381 genes were assessed. PET features were associated with genetic mutations: 41 genes with 24 features in ADC; 35 genes with 22 features in SQCC; and 43 genes with 25 features in SCLC (FDR < 0.05). Clusters based on PET features showed an association with alterations in oncogenic signaling pathways: Cell cycle and WNT signaling pathways in ADC (p = 0.023, p = 0.035, respectively); Cell cycle, p53, and WNT in SQCC (p = 0.045, 0.009, and 0.029, respectively); and TGFβ in SCLC (p = 0.030). In addition, SUVpeak and SUVmax were associated with a mutation of the TGFβ signaling pathway in ADC (FDR = 0.001, < 0.001). In this study, PET image features had significant associations with alterations in genes and oncogenic signaling pathways in patients with lung cancer.

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.

FDG PET versus CT radiomics to predict outcome in malignant pleural mesothelioma patients

BackgroundCareful selection of malignant pleural mesothelioma (MPM) patients for curative treatment is of highest importance, as the multimodal treatment regimen is challenging for patients and harbors a high risk of substantial toxicity. Radiomics—a quantitative method for image analysis—has shown its prognostic ability in different tumor entities and could therefore play an important role in optimizing patient selection for radical cancer treatment. So far, radiomics as a prognostic tool in MPM was not investigated.Materials and methodsThis study is based on 72 MPM patients treated with surgery in a curative intent at our institution between 2009 and 2017. Pre-treatment Fluorine-18 fluorodeoxyglucose (FDG) PET and CT scans were used for radiomics outcome modeling. After extraction of 1404 CT and 1410 FDG PET features from each image, a preselection by principal component analysis was performed to include only robust, non-redundant features for the cox regression to predict the progression-free survival (PFS) and the overall survival (OS). Results were validated on a separate cohort. Additionally, SUVmax and SUVmean, and volume were tested for their prognostic ability for PFS and OS.ResultsFor the PFS a concordance index (c-index) of 0.67 (95% CI 0.52–0.82) and 0.66 (95% CI 0.57–0.78) for the training cohort (n = 36) and internal validation cohort (n = 36), respectively, were obtained for the PET radiomics model. The PFS advantage of the low-risk group translated also into an OS advantage. On CT images, no radiomics model could be trained. SUV max and SUV mean were also not prognostic in terms of PFS and OS.ConclusionWe were able to build a successful FDG PET radiomics model for the prediction of PFS in MPM. Radiomics could serve as a tool to aid clinical decision support systems for treatment of MPM in future.

PD-2 Role of pretreatment SUVmax on 18F-FDG PET and clinicopathological features in the prognostic stratification of newly diagnosed intrahepatic cholangiocarcinoma

Cholangiocarcinoma (CCA) is the second most common primary liver cancer, following hepatocellular carcinoma, and it accounts for approximately 3% of all gastrointestinal cancers. CCAs are usually grouped in intrahepatic CCA (iCCA), perihilar CCA (pCCA) and distal CCA (dCCA), according to anatomical location. Several recent studies have suggested an increase in the incidence of iCCA in both western and eastern countries, and unfortunately, most of the patients with iCCA present with inoperable or metastatic disease.