Prediction of malignant esophageal fistula in esophageal cancer using a radiomics-clinical nomogram.

BACKGROUND: Malignant esophageal fistula (MEF), which occurs in 5% to 15% of esophageal cancer (EC) patients, has a poor prognosis. Accurate identification of esophageal cancer patients at high risk of MEF is challenging. The goal of this study was to build and validate a model to predict the occurrence of esophageal fistula in EC patients. METHODS: This study retrospectively enrolled 122 esophageal cancer patients treated by chemotherapy or chemoradiotherapy (53 with fistula, 69 without), and all patients were randomly assigned to a training (n = 86) and a validation (n = 36) cohort. Radiomic features were extracted from pre-treatment CTs, clinically predictors were identified by logistic regression analysis. Lasso regression model was used for feature selection, and radiomics signature building. Multivariable logistic regression analysis was used to develop the clinical nomogram, radiomics-clinical nomogram and radiomics prediction model. The models were validated and compared by discrimination, calibration, reclassification, and clinical benefit. RESULTS: The radiomic signature consisting of ten selected features, was significantly associated with esophageal fistula (P = 0.001). Radiomics-clinical nomogram was created by two predictors including radiomics signature and stenosis, which was identified by logistic regression analysis. The model showed good discrimination with an AUC = 0.782 (95% CI 0.684-0.8796) in the training set and 0.867 (95% CI 0.7461-0.987) in the validation set, with an AIC = 101.1, and good calibration. When compared to the clinical prediction model, the radiomics-clinical nomogram improved NRI by 0.236 (95% CI 0.153, 0.614) and IDI by 0.125 (95% CI 0.040, 0.210), P = 0.004. CONCLUSION: We developed and validated the first radiomics-clinical nomogram for malignant esophageal fistula, which could assist clinicians in identifying patients at high risk of MEF.

Research on multi-model imaging machine learning for distinguishing early hepatocellular carcinoma.

OBJECTIVE: To investigate the value of differential diagnosis of hepatocellular carcinoma (HCC) and non-hepatocellular carcinoma (non-HCC) based on CT and MR multiphase radiomics combined with different machine learning models and compare the diagnostic efficacy between different radiomics models. BACKGROUND: Primary liver cancer is one of the most common clinical malignancies, hepatocellular carcinoma (HCC) is the most common subtype of primary liver cancer, accounting for approximately 90% of cases. A clear diagnosis of HCC is important for the individualized treatment of patients with HCC. However, more sophisticated diagnostic modalities need to be explored. METHODS: This retrospective study included 211 patients with liver lesions: 97 HCC and 124 non-hepatocellular carcinoma (non-HCC) who underwent CT and MRI. Imaging data were used to obtain imaging features of lesions and radiomics regions of interest (ROI). The extracted imaging features were combined to construct different radiomics models. The clinical data and imaging features were then combined with radiomics features to construct the combined models. Support Vector Machine (SVM), K-nearest Neighbor (KNN), RandomForest (RF), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Multilayer Perceptron (MLP) six machine learning models were used for training. Five-fold cross-validation was used to train the models, and ROC curves were used to analyze the diagnostic efficacy of each model and calculate the accuracy rate. Model training and efficacy test were performed as before. RESULTS: Statistical analysis showed that some clinical data (gender and concomitant cirrhosis) and imaging features (presence of envelope, marked enhancement in the arterial phase, rapid contouring in the portal phase, uniform density/signal and concomitant steatosis) were statistical differences (P < 0.001). The results of machine learning models showed that KNN had the best diagnostic efficacy. The results of the combined model showed that SVM had the best diagnostic efficacy, indicating that the combined model (accuracy 0.824) had better diagnostic efficacy than the radiomics-only model. CONCLUSIONS: Our results demonstrate that the radiomic features of CT and MRI combined with machine learning models enable differential diagnosis of HCC and non-HCC (malignant, benign). The diagnostic model with dual radiomic had better diagnostic efficacy. The combined model was superior to the radiomic model alone.

A multi-task deep learning model for EGFR genotyping prediction and GTV segmentation of brain metastasis.

BACKGROUND: The precise prediction of epidermal growth factor receptor (EGFR) mutation status and gross tumor volume (GTV) segmentation are crucial goals in computer-aided lung adenocarcinoma brain metastasis diagnosis. However, these two tasks present continuous difficulties due to the nonuniform intensity distributions, ambiguous boundaries, and variable shapes of brain metastasis (BM) in MR images.The existing approaches for tackling these challenges mainly rely on single-task algorithms, which overlook the interdependence between these two tasks. METHODS: To comprehensively address these challenges, we propose a multi-task deep learning model that simultaneously enables GTV segmentation and EGFR subtype classification. Specifically, a multi-scale self-attention encoder that consists of a convolutional self-attention module is designed to extract the shared spatial and global information for a GTV segmentation decoder and an EGFR genotype classifier. Then, a hybrid CNN-Transformer classifier consisting of a convolutional block and a Transformer block is designed to combine the global and local information. Furthermore, the task correlation and heterogeneity issues are solved with a multi-task loss function, aiming to balance the above two tasks by incorporating segmentation and classification loss functions with learnable weights. RESULTS: The experimental results demonstrate that our proposed model achieves excellent performance, surpassing that of single-task learning approaches. Our proposed model achieves a mean Dice score of 0.89 for GTV segmentation and an EGFR genotyping accuracy of 0.88 on an internal testing set, and attains an accuracy of 0.81 in the EGFR genotype prediction task and an average Dice score of 0.85 in the GTV segmentation task on the external testing set. This shows that our proposed method has outstanding performance and generalization. CONCLUSION: With the introduction of an efficient feature extraction module, a hybrid CNN-Transformer classifier, and a multi-task loss function, the proposed multi-task deep learning network significantly enhances the performance achieved in both GTV segmentation and EGFR genotyping tasks. Thus, the model can serve as a noninvasive tool for facilitating clinical treatment.

Comparative analysis of dose calculation algorithms for CyberKnife-based stereotactic radiotherapy in lung cancer.

Purpose: The accuracy of dose calculation is the prerequisite for CyberKnife (CK) to implement precise stereotactic body radiotherapy (SBRT). In this study, CK-MLC treatment planning for early-stage non-small cell lung cancer (NSCLC) were compared using finite-size pencil beam (FSPB) algorithm, FSPB with lateral scaling option (FSPB_LS) and Monte Carlo (MC) algorithms, respectively. We concentrated on the enhancement of accuracy with the FSPB_LS algorithm over the conventional FSPB algorithm and the dose consistency with the MC algorithm. Methods: In this study, 54 cases of NSCLC were subdivided into central lung cancer (CLC, n=26) and ultra-central lung cancer (UCLC, n=28). For each patient, we used the FSPB algorithm to generate a treatment plan. Then the dose was recalculated using FSPB_LS and MC dose algorithms based on the plans computed using the FSPB algorithm. The resultant plans were assessed by calculating the mean value of pertinent comparative parameters, including PTV prescription isodose, conformity index (CI), homogeneity index (HI), and dose-volume statistics of organs at risk (OARs). Results: In this study, most dose parameters of PTV and OARs demonstrated a trend of MC > FSPB_LS > FSPB. The FSPB_LS algorithm aligns better with the dose parameters of the target compared to the MC algorithm, which is particularly evident in UCLC. However, the FSPB algorithm significantly underestimated the does of the target. Regarding the OARs in CLC, differences in dose parameters were observed between FSPB and FSPB_LS for V10 of the contralateral lung, as well as between FSPB and MC for mean dose (Dmean) of the contralateral lung and maximum dose (Dmax) of the aorta, exhibiting statistical differences. There were no statistically significant differences observed between FSPB_LS and MC for the OARs. However, the average dose deviation between FSPB_LS and MC algorithms for OARs ranged from 2.79% to 11.93%. No significant dose differences were observed among the three algorithms in UCLC. Conclusion: For CLC, the FSPB_LS algorithm exhibited good consistency with the MC algorithm in PTV and demonstrated a significant improvement in accuracy when compared to the traditional FSPB algorithm. However, the FSPB_LS algorithm and the MC algorithm showed a significant dose deviation in OARs of CLC. In the case of UCLC, FSPB_LS showed better consistency with the MC algorithm than observed in CLC. Notwithstanding, UCLC's OARs were highly sensitive to radiation dose and could result in potentially serious adverse reactions. Consequently, it is advisable to use the MC algorithm for dose calculation in both CLC and UCLC, while the application of FSPB_LS algorithm should be carefully considered.

Automatic Detection of Brain Metastases in T1-Weighted Construct-Enhanced MRI Using Deep Learning Model.

As a complication of malignant tumors, brain metastasis (BM) seriously threatens patients' survival and quality of life. Accurate detection of BM before determining radiation therapy plans is a paramount task. Due to the small size and heterogeneous number of BMs, their manual diagnosis faces enormous challenges. Thus, MRI-based artificial intelligence-assisted BM diagnosis is significant. Most of the existing deep learning (DL) methods for automatic BM detection try to ensure a good trade-off between precision and recall. However, due to the objective factors of the models, higher recall is often accompanied by higher number of false positive results. In real clinical auxiliary diagnosis, radiation oncologists are required to spend much effort to review these false positive results. In order to reduce false positive results while retaining high accuracy, a modified YOLOv5 algorithm is proposed in this paper. First, in order to focus on the important channels of the feature map, we add a convolutional block attention model to the neck structure. Furthermore, an additional prediction head is introduced for detecting small-size BMs. Finally, to distinguish between cerebral vessels and small-size BMs, a Swin transformer block is embedded into the smallest prediction head. With the introduction of the F2-score index to determine the most appropriate confidence threshold, the proposed method achieves a precision of 0.612 and recall of 0.904. Compared with existing methods, our proposed method shows superior performance with fewer false positive results. It is anticipated that the proposed method could reduce the workload of radiation oncologists in real clinical auxiliary diagnosis.

Deep learning-based combination of [18F]-FDG PET and CT images for producing pulmonary perfusion image.

BACKGROUND: The main application of [18F] FDG-PET (18 FDG-PET) and CT images in oncology is tumor identification and quantification. Combining PET and CT images to mine pulmonary perfusion information for functional lung avoidance radiation therapy (FLART) is desirable but remains challenging. PURPOSE: To develop a deep-learning-based (DL) method to combine 18 FDG-PET and CT images for producing pulmonary perfusion images (PPI). METHODS: Pulmonary technetium-99 m-labeled macroaggregated albumin SPECT (PPISPECT ), 18 FDG-PET, and CT images obtained from 53 patients were enrolled. CT and PPISPECT images were rigidly registered, and registration displacement was subsequently used to align 18 FDG-PET and PPISPECT images. The left/right lung was separated and rigidly registered again to improve the registration accuracy. A DL model based on 3D Unet architecture was constructed to directly combine multi-modality 18 FDG-PET and CT images for producing PPI (PPIDLM ). 3D Unet architecture was used as the basic architecture, and the input was expanded from a single-channel to a dual-channel to combine multi-modality images. For comparative evaluation, 18 FDG-PET images were also used alone to generate PPIDLPET . Sixty-seven samples were randomly selected for training and cross-validation, and 36 were used for testing. The Spearman correlation coefficient (rs ) and multi-scale structural similarity index measure (MS-SSIM) between PPIDLM /PPIDLPET and PPISPECT were computed to assess the statistical and perceptual image similarities. The Dice similarity coefficient (DSC) was calculated to determine the similarity between high-/low- functional lung (HFL/LFL) volumes. RESULTS: The voxel-wise rs and MS-SSIM of PPIDLM /PPIDLPET were 0.78 ± 0.04/0.57 ± 0.03, 0.93 ± 0.01/0.89 ± 0.01 for cross-validation and 0.78 ± 0.11/0.55 ± 0.18, 0.93 ± 0.03/0.90 ± 0.04 for testing. PPIDLM /PPIDLPET achieved averaged DSC values of 0.78 ± 0.03/0.64 ± 0.02 for HFL and 0.83 ± 0.01/0.72 ± 0.03 for LFL in the training dataset and 0.77 ± 0.11/0.64 ± 0.12, 0.82 ± 0.05/0.72 ± 0.06 in the testing dataset. PPIDLM yielded a stronger correlation and higher MS-SSIM with PPISPECT than PPIDLPET (p < 0.001). CONCLUSIONS: The DL-based method integrates lung metabolic and anatomy information for producing PPI and significantly improved the accuracy over methods based on metabolic information alone. The generated PPIDLM can be applied for pulmonary perfusion volume segmentation, which is potentially beneficial for FLART treatment plan optimization.

Volumetric modulated arc therapy for hippocampal-sparing prophylactic cranial irradiation: Planning comparison of Halcyon and C-arm accelerators.

Background: The purpose of the study was to evaluate the dosimetry of the Halcyon in prophylactic cranial irradiation (PCI) with volumetric modulated arc therapy (VMAT) and hippocampal-sparing for small cell lung cancer (SCLC). Methods: Five VMAT plans were designed on CT images of 15 patients diagnosed with SCLC and received PCI. Three plans with two full arcs were generated on the Trilogy and the TrueBeam accelerators, and flattening filter (FF) and flattening filter free (FFF) modes were used on TrueBeam. Two Halcyon plans with two and three full arcs were generated, referred to as H-2A and H-3A, respectively. The prescription dose was 25 Gy in 2.5-Gy fractions. The dose limit for hippocampus were D100 ≤ 9Gy and Dmax ≤ 16Gy. The Wilcoxon matched-paired signed-rank test was used to evaluate the significance of the observed differences between the five plans. Results: H-2A plans significantly increased the D2 of PTV, and H-3A plans showed comparable or even better target dosimetry (better conformity) compared to the three plans on C-arm accelerators. Compared to T and TB plans, the two Halcyon plans significantly reduced the D100 and mean doses of bilateral hippocampus, the mean doses of eyeballs, and the maximum doses of lenses. D100 of hippocampus was reduced in TrueBeam plans comparing to Trilogy plans. The FFF plans on TrueBeam also represented advantages in Dmean and D100 of hippocampas, Dmean and Dmax of eyeballs, and the Dmax of lenses compared to FF plans. Halcyon plans and TrueBeam plans with FFF mode increased the MUs compared to FF plans. Comparing to H-2A, the H-3A plans exhibited additional dosimetric advantages, including D2, CI and HI of PTV, as well as the maximum and mean doses of hippocampus and eyeballs, and the maximum doses of optic nerves and brainstem. The two Halcyon plans significantly reduced the delivery time and showed the higher gamma passing rate than the three plans of C-arm accelerators. Conclusions: Compared with the C-arm accelerators, the dose of hippocampus and the delivery times on Halcyon are relatively significantly reduced for hippocampal-sparing PCI. Three arcs are recommended for VMAT plans with the Halcyon in hippocampal-sparing PCI.

Prediction of distant metastasis in esophageal cancer using a radiomics-clinical model.

BACKGROUND: Distant metastasis, which occurs at a rate of 25% in patients with esophageal cancer (EC), has a poor prognosis, with previous studies reporting an overall survival of only 3-10 months. However, few studies have been conducted to predict distant metastasis in EC, owing to a dearth of reliable biomarkers. The purpose of this study was to develop and validate an accurate model for predicting distant metastasis in patients with EC. METHODS: A total of 299 EC patients were enrolled and randomly assigned to a training cohort (n = 207) and a validation cohort (n = 92). Logistic univariate and multivariate regression analyses were used to identify clinical independent predictors and create a clinical nomogram. Radiomic features were extracted from contrast-enhanced computed tomography (CT) images taken prior to treatment, and least absolute shrinkage and selection operator (Lasso) regression was used to screen the associated features, which were then used to develop a radiomic signature. Based on the screened features, four machine learning algorithms were used to build radiomics models. The joint nomogram with radiomic signature and clinically independent risk factors was developed using the logical regression algorithm. All models were validated and compared by discrimination, calibration, reclassification, and clinical benefit. RESULTS: Multivariable analyses revealed that age, N stage, and degree of pathological differentiation were independent predictors of distant metastasis, and a clinical nomogram incorporating these factors was established. A radiomic signature was developed by a set of sixteen features chosen from 851 radiomic features. The joint nomogram incorporating clinical factors and radiomic signature performed better [AUC(95% CI) 0.827(0.742-0.912)] than the clinical nomogram [AUC(95% CI) 0.731(0.626-0.836)] and radiomics predictive models [AUC(95% CI) 0.754(0.652-0.855), LR algorithms]. Calibration and decision curve analyses revealed that the radiomics-clinical nomogram outperformed the other models. In comparison with the clinical nomogram, the joint nomogram's NRI was 0.114 (95% CI 0.075-0.345), and its IDI was 0.071 (95% CI 0.030-0.112), P = 0.001. CONCLUSIONS: We developed and validated the first radiomics-clinical nomogram for distant metastasis in EC which may aid clinicians in identifying patients at high risk of distant metastasis.

Development and verification of radiomics framework for computed tomography image segmentation.

BACKGROUND: Radiomics has been considered an imaging marker for capturing quantitative image information (QII). The introduction of radiomics to image segmentation is desirable but challenging. PURPOSE: This study aims to develop and validate a radiomics-based framework for image segmentation (RFIS). METHODS: RFIS is designed using features extracted from volume (svfeatures) created by sliding window (swvolume). The 53 svfeatures are extracted from 11 phantom series. Outliers in the svfeature datasets are detected by isolation forest (iForest) and specified as the mean value. The percentage coefficient of variation (%COV) is calculated to evaluate the reproducibility of svfeatures. RFIS is constructed and applied to the gross target volume (GTV) segmentation from the peritumoral region (GTV with a 10 mm margin) to assess its feasibility. The 127 lung cancer images are enrolled. The test-retest method, correlation matrix, and Mann-Whitney U test (p < 0.05) are used to select non-redundant svfeatures of statistical significance from the reproducible svfeatures. The synthetic minority over-sampling technique is utilized to balance the minority group in the training sets. The support vector machine is employed for RFIS construction, which is tuned in the training set using 10-fold stratified cross-validation and then evaluated in the test sets. The swvolumes with the consistent classification results are grouped and merged. Mode filtering is performed to remove very small subvolumes and create relatively large regions of completely uniform character. In addition, RFIS performance is evaluated by the area under the receiver operating characteristic (ROC) curve (AUC), accuracy, sensitivity, specificity, and Dice similarity coefficient (DSC). RESULTS: 30249 phantom and 145008 patient image swvolumes were analyzed. Forty-nine (92.45% of 53) svfeatures represented excellent reproducibility(%COV<15). Forty-five features (91.84% of 49) included five categories that passed test-retest analysis. Thirteen svfeatures (28.89% of 45) svfeatures were selected for RFIS construction. RFIS showed an average (95% confidence interval) sensitivity of 0.848 (95% CI:0.844-0.883), a specificity of 0.821 (95% CI: 0.818-0.825), an accuracy of 83.48% (95% CI: 83.27%-83.70%), and an AUC of 0.906 (95% CI: 0.904-0.908) with cross-validation. The sensitivity, specificity, accuracy, and AUC were equal to 0.762 (95% CI: 0.754-0.770), 0.840 (95% CI: 0.837-0.844), 82.29% (95% CI: 81.90%-82.60%), and 0.877 (95% CI: 0.873-0.881) in the test set, respectively. GTV was segmented by grouping and merging swvolume with identical classification results. The mean DSC after mode filtering was 0.707 ± 0.093 in the training sets and 0.688 ± 0.072 in the test sets. CONCLUSION: Reproducible svfeatures can capture the differences in QII among swvolumes. RFIS can be applied to swvolume classification, which achieves image segmentation by grouping and merging the swvolume with similar QII.

Demonstration of quasi-real-time intrafractional motion in esophageal cancer radiotherapy provided by ExacTrac X-ray snap verification.

OBJECTIVE: To investigate the following hypotheses: (1) ExacTrac X-ray Snap Verification (ET-SV) is an alternative to CBCT for positioning patients with esophageal carcinoma (EC), (2) ET-SV can detect displacement in EC patients during radiotherapy (RT) and (3) EC patients can be feasibly monitored in quasi-real-time with ET-SV during RT. METHODS: Anthropomorphic phantoms and 13 patients were included in this study. CBCT and ET-SV were both implemented before treatment delivery to detect displacement, and their correction results were compared. For the patient tests, positional correction in 3 translational directions and the yaw direction were applied using the ET-SV correction results. The residual error was detected immediately using ET-SV. Finally, to acquire the intrafractional motion, ET-SV was implemented when the gantry was at 0°, 90°, 180° and 270°, respectively. RESULTS: In phantom tests, the maximum value of the difference in displacement between the CBCT and ET systems was 1.16 mm for translation and 0.31° for yaw. According to Bland-Altman analysis of the patient test results, 5% (5/98), 5% (5/98), 5% (5/98), and 4% (4/98) of points were beyond the upper and lower limits of agreement in the AP, SI, LR and yaw directions, respectively. The mean residual error was -0.482 mm, 1.215 mm, 1.0 mm, -0.487°, 0.105°, and 0.003° in the AP, SI, LR, pitch, roll and yaw directions, respectively. The intrafractional displacement ranged from -0.21 mm to 0 mm for translation and from -0.63° to 0.21° for rotation. The mean total translational error for intrafractional motion increased from 0.47 mm to 1.14 mm during the treatment. CONCLUSION: The accuracy of ET-SV for EC RT positional correction is comparable to that of CBCT. Thus, Quasi-real-time intrafractional monitoring can be used to detect EC patient displacement during radiotherapy.

Reproducibility for Hepatocellular Carcinoma CT Radiomic Features: Influence of Delineation Variability Based on 3D-CT, 4D-CT and Multiple-Parameter MR Images.

Purpose: Accurate lesion segmentation is a prerequisite for radiomic feature extraction. It helps to reduce the features variability so as to improve the reporting quality of radiomics study. In this research, we aimed to conduct a radiomic feature reproducibility test of inter-/intra-observer delineation variability in hepatocellular carcinoma using 3D-CT images, 4D-CT images and multiple-parameter MR images. Materials and Methods: For this retrospective study, 19 HCC patients undergoing 3D-CT, 4D-CT and multiple-parameter MR scans were included in this study. The gross tumor volume (GTV) was independently delineated twice by two observers based on contrast-enhanced computed tomography (CECT), maximum intensity projection (MIP), LAVA-Flex, T2W FRFSE and DWI-EPI images. We also delineated the peritumoral region, which was defined as 0 to 5 mm radius surrounding the GTV. 107 radiomic features were automatically extracted from CECT images using 3D-Slicer software. Quartile coefficient of dispersion (QCD) and intraclass correlation coefficient (ICC) were applied to assess the variability of each radiomic feature. QCD<10% and ICC≥0.75 were considered small variations and excellent reliability. Finally, the principal component analysis (PCA) was used to test the feasibility of dimensionality reduction. Results: For tumor tissues, the numbers of radiomic features with QCD<10% indicated no obvious inter-/intra-observer differences or discrepancies in 3D-CT, 4D-CT and multiple-parameter MR delineation. However, the number of radiomic features (mean 89) with ICC≥0.75 was the highest in the multiple-parameter MR group, followed by the 3DCT group (mean 77) and the MIP group (mean 73). The peritumor tissues also showed similar results. A total of 15 and 7 radiomic features presented excellent reproducibility and small variation in tumor and peritumoral tissues, respectively. Two robust features showed excellent reproducibility and small variation in tumor and peritumoral tissues. In addition, the values of the two features both represented statistically significant differences among tumor and peritumoral tissues (P<0.05). The PCA results indicated that the first seven principal components could preserve at least 90% of the variance of the original set of features. Conclusion: Delineation on multiple-parameter MR images could help to improve the reproducibility of the HCC CT radiomic features and weaken the inter-/intra-observer influence.

Development and Validation of a Radiomics Nomogram Using Computed Tomography for Differentiating Immune Checkpoint Inhibitor-Related Pneumonitis From Radiation Pneumonitis for Patients With Non-Small Cell Lung Cancer.

Background: The combination of immunotherapy and chemoradiotherapy has become the standard therapeutic strategy for patients with unresected locally advance-stage non-small cell lung cancer (NSCLC) and induced treatment-related adverse effects, particularly immune checkpoint inhibitor-related pneumonitis (CIP) and radiation pneumonitis (RP). The aim of this study is to differentiate between CIP and RP by pretreatment CT radiomics and clinical or radiological parameters. Methods: A total of 126 advance-stage NSCLC patients with pneumonitis were enrolled in this retrospective study and divided into the training dataset (n =88) and the validation dataset (n = 38). A total of 837 radiomics features were extracted from regions of interest based on the lung parenchyma window of CT images. A radiomics signature was constructed on the basis of the predictive features by the least absolute shrinkage and selection operator. A logistic regression was applied to develop a radiomics nomogram. Receiver operating characteristics curve and area under the curve (AUC) were applied to evaluate the performance of pneumonitis etiology identification. Results: There was no significant difference between the training and the validation datasets for any clinicopathological parameters in this study. The radiomics signature, named Rad-score, consisting of 11 selected radiomics features, has potential ability to differentiate between CIP and RP with the empirical and α-binormal-based AUCs of 0.891 and 0.896. These results were verified in the validation dataset with AUC = 0.901 and 0.874, respectively. The clinical and radiological parameters of bilateral changes (p < 0.001) and sharp border (p = 0.001) were associated with the identification of CIP and RP. The nomogram model showed good performance on discrimination in the training dataset (AUC = 0.953 and 0.950) and in the validation dataset (AUC = 0.947 and 0.936). Conclusions: CT-based radiomics features have potential values for differentiating between patients with CIP and patients with RP. The addition of bilateral changes and sharp border produced superior model performance on classifying, which could be a useful method to improve related clinical decision-making.

Radiomics Nomogram Based on Multiple-Sequence Magnetic Resonance Imaging Predicts Long-Term Survival in Patients Diagnosed With Nasopharyngeal Carcinoma.

Purpose: Although the tumor-node-metastasis staging system is widely used for survival analysis of nasopharyngeal carcinoma (NPC), tumor heterogeneity limits its utility. In this study, we aimed to develop and validate a radiomics model, based on multiple-sequence magnetic resonance imaging (MRI), to estimate the probability of overall survival in patients diagnosed with NPC. Methods: Multiple-sequence MRIs, including T1-weighted, T1 contrast, and T2-weighted imaging, were collected from patients diagnosed with NPC. Radiomics features were extracted from the contoured gross tumor volume of three sequences from each patient using the least absolute shrinkage and selection operator with the Cox regression model. The optimal Rad score was determined using 12 of the 851 radiomics features derived from the multiple-sequence MRI and its discrimination power was compared in the training and validation cohorts. For better prediction performance, an optimal nomogram (radiomics nomogram-MS) that incorporated the optimal Rad score and clinical risk factors was developed, and a calibration curve and a decision curve were used to further evaluate the optimized discrimination power. Results: A total of 504 patients diagnosed with NPC were included in this study. The optimal Rad score was significantly correlated with overall survival in both the training [C-index: 0.731, 95% confidence interval (CI): 0.709-0.753] and validation cohorts (C-index: 0.807, 95% CI: 0.782-0.832). Compared with the nomogram developed with only single-sequence MRI, the radiomics nomogram-MS had a higher discrimination power in both the training (C-index: 0.827, 95% CI: 0.809-0.845) and validation cohorts (C-index: 0.836, 95% CI: 0.815-0.857). Analysis of the calibration and decision curves confirmed the effectiveness and utility of the optimal radiomics nomogram-MS. Conclusions: The radiomics nomogram model that incorporates multiple-sequence MRI and clinical factors may be a useful tool for the early assessment of the long-term prognosis of patients diagnosed with NPC.

Developing and validating a deep learning and radiomic model for glioma grading using multiplanar reconstructed magnetic resonance contrast-enhanced T1-weighted imaging: a robust, multi-institutional study.

BACKGROUND: Although surgical pathology or biopsy are considered the gold standard for glioma grading, these procedures have limitations. This study set out to evaluate and validate the predictive performance of a deep learning radiomics model based on contrast-enhanced T1-weighted multiplanar reconstruction images for grading gliomas. METHODS: Patients from three institutions who diagnosed with gliomas by surgical specimen and multiplanar reconstructed (MPR) images were enrolled in this study. The training cohort included 101 patients from institution 1, including 43 high-grade glioma (HGG) patients and 58 low-grade glioma (LGG) patients, while the test cohorts consisted of 50 patients from institutions 2 and 3 (25 HGG patients, 25 LGG patients). We then extracted radiomics features and deep learning features using six pretrained models from the MPR images. The Spearman correlation test and the recursive elimination feature selection method were used to reduce the redundancy and select most predictive features. Subsequently, three classifiers were used to construct classification models. The performance of the grading models was evaluated using the area under the receiver operating curve, sensitivity, specificity, accuracy, precision, and negative predictive value. Finally, the prediction performances of the test cohort were compared to determine the optimal classification model. RESULTS: For the training cohort, 62% (13 out of 21) of the classification models constructed with MPR images from multiple planes outperformed those constructed with single-plane MPR images, and 61% (11 out of 18) of classification models constructed with both radiomics features and deep learning features had higher area under the curve (AUC) values than those constructed with only radiomics or deep learning features. The optimal model was a random forest model that combined radiomic features and VGG16 deep learning features derived from MPR images, which achieved AUC of 0.847 in the training cohort and 0.898 in the test cohort. In the test cohort, the sensitivity, specificity, and accuracy of the optimal model were 0.840, 0.760, and 0.800, respectively. CONCLUSIONS: Multiplanar CE-T1W MPR imaging features are more effective than features from single planes when differentiating HGG and LGG. The combination of deep learning features and radiomics features can effectively grade glioma and assist clinical decision-making.

Radiomics Nomogram for Predicting Locoregional Failure in Locally Advanced Non-small Cell Lung Cancer Treated with Definitive Chemoradiotherapy.

RATIONALE AND OBJECTIVES: To develop and validate a computed tomography (CT)-based radiomics nomogram for predicting locoregional failure (LRF) in patients with locally advanced non-small cell lung cancer (NSCLC) treated with definitive chemoradiotherapy (CRT). MATERIALS AND METHODS: A total of 141 patients with locally advanced NSCLC treated with definitive CRT from January 2014 to December 2017 were included and divided into testing cohort (n = 100) and validation (n = 41) cohort. Radiomics features were extracted from pretreatment contrast enhanced CT. The least absolute shrinkage and selection operator logistic regression was processed to select predictive features from the testing cohort and constructed a radiomics signature. Clinical characteristics and the radiomics signature were analyzed using univariable and multivariate Cox regression. The radiomics nomogram was established with the radiomics signature and independent clinical factors. Harrell's C-index, calibration curves and decision curves were used to assess the performance of the radiomics nomogram. RESULTS: The radiomics signature, which consisted of eight selected features, was an independent factor of LRF. The clinical predictors of LRF were the histologic type and clinical stage. The radiomics nomogram combined with the radiomics signature and clinical prognostic factors showed good performance with C-indexes of 0.796 (95% confidence interval [CI]: 0.709-0.883) and 0.756 (95% CI: 0.674-0.838) in the testing and validation cohorts respectively. Additionally, the combined nomogram resulted in better performance (p < 0.001) for the estimation of LRF than the nomograms with the radiomics signature (C-index: 0.776; 95% CI: 0.686-0.866) or clinical predictors (C-index: 0.641; 95% CI: 0.542-0.740) alone. CONCLUSION: The radiomics nomogram provided the best performance for LRF prediction in patients with locally advanced NSCLC, which may help optimize individual treatments.

Development and validation of a prognostic nomogram for malignant esophageal fistula based on radiomics and clinical factors.

BACKGROUND: The current study aimed to comprehensively analyze the clinical prognostic factors of malignant esophageal fistula (MEF). Furthermore, this study sought to establish and validate prognostic nomograms incorporating radiomics and clinical factors to predict overall survival and median survival after fistula for patients with MEF. METHODS: The records of 76 patients with MEF were retrospectively analyzed. A stepwise Cox proportional hazards regression model was employed to screen independent prognostic factors and develop clinical nomograms. Radiomic features were extracted from prefistula CT images and post fistula CT images. Least absolute shrinkage and selection operator (LASSO) regression and Cox regression algorithm was used to filter radiomic features and avoid overfitting. Radiomic signature was a linear combination of optimal features and corresponding coefficients. The joint prognostic nomograms was constructed by radiomic signatures and clinical features. All models were validated by Harrell's concordance index (C-index), caliberation and bootstrap validation. RESULTS: For overall survival, age, prealbumin, KPS and interval between diagnosis of esophageal cancer and fistula were identified as independent prognostic factors and incorporated into the clinical nomogram. Age, prealbumin, serum albumin, KPS and neutrophil proportion were selected for the clinical nomogram of post fistula survival. The C-index of overall survival nomogram was 0.719 (95% CI: 0.645-0.793) and that was 0.722 (95% CI: 0.653-0.791) in the post fistula survival nomogram. The radiomic signature developed by radiomic features of prefistula CT showed a significant correlation with both overall survival and post fistula survival. The C-index of joint nomogarm for overall survival and post fistula survival was 0.831 (95% CI: 0.757-0.905) and 0.77 (95% CI: 0.686-0.854), respectively. The calibration curve showed the joint nomograms outperformed the clinical ones. CONCLUSIONS: The study presents nomograms incorporating independent clinical risk factors and radiomic signature to predict the prognosis of MEF. This prognostic classification system has the potential to guide therapeutic decisions for patients with malignant esophageal fistulas.

Bone marrow dosimetric analysis of lymphopenia in patients with esophageal squamous cell carcinoma treated with chemoradiotherapy.

BACKGROUND: We analyzed the relationship among peripheral blood lymphocytes, exposed sternum and vertebra body bone marrow (BM), and overall survival (OS) to find BM dosimetric parameters of lymphopenia during chemoradiotherapy (CRT) for patients with esophageal squamous cell carcinoma (ESCC). METHODS: We examined 476 ESCC patients from January 2012 to January 2015, all of whom received concurrent or sequential CRT. Absolute lymphocyte counts (ALC) during radiotherapy (RT) of each patient were collected from the routine workup at the following RT times: pretreatment ALC (ALC0), at 1-5, 6-10, 11-15, 16-20, and 21-25, and more than 26 sessions (called ALC1-6, respectively). The sternum and vertebral body BM were delineated in accordance with uniform standards, and the irradiated volumes were calculated by dose-volume histograms (DVH). The Kaplan-Meier method and Cox proportional hazards regression were used to analyze the survival of the patients. Comparisons of DVH were performed using the Mann-Whitney U test or two-sample t-test where appropriate. RESULTS: A relative volume of sternum BM irradiated by more than 20 Gy could clearly affect the peripheral blood lymphocytes. The V20 of sternum BM and V50 of vertebra body BM were related to the OS of the patients, and the level of ALC2 (at 6-10 times of RT) could predict the outcomes of patients. The Cox regression analyses showed that the 218 patients with ALC2 ≥ 0.8 × 109 /L had a significantly higher OS (47.0 months vs. 30.9 months, p < 0.0001) than the 258 patients with ALC2 < 0.8×109 /L. CONCLUSION: In patients with ESCC, the relative volume of sternum BM irradiated by more than 20 Gy was associated with lymphocytes. Patients with ALC2 ≥ 0.8 × 109 /L had a significantly higher OS. The V20 of the sternum BM, the V50 of the vertebra body BM, and the level of ALC2 were significant prognostic factors in patients with ESCC.

Treatment planning comparison of volumetric modulated arc therapy with the trilogy and the Halcyon for bilateral breast cancer.

BACKGROUND: The Halcyon is a new machine from the Varian company. The purpose of this study was to evaluate the dosimetry of the Halcyon in treatment of bilateral breast cancer with volumetric modulated arc therapy. METHODS: On CT images of 10 patients with bilateral breast cancer, four Halcyon plans with different setup fields were generated, and dosimetric comparisons using Bonferroni's multiple comparisons test were conducted among the four plans. Whole and partial arc plans on the Trilogy and the Halcyon, referred to as T-4arc, T-8arc, H-4arc and H-8arc, were designed. The prescription dose was 50 Gy in 2-Gy fractions. All plans were designed with the Eclipse version 15.5 treatment planning system. The dosimetric differences between whole and partial arc plans in the same accelerator were compared using the Mann-Whitney U test. The better Halcyon plan was selected for the further dosimetric comparison of the plan quality and delivery efficiency between the Trilogy and the Halcyon. RESULTS: Halcyon plans with high-quality megavoltage cone beam CT setup fields increased the Dmean, D2 and V107 of the planning target volume (PTV) and the V5 and Dmean of the heart, left ventricle (LV) and lungs compared with other Halcyon setup plans. The mean dose and low dose volume of the heart, lungs and liver were significantly decreased in T-8arc plans compared to T-4arc plans. In terms of the V5, V20, V30, V40 and Dmean of the heart, the V20, V30, V40 and Dmean of the LV, the V30, V40, Dmax and Dmean of the left anterior descending artery (LAD), and the V5 and V40 of lungs, H-8arc was significantly higher than H-4arc (p < 0.05). Compared with the Trilogy's plans, the Halcyon's plans reduced the high-dose volume of the heart and LV but increased the mean dose of the heart. For the dose of the LAD and the V20 and V30 of lungs, there was no significant difference between the two accelerators. Compared with the Trilogy, plans on the Halcyon significantly increased the skin dose but also significantly reduced the delivery time. CONCLUSION: For the Halcyon, the whole-arc plans have more dosimetric advantages than partial-arc plans in bilateral breast cancer radiotherapy. Although the mean dose of the heart and the skin dose are increased, the doses of the cardiac substructure and other OARs are comparable to the Trilogy, and the delivery time is significantly reduced.

Automatic segmentation of hippocampus in hippocampal sparing whole brain radiotherapy: A multitask edge-aware learning.

PURPOSE: This study aimed to improve the accuracy of the hippocampus segmentation through multitask edge-aware learning. METHOD: We developed a multitask framework for computerized hippocampus segmentation. We used three-dimensional (3D) U-net as our backbone model with two training objectives: (a) to minimize the difference between the targeted binary mask and the model prediction; and (b) to optimize an auxiliary edge-prediction task which is designed to guide the model detection of the weak boundary of the hippocampus in model optimization. To balance the multiple task objectives, we proposed an improved gradient normalization by adaptively adjusting the weight of losses from different tasks. A total of 247 T1-weighted MRIs including 131 without contrast and 116 with contrast were collected from 247 patients to train and validate the proposed method. Segmentation was quantitatively evaluated with the dice coefficient (Dice), Hausdorff distance (HD), and average Hausdorff distance (AVD). The 3D U-net was used for baseline comparison. We used a Wilcoxon signed-rank test to compare repeated measurements (Dice, HD, and AVD) by different segmentations. RESULTS: Through fivefold cross-validation, our multitask edge-aware learning achieved Dice of 0.8483 ± 0.0036, HD of 7.5706 ± 1.2330 mm, and AVD of 0.1522 ± 0.0165 mm, respectively. Conversely, the baseline results were 0.8340 ± 0.0072, 10.4631 ± 2.3736 mm, and 0.1884 ± 0.0286 mm, respectively. With a Wilcoxon signed-rank test, we found that the differences between our method and the baseline were statistically significant (P < 0.05). CONCLUSION: Our results demonstrated the efficiency of multitask edge-aware learning in hippocampus segmentation for hippocampal sparing whole-brain radiotherapy. The proposed framework may also be useful for other low-contrast small organ segmentations on medical imaging modalities.