Value of multi-center 18F-FDG PET/CT radiomics in predicting EGFR mutation status in lung adenocarcinoma.

BACKGROUND: Accurate, noninvasive, and reliable assessment of epidermal growth factor receptor (EGFR) mutation status and EGFR molecular subtypes is essential for treatment plan selection and individualized therapy in lung adenocarcinoma (LUAD). Radiomics models based on 18F-FDG PET/CT have great potential in identifying EGFR mutation status and EGFR subtypes in patients with LUAD. The validation of multi-center data, model visualization, and interpretation are significantly important for the management, application and trust of machine learning predictive models. However, few EGFR-related research involved model visualization and interpretation, and multi-center trial. PURPOSE: To develop explainable optimal predictive models based on handcrafted radiomics features (HRFs) extracted from multi-center 18F-FDG PET/CT to predict EGFR mutation status and molecular subtypes in LUAD. METHODS: Baseline 18F-FDG PET/CT images of 383 LUAD patients from three hospitals and one public data set were collected. Further, 1808 HRFs were extracted from the primary tumor regions using Pyradiomics. Predictive models were built based on cross-combination of seven feature selection methods and seven machine learning algorithms. Yellowbrick and explainable artificial intelligence technology were used for model visualization and interpretation. Receiver operating characteristic curve, classification report and confusion matrix were used for model performance evaluation. Clinical applicability of the optimal models was assessed by decision curve analysis. RESULTS: STACK feature selection method combined with light gradient boosting machine (LGBM) reached optimal performance in identifying EGFR mutation status ([area under the curve] AUC = 0.81 in the internal test cohort; AUC = 0.62 in the external test cohort). Random forest feature selection method combined with LGBM reached optimal performance in predicting EGFR mutation molecular subtypes (AUC = 0.89 in the internal test cohort; AUC = 0.61 in the external test cohort). CONCLUSIONS: Explainable machine learning models combined with radiomics features extracted from multi-center/scanner 18F-FDG PET/CT have certain potential to identify EGFR mutation status and subtypes in LUAD, which might be helpful to the treatment of LUAD.

Deep radiomics-based fusion model for prediction of bevacizumab treatment response and outcome in patients with colorectal cancer liver metastases: a multicentre cohort study.

Background: Accurate tumour response prediction to targeted therapy allows for personalised conversion therapy for patients with unresectable colorectal cancer liver metastases (CRLM). In this study, we aimed to develop and validate a multi-modal deep learning model to predict the efficacy of bevacizumab in patients with initially unresectable CRLM using baseline PET/CT, clinical data, and colonoscopy biopsy specimens. Methods: In this multicentre cohort study, we retrospectively collected data of 307 patients with CRLM from the BECOME study (NCT01972490) (Zhongshan Hospital of Fudan University, Shanghai) and two independent Chinese cohorts (internal validation cohort from January 1, 2018 to December 31, 2018 at Zhongshan Hospital of Fudan University; external validation cohort from January 1, 2020 to December 31, 2020 at Zhongshan Hospital-Xiamen, Shanghai, and the First Hospital of Wenzhou Medical University, Wenzhou). The main inclusion criteria were that patients with CRLM had pre-treatment PET/CT images as well as colonoscopy specimens. After extracting PET/CT features with deep neural networks (DNN) and selecting related clinical factors using LASSO analysis, a random forest classifier was built as the Deep Radiomics Bevacizumab efficacy predicting model (DERBY). Furthermore, by combining histopathological biomarkers into DERBY, we established DERBY+. The performance of model was evaluated using area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value. Findings: DERBY achieved promising performance in predicting bevacizumab sensitivity with an AUC of 0.77 and 95% confidence interval (CI) [0.67-0.87]. After combining histopathological features, we developed DERBY+, which had more robust accuracy for predicting tumour response in external validation cohort (AUC 0.83 and 95% CI [0.75-0.92], sensitivity 80.4%, specificity 76.8%). DERBY+ also had prognostic value: the responders had longer progression-free survival (median progression-free survival: 9.6 vs 6.3 months, p = 0.002) and overall survival (median overall survival: 27.6 vs 18.5 months, p = 0.010) than non-responders. Interpretation: This multi-modal deep radiomics model, using PET/CT, clinical data and histopathological data, was able to identify patients with bevacizumab-sensitive CRLM, providing a favourable approach for precise patient treatment. To further validate and explore the clinical impact of this work, future prospective studies with larger patient cohorts are warranted. Funding: The National Natural Science Foundation of China; Fujian Provincial Health Commission Project; Xiamen Science and Technology Agency Program; Clinical Research Plan of SHDC; Shanghai Science and Technology Committee Project; Clinical Research Plan of SHDC; Zhejiang Provincial Natural Science Foundation of China; and National Science Foundation of Xiamen.

Anatomical attention-based prediction of postoperative pulmonary venous obstruction via CTA images.

Total anomalous pulmonary venous connection (TAPVC) is a rare congenital heart disease, with which some patients suffer from postoperative pulmonary venous obstruction (PPVO), requiring particular follow-up strategies and treatments. PPVO prediction has important clinical significance, while building a PPVO prediction model is challenging due to limited data and class imbalance distribution. Inspired by the anatomical evidence of PPVO, which is related to the structure of the left atrium (LA) and pulmonary vein (PV), we design an effective multi-task network for PPVO classification. The proposed method incorporates clinical priors and merits of the segmentation-based network into the classification task. The features learned from segmenting LA and PV are concatenated into the PPVO classification branch to constrain the learning of discriminative features. Anatomical-guided attention is applied in the aggregation of these features to restrict them focusing on TAPVC-related regions. To deal with the imbalance classification problem of PPVO, a novel classification loss derived by masked class activation map (MCAM) is designed to improve the classification performance. Computed tomography angiography (CTA) images of 146 patients diagnosed with supracardiac TAPVC in Shanghai Children's Medical Center and Guangdong Provincial People's Hospital were enrolled in this work. The comprehensive experiments demonstrate the effectiveness and generalization of our proposed method. The automatic PPVO prediction model shows the potential application in helping clinicians develop follow-up strategies, thereby improving the survival rate of TAPVC patients.

Progression-Free Survival Prediction in Small Cell Lung Cancer Based on Radiomics Analysis of Contrast-Enhanced CT.

Purposes and Objectives: The aim of this study was to predict the progression-free survival (PFS) in patients with small cell lung cancer (SCLC) by radiomic signature from the contrast-enhanced computed tomography (CT). Methods: A total of 186 cases with pathological confirmed small cell lung cancer were retrospectively assembled. First, 1,218 radiomic features were automatically extracted from tumor region of interests (ROIs) on the lung window and mediastinal window, respectively. Then, the prognostic and robust features were selected by machine learning methods, such as (1) univariate analysis based on a Cox proportional hazard (CPH) model, (2) redundancy removing using the variance inflation factor (VIF), and (3) multivariate importance analysis based on random survival forests (RSF). Finally, PFS predictive models were established based on RSF, and their performances were evaluated using the concordance index (C-index) and the cumulative/dynamic area under the curve (C/D AUC). Results: In total, 11 radiomic features (6 for mediastinal window and 5 for lung window) were finally selected, and the predictive model constructed from them achieved a C-index of 0.7531 and a mean C/D AUC of 0.8487 on the independent test set, better than the predictions by single clinical features (C-index = 0.6026, mean C/D AUC = 0.6312), and single radiomic features computed in lung window (C-index = 0.6951, mean C/D AUC = 0.7836) or mediastinal window (C-index = 0.7192, mean C/D AUC = 0.7964). Conclusion: The radiomic features computed from tumor ROIs on both lung window and mediastinal window can predict the PFS for patients with SCLC by a high accuracy, which could be used as a useful tool to support the personalized clinical decision for the diagnosis and patient management of patients with SCLC.

Building a Risk Prediction Model for Postoperative Pulmonary Vein Obstruction via Quantitative Analysis of CTA Images.

Total anomalous pulmonary venous connection (TAPVC) is a rare but mortal congenital heart disease in children and can be repaired by surgical operations. However, some patients may suffer from pulmonary venous obstruction (PVO) after surgery with insufficient blood supply, necessitating special follow-up strategy and treatment. Therefore, it is a clinically important yet challenging problem to predict such patients before surgery. In this paper, we address this issue and propose a computational framework to determine the risk factors for postoperative PVO (PPVO) from computed tomography angiography (CTA) images and build the PPVO risk prediction model. From clinical experiences, such risk factors are likely from the left atrium (LA) and pulmonary vein (PV) of the patient. Thus, 3D models of LA and PV are first reconstructed from low-dose CTA images. Then, a feature pool is built by computing different morphological features from 3D models of LA and PV, and the coupling spatial features of LA and PV. Finally, four risk factors are identified from the feature pool using the machine learning techniques, followed by a risk prediction model. As a result, not only PPVO patients can be effectively predicted but also qualitative risk factors reported in the literature can now be quantified. Finally, the risk prediction model is evaluated on two independent clinical datasets from two hospitals. The model can achieve the AUC values of 0.88 and 0.87 respectively, demonstrating its effectiveness in risk prediction.

Automatic identification of myopic maculopathy related imaging features in optic disc region via machine learning methods.

BACKGROUND: Myopic maculopathy (MM) is the most serious and irreversible complication of pathologic myopia, which is a major cause of visual impairment and blindness. Clinic proposed limited number of factors related to MM. To explore additional features strongly related with MM from optic disc region, we employ a machine learning based radiomics analysis method, which could explore and quantify more hidden or imperceptible MM-related features to the naked eyes and contribute to a more comprehensive understanding of MM and therefore may assist to distinguish the high-risk population in an early stage. METHODS: A total of 457 eyes (313 patients) were enrolled and were divided into severe MM group and without severe MM group. Radiomics analysis was applied to depict features significantly correlated with severe MM from optic disc region. Receiver Operating Characteristic were used to evaluate these features' performance of classifying severe MM. RESULTS: Eight new MM-related image features were discovered from the optic disc region, which described the shapes, textural patterns and intensity distributions of optic disc region. Compared with clinically reported MM-related features, these newly discovered features exhibited better abilities on severe MM classification. And the mean values of most features were markedly changed between patients with peripapillary diffuse chorioretinal atrophy (PDCA) and macular diffuse chorioretinal atrophy (MDCA). CONCLUSIONS: Machine learning and radiomics method are useful tools for mining more MM-related features from the optic disc region, by which complex or even hidden MM-related features can be discovered and decoded. In this paper, eight new MM-related image features were found, which would be useful for further quantitative study of MM-progression. As a nontrivial byproduct, marked changes between PDCA and MDCA was discovered by both new image features and clinic features.

Predicting EGFR mutation subtypes in lung adenocarcinoma using 18F-FDG PET/CT radiomic features.

BACKGROUND: Identification of epidermal growth factor receptor (EGFR) mutation types is crucial before tyrosine kinase inhibitors (TKIs) treatment. Radiomics is a new strategy to noninvasively predict the genetic status of cancer. In this study, we aimed to develop a predictive model based on 18F-fluorodeoxyglucose positron emission tomography-computed tomography (18F-FDG PET/CT) radiomic features to identify the specific EGFR mutation subtypes. METHODS: We retrospectively studied 18F-FDG PET/CT images of 148 patients with isolated lung lesions, which were scanned in two hospitals with different CT scan setting (slice thickness: 3 and 5 mm, respectively). The tumor regions were manually segmented on PET/CT images, and 1,570 radiomic features (1,470 from CT and 100 from PET) were extracted from the tumor regions. Seven hundred and ninety-four radiomic features insensitive to different CT settings were first selected using the Mann white U test, and collinear features were further removed from them by recursively calculating the variation inflation factor. Then, multiple supervised machine learning models were applied to identify prognostic radiomic features through: (I) a multi-variate random forest to select features of high importance in discriminating different EGFR mutation status; (II) a logistic regression model to select features of the highest predictive value of the EGFR subtypes. The EGFR mutation predicting model was constructed from prognostic radiomic features using the popular Xgboost machine-learning algorithm and validated using 3-fold cross-validation. The performance of predicting model was analyzed using the receiver operating characteristic curve (ROC) and measured with the area under the curve (AUC). RESULTS: Two sets of prognostic radiomic features were found for specific EGFR mutation subtypes: 5 radiomic features for EGFR exon 19 deletions, and 5 radiomic features for EGFR exon 21 L858R missense. The corresponding radiomic predictors achieved the prediction accuracies of 0.77 and 0.92 in terms of AUC, respectively. Combing these two predictors, the overall model for predicting EGFR mutation positivity was also constructed, and the AUC was 0.87. CONCLUSIONS: In our study, we established predictive models based on radiomic analysis of 18F-FDG PET/CT images. And it achieved a satisfying prediction power in the identification of EGFR mutation status as well as the certain EGFR mutation subtypes in lung cancer.

Assessing PD-L1 Expression Level by Radiomic Features From PET/CT in Nonsmall Cell Lung Cancer Patients: An Initial Result.

RATIONALE AND OBJECTIVES: To explore the potential value of radiomic features-derived approach in assessing PD-L1 expression status in nonsmall cell lung cancer (NSCLC) patients. MATERIALS AND METHODS: A cohort of 399 stage I-IV NSCLC patients were enrolled. Tumor segmentation was performed to select essential primary lesions of NSCLC cases after PET/CT images acquisition. Features were extracted, then filtered with automatic relevance determination and minimized with LASSO model based on its relevance of PD-L1 expression status. Finally, we built predictive models with features from the CT, the PET, and the PET/CT images, respectively, for differentiating different status of specific PD-L1 types. Five-fold cross validation was practiced to evaluate the signatures' accuracy, and the receiver operating characteristic as well as the corresponding area under the curve (AUC) was reckoned for each model. RESULTS: With the total of 24 selected features which were significantly associated with PD-L1 expression levels, models based on CT-, PET-, PET/CT-derived features were built and compared. For PD-L1 (SP142) expression level over 1% prediction, models that comprised radiomic features from the CT, the PET, and the PET/CT images resulted in an AUC of 0.97, 0.61, and 0.97, respectively; models for over 50% prediction resulted with AUC of 0.80, 0.65, and 0.77. For PD-L1 (28-8) expression level prediction, predictive models of over 1% expression scored at 0.86, 0.62, and 0.85; and signatures of over 50% expression reached the score of AUCs at 0.91, 0.75, and 0.88, respectively. CONCLUSION: The radiomic-based predictive approach, especially CT-derived predictive model, may anticipate PD-L1 expression status in NSCLC patients relatively accurate. It may be helpful in guiding immunotherapy in clinical practice and deserves further analysis.