Preoperative Risk Stratification for Gastric Cancer: The Establishment of Dual-Energy CT-Based Radiomics Using Prospective Datasets at Two Centers.

RATIONALE AND OBJECTIVES: To evaluate the performance of dual-energy CT (DECT)-based radiomics models for identifying high-risk histopathologic phenotypes-serosal invasion (pT4a), lymph node metastasis (LNM), lymphovascular invasion (LVI) and perineural invasion (PNI) in gastric cancer. MATERIAL AND METHODS: This prospective bi-center study recruited histologically confirmed gastric adenocarcinoma patients who underwent triple-phase enhanced DECT before gastrectomy between January 2021 and July 2023. Radiomics features were extracted from polychromatic/monochromatic (40 keV, 100 keV)/iodine images at arterial/venous/delay phase, respectively. Predictive features were selected in the training dataset using logistic regression classifier, and trained models were applied to the external validation dataset. Performances of clinical models, conventional contrast enhanced CT (CECT) models and DECT models were evaluated using areas under the receiver operating characteristic curve (AUCs). RESULTS: In total, 503 patients were recruited: 396 at training dataset (60.1 ± 10.8 years, 110 females, 286 males) and 107 at validation dataset (61.4 ± 9.5 years, 29 females, 78 males). DECT models dichotomizing pT4a, LNM, LVI, and PNI achieved AUCs of 0.891, 0.817, 0.834, and 0.889, respectively, in the validation dataset, similar with the CECT models. In the training dataset, compared to the CECT model, the DECT model provided increased performance for identifying pT4a, LNM, LVI (all P＜0.05), and similar performance for stratifying PNI (P = 0.104). The DECT models was associated with patient disease-free survival (all P＜0.05). CONCLUSION: DECT radiomics can stratify patients preoperatively according to high-risk histopathologic phenotypes for gastric cancer and are associated with patient disease-free survival in the training dataset.

Predicting tumor deposits in rectal cancer: a combined deep learning model using T2-MR imaging and clinical features.

BACKGROUND: Tumor deposits (TDs) are associated with poor prognosis in rectal cancer (RC). This study aims to develop and validate a deep learning (DL) model incorporating T2-MR image and clinical factors for the preoperative prediction of TDs in RC patients. METHODS AND METHODS: A total of 327 RC patients with pathologically confirmed TDs status from January 2016 to December 2019 were retrospectively recruited, and the T2-MR images and clinical variables were collected. Patients were randomly split into a development dataset (n = 246) and an independent testing dataset (n = 81). A single-channel DL model, a multi-channel DL model, a hybrid DL model, and a clinical model were constructed. The performance of these predictive models was assessed by using receiver operating characteristics (ROC) analysis and decision curve analysis (DCA). RESULTS: The areas under the curves (AUCs) of the clinical, single-DL, multi-DL, and hybrid-DL models were 0.734 (95% CI, 0.674-0.788), 0.710 (95% CI, 0.649-0.766), 0.767 (95% CI, 0.710-0.819), and 0.857 (95% CI, 0.807-0.898) in the development dataset. The AUC of the hybrid-DL model was significantly higher than the single-DL and multi-DL models (both p < 0.001) in the development dataset, and the single-DL model (p = 0.028) in the testing dataset. Decision curve analysis demonstrated the hybrid-DL model had higher net benefit than other models across the majority range of threshold probabilities. CONCLUSIONS: The proposed hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer. CRITICAL RELEVANCE STATEMENT: The proposed hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer. KEY POINTS: • Preoperative non-invasive identification of TDs is of great clinical significance. • The combined hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer. • A preoperative nomogram provides gastroenterologist with an accurate and effective tool.

Correlation Between Intranodular Vessels and Tumor Invasiveness of Lung Adenocarcinoma Presenting as Ground-glass Nodules: A Deep Learning 3-Dimensional Reconstruction Algorithm-based Quantitative Analysis on Noncontrast Computed Tomography Images.

PURPOSE: To evaluate the role of quantitative features of intranodular vessels based on deep learning in distinguishing pulmonary adenocarcinoma invasiveness. MATERIALS AND METHODS: This retrospective study included 512 confirmed ground-glass nodules from 474 patients with 241 precursor glandular lesions (PGL), 126 minimally invasive adenocarcinomas (MIA), and 145 invasive adenocarcinomas (IAC). The pulmonary blood vessels were reconstructed on noncontrast computed tomography images using deep learning-based region-segmentation and region-growing techniques. The presence of intranodular vessels was evaluated based on the automatic calculation of vessel prevalence, vascular categories, and vessel volume percentage. Further comparisons were made between different invasive groups by the Mantel-Haenszel χ 2 test, χ 2 test, and analysis of variance. RESULTS: The detection rate of intranodular vessels in PGL (33.2%) was significantly lower than that of MIA (46.8%, P = 0.011) and IAC (55.2%, P < 0.001), while the vascular categories were similar (all P > 0.05). Vascular changes were more common in IAC and MIA than in PGL, mainly in increased vessel volume percentage (12.4 ± 19.0% vs. 6.3 ± 13.1% vs. 3.9 ± 9.4%, P < 0.001). The average intranodular artery and vein volume percentage of IAC (7.5 ± 14.0% and 5.0 ± 10.1%) was higher than that of PGL (2.1 ± 6.9% and 1.7 ± 5.8%) and MIA (3.2 ± 9.1% and 3.1 ± 8.7%), with statistical significance (all P < 0.05). CONCLUSIONS: The quantitative analysis of intranodular vessels on noncontrast computed tomography images demonstrated that the ground-glass nodules with increased internal vessel prevalence and volume percentages had higher possibility of tumor invasiveness.

Deep Learning Models for Severity Prediction of Acute Pancreatitis in the Early Phase From Abdominal Nonenhanced Computed Tomography Images.

OBJECTIVES: To develop and validate deep learning (DL) models for predicting the severity of acute pancreatitis (AP) by using abdominal nonenhanced computed tomography (CT) images. METHODS: The study included 978 AP patients admitted within 72 hours after onset and performed abdominal CT on admission. The image DL model was built by the convolutional neural networks. The combined model was developed by integrating CT images and clinical markers. The performance of the models was evaluated by using the area under the receiver operating characteristic curve. RESULTS: The clinical, Image DL, and the combined DL models were developed in 783 AP patients and validated in 195 AP patients. The combined models possessed the predictive accuracy of 90.0%, 32.4%, and 74.2% for mild, moderately severe, and severe AP. The combined DL model outperformed clinical and image DL models with 0.820 (95% confidence interval, 0.759-0.871), the sensitivity of 84.76% and the specificity of 66.67% for predicting mild AP and the area under the receiver operating characteristic curve of 0.920 (95% confidence interval, 0.873-0.954), the sensitivity of 90.32%, and the specificity of 82.93% for predicting severe AP. CONCLUSIONS: The DL technology allows nonenhanced CT images as a novel tool for predicting the severity of AP.

Energy spectrum CT index-based machine learning model predicts the effect of intravenous thrombolysis in lower limbs.

To develop a noninvasive machine learning (ML) model based on energy spectrum computed tomography venography (CTV) indices for preoperatively predicting the effect of intravenous thrombolytic treatment in lower limbs. A total of 3492 slices containing thrombus regions from 58 veins in lower limbs in a cohort of 18 patients, divided in good and poor thrombolysis prognosis groups, were analyzed. Key indices were selected by univariate analysis and Pearson correlation coefficient test. A support vector machine classifier-based model was developed through ten-fold cross validation. Model performance was assessed in terms of discrimination, calibration, and clinical usefulness at both per-slice and per-vessel levels. Continuous variables and categorical variables were compared between good and poor thrombolysis prognosis group by Mann-Whitney U-test and chi-square test, respectively. A nomogram was built by integrating clinical factors and the energy spectrum CTV index-based score calculated by the model. Six indices selected from 192 indices were used to build the predictive model. The ML model achieved area under the curves (AUCs) of 0.838 and 0.767 [95% CI (confidence interval), 0.825-0.850, 0.752-0.781] in the training and validation datasets at the per-slice level, and the per-vessel level AUCs were 0.945 and 0.876 (95% CI, 0.852-0.988, 0.763-0.948) in the training and validation datasets, respectively. The nomogram showed better performance with the per-vessel level AUC, accuracy, sensitivity and specificity, yielding 0.901(95% CI, 0.793-0.964), 86.2%, 87.9% and 84.0% in the validation dataset, respectively. There was no significant difference in the vessel distribution between good and poor thrombolysis prognosis groups (chi-square test, p = 0.671). The energy spectrum CTV index-based ML model achieved favorable effectiveness in predicting the outcome of vessel-level intravenous thrombolysis. A nomogram integrating clinical factors, and risk score calculated by the developed model showed improved performance and had potential to be used as a noninvasive preoperative tool for clinicians.

Multi-channel deep learning model-based myocardial spatial-temporal morphology feature on cardiac MRI cine images diagnoses the cause of LVH.

BACKGROUND: To develop a fully automatic framework for the diagnosis of cause for left ventricular hypertrophy (LVH) via cardiac cine images. METHODS: A total of 302 LVH patients with cine MRI images were recruited as the primary cohort. Another 53 LVH patients prospectively collected or from multi-centers were used as the external test dataset. Different models based on the cardiac regions (Model 1), segmented ventricle (Model 2) and ventricle mask (Model 3) were constructed. The diagnostic performance was accessed by the confusion matrix with respect to overall accuracy. The capability of the predictive models for binary classification of cardiac amyloidosis (CA), hypertrophic cardiomyopathy (HCM) or hypertensive heart disease (HHD) were also evaluated. Additionally, the diagnostic performance of best Model was compared with that of 7 radiologists/cardiologists. RESULTS: Model 3 showed the best performance with an overall classification accuracy up to 77.4% in the external test datasets. On the subtasks for identifying CA, HCM or HHD only, Model 3 also achieved the best performance with AUCs yielding 0.895-0.980, 0.879-0.984 and 0.848-0.983 in the validation, internal test and external test datasets, respectively. The deep learning model showed non-inferior diagnostic capability to the cardiovascular imaging expert and outperformed other radiologists/cardiologists. CONCLUSION: The combined model based on the mask of left ventricular segmented from multi-sequences cine MR images shows favorable and robust performance in diagnosing the cause of left ventricular hypertrophy, which could be served as a noninvasive tool and help clinical decision.

Deep learning-based high-accuracy detection for lumbar and cervical degenerative disease on T2-weighted MR images.

PURPOSE: To develop and validate a deep learning (DL) model for detecting lumbar degenerative disease in both sagittal and axial views of T2-weighted MRI and evaluate its generalized performance in detecting cervical degenerative disease. METHODS: T2-weighted MRI scans of 804 patients with symptoms of lumbar degenerative disease were retrospectively collected from three hospitals. The training dataset (n = 456) and internal validation dataset (n = 134) were randomly selected from the center I. Two external validation datasets comprising 100 and 114 patients were from center II and center III, respectively. A DL model based on 3D ResNet18 and transformer architecture was proposed to detect lumbar degenerative disease. In addition, a cervical MR image dataset comprising 200 patients from an independent hospital was used to evaluate the generalized performance of the DL model. The diagnostic performance was assessed by the free-response receiver operating characteristic (fROC) curve and precision-recall (PR) curve. Precision, recall, and F1-score were used to measure the DL model. RESULTS: A total of 2497 three-dimension retrogression annotations were labeled for training (n = 1157) and multicenter validation (n = 1340). The DL model showed excellent detection efficiency in the internal validation dataset, with F1-score achieving 0.971 and 0.903 on the sagittal and axial MR images, respectively. Good performance was also observed in the external validation dataset I (F1-score, 0.768 on sagittal MR images and 0.837 on axial MR images) and external validation dataset II (F1-score, 0.787 on sagittal MR images and 0.770 on axial MR images). Furthermore, the robustness of the DL model was demonstrated via transfer learning and generalized performance evaluation on the external cervical dataset, with the F1-score yielding 0.931 and 0.919 on the sagittal and axial MR images, respectively. CONCLUSION: The proposed DL model can automatically detect lumbar and cervical degenerative disease on T2-weighted MR images with good performance, robustness, and feasibility in clinical practice.

Comparison of MRI and CT-Based Radiomics and Their Combination for Early Identification of Pathological Response to Neoadjuvant Chemotherapy in Locally Advanced Gastric Cancer.

BACKGROUND: Current radiomics for treatment response assessment in gastric cancer (GC) have focused solely on Computed tomography (CT). The importance of multi-parametric magnetic resonance imaging (mp-MRI) radiomics in GC is less clear. PURPOSE: To compare and combine CT and mp-MRI radiomics for pretreatment identification of pathological response to neoadjuvant chemotherapy in GC. STUDY TYPE: Retrospective. POPULATION: Two hundred twenty-five GC patients were recruited and split into training (157) and validation dataset (68) in the ratio of 7:3 randomly. FIELD/SEQUENCE: T2-weighted fast spin echo (fat suppressed T2-weighted imaging [fs-T2WI]), diffusion weighted echo planar imaging (DWI), and fast gradient echo (dynamic contrast enhanced [DCE]) sequences at 3.0T. ASSESSMENT: Apparent diffusion coefficient (ADC) maps were generated from DWI. CT, fs-T2WI, ADC, DCE, and mp-MRI Radiomics score (Radscores) were compared between responders and non-responders. A multimodal nomogram combining CT and mp-MRI Radscores was developed. Patients were followed up for 3-65 months (median 19) after surgery, the overall survival (OS) and progression free survival (PFS) were calculated. STATISTICAL TESTS: A logistic regression classifier was applied to construct the five models. Each model's performance was evaluated using a receiver operating characteristic curve. The association of the nomogram with OS/PFS was evaluated by Kaplan-Meier survival analysis and C-index. A P value <0.05 was considered statistically significant. RESULTS: CT Radscore, mp-MRI Radscore and nomogram were significantly associated with tumor regression grading. The nomogram achieved the highest area under the curves (AUCs) of 0.893 (0.834-0.937) and 0.871 (0.767-0.940) in training and validation datasets, respectively. The C-index was 0.589 for OS and 0.601 for PFS. The AUCs of the mp-MRI model were not significantly different to that of the CT model in training (0.831 vs. 0.770, P = 0.267) and validation dataset (0.797 vs. 0.746, P = 0.137). DATA CONCLUSIONS: mp-MRI radiomics provides similar results to CT radiomics for early identification of pathologic response to neoadjuvant chemotherapy. The multimodal radiomics nomogram further improved the capability. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: 2.

Multiparametric MRI-based radiomics nomogram for early prediction of pathological response to neoadjuvant chemotherapy in locally advanced gastric cancer.

OBJECTIVES: To build and validate a multi-parametric MRI (mpMRI)-based radiomics nomogram for early prediction of treatment response to neoadjuvant chemotherapy (NAC) in locally advanced gastric cancer. METHODS: Baseline MRI were retrospectively enrolled from 141 patients with gastric adenocarcinoma who received NAC followed by radical gastrectomy. According to pathologic response of tumor regression grading (TRG), patients were labeled as responders (TRG = 0 + 1) and non-responders (TRG = 2 + 3) and further divided into a training (n = 85) and validation dataset (n = 56). Radiomics score (Radscore) were built from T2WI, ADC, and venous phase of dynamic-contrasted-enhanced MR imaging. Clinical information, laboratory indicators, MRI parameters, and follow-up data were also recorded. According to multivariable regression analysis, an mpMRI radiomics nomogram was built and its predictive ability was evaluated by ROC analysis. Decision curve analysis was applied to evaluate the clinical usefulness. Kaplan-Meier survival curves based on the nomogram were used to estimate the progression-free survival (PFS) and overall survival (OS) in the validation dataset. RESULTS: Both single sequence-based Radscores and mpMRI radiomics nomogram were associated with pathologic response (p < 0.001). The nomogram achieved the highest diagnostic ability with area under ROC curve of 0.844 (95% CI, 0.749-0.914) and 0.820 (95% CI, 0.695-0.910) in the training and validation datasets. The hazard ratio of the nomogram for PFS and OS prediction was 2.597 (95% CI: 1.046-6.451, log-rank p = 0.023) and 2.570 (95% CI: 1.166-5.666, log-rank p = 0.011). CONCLUSIONS: The mpMRI-based radiomics nomogram showed preferable performance in predicting pathologic response to NAC in LAGC. KEY POINTS: • This study investigated the value of multi-parametric MRI-based radiomics in predicting pathologic response to neoadjuvant chemotherapy in locally advanced gastric cancer. • The nomogram incorporating T2WI Radscore, ADC Radscore, and DCE Radscore as well as Borrmann classification outperformed the single sequence-based Radscore. • The nomogram also exhibited a promising prognostic ability for patient survival and enriched radiomics studies in gastric cancer.

Novel Radiomics-Clinical Model for the Noninvasive Prediction of New Fractures After Vertebral Augmentation.

PURPOSE: To investigate the noninvasive prediction model for new fractures after percutaneous vertebral augmentation (PVA) based on radiomics signature and clinical parameters. METHODS: Data from patients who were diagnosed with osteoporotic vertebral compression fracture (OVCF) and treated with PVA in our hospital between May 2014 and April 2019 were retrospectively analyzed. Radiomics features were extracted from T1-weighted magnetic resonance imaging (MRI) of the T11-L5 segments taken before PVA. Different radiomics models was developed by using linear discriminant analysis (LDA), multilayer perceptron (MLP), and stochastic gradient descent (SGD) classifiers. A nomogram was constructed by integrating clinical parameters and Radscore that calculated by the best radiomics model. The model performance was quantified in terms of discrimination, calibration and clinical usefulness. RESULT: Four clinical parameters and 16 selected radiomics features were used for model development. The clinical model showed poor discrimination capability with area under the curves (AUCs) yielding of 0.522 in the training dataset and 0.517 in the validation dataset. The LDA, MLP and SGD classifier-based radiomics model had achieved AUCs of 0.793, 0.810, and 0.797 in the training dataset, and 0.719, 0.704, and 0.725 in the validation dataset, respectively. The nomogram showed the best performance with AUCs achieving 0.810 and 0.754 in the training and validation datasets, respectively. The decision curve analysis demonstrated the net benefit of the nomogram was higher than that of other models. CONCLUSION: Our findings indicate that combining clinical features with radiomics features from pre-augmentation T1-weighted MRI can be used to develop a nomogram that can predict new fractures in patients after PVA.

Prediction of Chemotherapy Efficacy in Patients with Colorectal Cancer Ovarian Metastases: A Preliminary Study Using Contrast-Enhanced Computed-Tomography-Based Radiomics.

Ovarian metastasis (OM) from colorectal cancer (CRC) is infrequent and has a poor prognosis. The purpose of this study is to investigate the value of a contrast-enhanced CT-based radiomics model in predicting ovarian metastasis from colorectal cancer outcomes after systemic chemotherapy. A total of 52 ovarian metastatic CRC patients who received first-line systemic chemotherapy were retrospectively included in this study and were categorized into chemo-benefit (C+) and no-chemo-benefit (C-) groups, using Response Criteria in Solid Tumors (RECIST v1.1) as the standard. A total of 1743 radiomics features were extracted from baseline CT, three methods were adopted during the feature selection, and five prediction models were constructed. Receiver operating characteristic (ROC) analysis, calibration analysis, and decision curve analysis (DCA) were used to evaluate the diagnostic performance and clinical utility of each model. Among those machine-learning-based radiomics models, the SVM model showed the best performance on the validation dataset, with AUC, accuracy, sensitivity, and specificity of 0.903 (95% CI, 0.788-0.967), 88.5%, 95.7%, and 82.8%, respectively. All radiomics models exhibited good calibration, and the DCA demonstrated that the SVM model had a higher net benefit than other models across the majority of the range of threshold probabilities. Our findings showed that contrast-enhanced CT-based radiomics models have high discriminating power in predicting the outcome of colorectal cancer ovarian metastases patients receiving chemotherapy.

RESOLVE-DWI-based deep learning nomogram for prediction of normal-sized lymph node metastasis in cervical cancer: a preliminary study.

BACKGROUND: It is difficult to predict normal-sized lymph node metastasis (LNM) in cervical cancer clinically. We aimed to investigate the feasibility of using deep learning (DL) nomogram based on readout segmentation of long variable echo-trains diffusion weighted imaging (RESOLVE-DWI) and related patient information to preoperatively predict normal-sized LNM in patients with cervical cancer. METHODS: A dataset of MR images [RESOLVE-DWI and apparent diffusion coefficient (ADC)] and patient information (age, tumor size, International Federation of Gynecology and Obstetrics stage, ADC value and squamous cell carcinoma antigen level) of 169 patients with cervical cancer between November 2013 and January 2022 were retrospectively collected. The LNM status was determined by final histopathology. The collected studies were randomly divided into a development cohort (n = 126) and a test cohort (n = 43). A single-channel convolutional neural network (CNN) and a multi-channel CNN based on ResNeSt architectures were proposed for predicting normal-sized LNM from single or multi modalities of MR images, respectively. A DL nomogram was constructed by incorporating the clinical information and the multi-channel CNN. These models' performance was analyzed by the receiver operating characteristic analysis in the test cohort. RESULTS: Compared to the single-channel CNN model using RESOLVE-DWI and ADC respectively, the multi-channel CNN model that integrating both two MR modalities showed improved performance in development cohort [AUC 0.848; 95% confidence interval (CI) 0.774-0.906] and test cohort (AUC 0.767; 95% CI 0.613-0.882). The DL nomogram showed the best performance in development cohort (AUC 0.890; 95% CI 0.821-0.938) and test cohort (AUC 0.844; 95% CI 0.701-0.936). CONCLUSION: The DL nomogram incorporating RESOLVE-DWI and clinical information has the potential to preoperatively predict normal-sized LNM of cervical cancer.

Machine Learning Models for Diagnosis of Parkinson's Disease Using Multiple Structural Magnetic Resonance Imaging Features.

Purpose: This study aimed to develop machine learning models for the diagnosis of Parkinson's disease (PD) using multiple structural magnetic resonance imaging (MRI) features and validate their performance. Methods: Brain structural MRI scans of 60 patients with PD and 56 normal controls (NCs) were enrolled as development dataset and 69 patients with PD and 71 NCs from Parkinson's Progression Markers Initiative (PPMI) dataset as independent test dataset. First, multiple structural MRI features were extracted from cerebellar, subcortical, and cortical regions of the brain. Then, the Pearson's correlation test and least absolute shrinkage and selection operator (LASSO) regression were used to select the most discriminating features. Finally, using logistic regression (LR) classifier with the 5-fold cross-validation scheme in the development dataset, the cerebellar, subcortical, cortical, and a combined model based on all features were constructed separately. The diagnostic performance and clinical net benefit of each model were evaluated with the receiver operating characteristic (ROC) analysis and the decision curve analysis (DCA) in both datasets. Results: After feature selection, 5 cerebellar (absolute value of left lobule crus II cortical thickness (CT) and right lobule IV volume, relative value of right lobule VIIIA CT and lobule VI/VIIIA gray matter volume), 3 subcortical (asymmetry index of caudate volume, relative value of left caudate volume, and absolute value of right lateral ventricle), and 4 cortical features (local gyrification index of right anterior circular insular sulcus and anterior agranular insula complex, local fractal dimension of right middle insular area, and CT of left supplementary and cingulate eye field) were selected as the most distinguishing features. The area under the curve (AUC) values of the cerebellar, subcortical, cortical, and combined models were 0.679, 0.555, 0.767, and 0.781, respectively, for the development dataset and 0.646, 0.632, 0.690, and 0.756, respectively, for the independent test dataset, respectively. The combined model showed higher performance than the other models (Delong's test, all p-values < 0.05). All models showed good calibration, and the DCA demonstrated that the combined model has a higher net benefit than other models. Conclusion: The combined model showed favorable diagnostic performance and clinical net benefit and had the potential to be used as a non-invasive method for the diagnosis of PD.

Multiphase Dual-Energy Spectral CT-Based Deep Learning Method for the Noninvasive Prediction of Head and Neck Lymph Nodes Metastasis in Patients With Papillary Thyroid Cancer.

Purpose: To develop deep learning (DL) models based on multiphase dual-energy spectral CT for predicting lymph nodes metastasis preoperatively and noninvasively in papillary thyroid cancer patients. Methods: A total of 293 lymph nodes from 78 papillary thyroid cancer patients who underwent dual-energy spectral CT before lymphadenectomy were enrolled in this retrospective study. The lymph nodes were randomly divided into a development set and an independent testing set following a 4:1 ratio. Four single-modality DL models based on CT-A model, CT-V model, Iodine-A model and Iodine-V model and a multichannel DL model incorporating all modalities (Combined model) were proposed for the prediction of lymph nodes metastasis. A CT-feature model was also built on the selected CT image features. The model performance was evaluated with respect to discrimination, calibration and clinical usefulness. In addition, the diagnostic performance of the Combined model was also compared with four radiologists in the independent test set. Results: The AUCs of the CT-A, CT-V, Iodine-A, Iodine-V and CT-feature models were 0.865, 0.849, 0.791, 0.785 and 0.746 in the development set and 0.830, 0.822, 0.744, 0.739 and 0.732 in the testing set. The Combined model had outperformed the other models and achieved the best performance with AUCs yielding 0.890 in the development set and 0.865 in the independent testing set. The Combined model showed good calibration, and the decision curve analysis demonstrated that the net benefit of the Combined model was higher than that of the other models across the majority of threshold probabilities. The Combined model also showed noninferior diagnostic capability compared with the senior radiologists and significantly outperformed the junior radiologists, and the interobserver agreement of junior radiologists was also improved after artificial intelligence assistance. Conclusion: The Combined model integrating both CT images and iodine maps of the arterial and venous phases showed good performance in predicting lymph nodes metastasis in papillary thyroid cancer patients, which could facilitate clinical decision-making.

A Deep Learning Model Based on MRI and Clinical Factors Facilitates Noninvasive Evaluation of KRAS Mutation in Rectal Cancer.

BACKGROUND: Recent studies showed the potential of MRI-based deep learning (DL) for assessing treatment response in rectal cancer, but the role of MRI-based DL in evaluating Kirsten rat sarcoma viral oncogene homologue (KRAS) mutation remains unclear. PURPOSE: To develop a DL method based on T2-weighted imaging (T2WI) and clinical factors for noninvasively evaluating KRAS mutation in rectal cancer. STUDY TYPE: Retrospective. SUBJECTS: A total of 376 patients (108 women [28.7%]) with histopathology-confirmed rectal adenocarcinoma and KRAS mutation status. FIELD STRENGTH/SEQUENCE: A 3 T, turbo spin echo T2WI and single-shot echo-planar diffusion-weighted imaging (b = 0, 1000 sec/mm2 ). ASSESSMENT: A clinical model was constructed with clinical factors (age, gender, carcinoembryonic antigen level, and carbohydrate antigen 199 level) and MRI features (tumor length, tumor location, tumor stage, lymph node stage, and extramural vascular invasion), and two DL models based on modified MobileNetV2 architecture were evaluated for diagnosing KRAS mutation based on T2WI alone (image model) or both T2WI and clinical factors (combined model). The clinical usefulness of these models was evaluated through calibration analysis and decision curve analysis (DCA). STATISTICAL TESTS: Mann-Whitney U test, Chi-squared test, Fisher's exact test, logistic regression analysis, receiver operating characteristic curve (ROC), Delong's test, Hosmer-Lemeshow test, interclass correlation coefficients, and Fleiss kappa coefficients (P < 0.05 was considered statistically significant). RESULTS: All the nine clinical-MRI characteristics were included for clinical model development. The clinical model, image model, and combined model in the testing cohort demonstrated good calibration and achieved areas under the curve (AUCs) of 0.668, 0.765, and 0.841, respectively. The combined model showed improved performance compared to the clinical model and image model in two cohorts. DCA confirmed the higher net benefit of the combined model than the other two models when the threshold probability is between 0.05 and 0.85. DATA CONCLUSION: The proposed combined DL model incorporating T2WI and clinical factors may show good diagnostic performance. Thus, it could potentially serve as a supplementary approach for noninvasively evaluating KRAS mutation in rectal cancer. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Machine learning analysis for the noninvasive prediction of lymphovascular invasion in gastric cancer using PET/CT and enhanced CT-based radiomics and clinical variables.

PURPOSE: Lymphovascular invasion (LVI) is associated with metastasis and poor survival in patients with gastric cancer, yet the noninvasive diagnosis of LVI is difficult. This study aims to develop predictive models using different machine learning (ML) classifiers based on both enhanced CT and PET/CT images and clinical variables for preoperatively predicting lymphovascular invasion (LVI) status of gastric cancer. METHODS: A total of 101 patients with gastric cancer who underwent surgery were retrospectively recruited, and the LVI status was confirmed by pathological analysis. Patients were randomly divided into a training dataset (n = 76) and a validation dataset (n = 25). By 3D manual segmentation, radiomics features were extracted from the PET and venous phase CT images. Image models, clinical models, and combined models were constructed by selected enhanced CT-based and PET-based radiomics features, clinical factors, and a combination of both, respectively. Three ML classifiers including adaptive boosting (AdaBoost), linear discriminant analysis (LDA), and logistic regression (LR) were used for model development. The performance of these predictive models was evaluated with respect to discrimination, calibration, and clinical usefulness. RESULTS: Ten radiomics features and eight clinical factors were selected for the development of predictive models. In the validation dataset, the area under curve (AUC) values of clinical models using AdaBoost, LDA, and LR classifiers were 0.742, 0.706, and 0.690, respectively. The image models using AdaBoost, LDA, and LR classifiers achieved an AUC of 0.849, 0.778, and 0.810, respectively. The combined models showed improved performance than the image models and the clinical models, with the AUC values of AdaBoost, LDA, and LR classifier yielding 0.944, 0.929, and 0.921, respectively. The combined models also showed good calibration and clinical usefulness for LVI prediction. CONCLUSION: ML-based models integrating PET/CT and enhanced CT radiomics features and clinical factors have good discrimination capability, which could serve as a noninvasive, preoperative tool for the prediction of LVI and assist surgical treatment decisions in patients with gastric cancer.

Radiomic analysis of enhanced CMR cine images predicts left ventricular remodeling after TAVR in patients with symptomatic severe aortic stenosis.

Objective: This study aimed to develop enhanced cine image-based radiomic models for non-invasive prediction of left ventricular adverse remodeling following transcatheter aortic valve replacement (TAVR) in symptomatic severe aortic stenosis. Methods: A total of 69 patients (male:female = 37:32, median age: 66 years, range: 47-83 years) were retrospectively recruited, and severe aortic stenosis was confirmed via transthoracic echocardiography detection. The enhanced cine images and clinical variables were collected, and three types of regions of interest (ROIs) containing the left ventricular (LV) myocardium from the short-axis view at the basal, middle, and apical LV levels were manually labeled, respectively. The radiomic features were extracted and further selected by using the least absolute shrinkage and selection operator (LASSO) regression analysis. Clinical variables were also selected through univariate regression analysis. The predictive models using logistic regression classifier were developed and validated through leave-one-out cross-validation. The model performance was evaluated with respect to discrimination, calibration, and clinical usefulness. Results: Five basal levels, seven middle levels, eight apical level radiomic features, and three clinical factors were finally selected for model development. The radiomic models using features from basal level (Rad I), middle level (Rad II), and apical level (Rad III) had achieved areas under the curve (AUCs) of 0.761, 0.909, and 0.913 in the training dataset and 0.718, 0.836, and 0.845 in the validation dataset, respectively. The performance of these radiomic models was improved after integrating clinical factors, with AUCs of the Combined I, Combined II, and Combined III models increasing to 0.906, 0.956, and 0.959 in the training dataset and 0.784, 0.873, and 0.891 in the validation dataset, respectively. All models showed good calibration, and the decision curve analysis indicated that the Combined III model had a higher net benefit than other models across the majority of threshold probabilities. Conclusion: Radiomic models and combined models at the mid and apical slices showed outstanding and comparable predictive effectiveness of adverse remodeling for patients with symptomatic severe aortic stenosis after TAVR, and both models were significantly better than the models of basal slice. The cardiac magnetic resonance radiomic analysis might serve as an effective tool for accurately predicting left ventricular adverse remodeling following TAVR in patients with symptomatic severe aortic stenosis.

A Bounding Box-Based Radiomics Model for Detecting Occult Peritoneal Metastasis in Advanced Gastric Cancer: A Multicenter Study.

PURPOSE: To develop a bounding box (BBOX)-based radiomics model for the preoperative diagnosis of occult peritoneal metastasis (OPM) in advanced gastric cancer (AGC) patients. MATERIALS AND METHODS: 599 AGC patients from 3 centers were retrospectively enrolled and were divided into training, validation, and testing cohorts. The minimum circumscribed rectangle of the ROIs for the largest tumor area (R_BBOX), the nonoverlapping area between the tumor and R_BBOX (peritumoral area; PERI) and the smallest rectangle that could completely contain the tumor determined by a radiologist (M_BBOX) were used as inputs to extract radiomic features. Multivariate logistic regression was used to construct a radiomics model to estimate the preoperative probability of OPM in AGC patients. RESULTS: The M_BBOX model was not significantly different from R_BBOX in the validation cohort [AUC: M_BBOX model 0.871 (95% CI, 0.814-0.940) vs. R_BBOX model 0.873 (95% CI, 0.820-0.940); p = 0.937]. M_BBOX was selected as the final radiomics model because of its extremely low annotation cost and superior OPM discrimination performance (sensitivity of 85.7% and specificity of 82.8%) over the clinical model, and this radiomics model showed comparable diagnostic efficacy in the testing cohort. CONCLUSIONS: The BBOX-based radiomics could serve as a simpler reliable and powerful tool for the preoperative diagnosis of OPM in AGC patients. And M_BBOX-based radiomics is simpler and less time consuming.

Consecutive Serial Non-Contrast CT Scan-Based Deep Learning Model Facilitates the Prediction of Tumor Invasiveness of Ground-Glass Nodules.

INTRODUCTION: Tumors are continuously evolving biological systems which can be monitored by medical imaging. Previous studies only focus on single timepoint images, whether the performance could be further improved by using serial noncontrast CT imaging obtained during nodule follow-up management remains unclear. In this study, we evaluated DL model for predicting tumor invasiveness of GGNs through analyzing time series CT images. METHODS: A total of 168 pathologically confirmed GGN cases (48 noninvasive lesions and 120 invasive lesions) were retrospectively collected and randomly assigned to the development dataset (n = 123) and independent testing dataset (n = 45). All patients underwent consecutive noncontrast CT examinations, and the baseline CT and 3-month follow-up CT images were collected. The gross region of interest (ROI) patches containing only tumor region and the full ROI patches including both tumor and peritumor regions were cropped from CT images. A baseline model was built on the image features and demographic features. Four DL models were proposed: two single-DL model using gross ROI (model 1) or full ROI patches (model 3) from baseline CT images, and two serial-DL models using gross ROI (model 2) or full ROI patches (model 4) from consecutive CT images (baseline scan and 3-month follow-up scan). In addition, a combined model integrating serial full ROI patches and clinical information was also constructed. The performance of these predictive models was assessed with respect to discrimination and clinical usefulness. RESULTS: The area under the curve (AUC) of the baseline model, models 1, 2, 3, and 4 were 0.562 [(95% confidence interval (C)], 0.406~0.710), 0.693 (95% CI, 0.538-0.822), 0.787 (95% CI, 0.639-0.895), 0.727 (95% CI, 0.573-0.849), and 0.811 (95% CI, 0.667-0.912) in the independent testing dataset, respectively. The results indicated that the peritumor region had potential to contribute to tumor invasiveness prediction, and the model performance was further improved by integrating imaging scans at multiple timepoints. Furthermore, the combined model showed best discrimination ability, with AUC, sensitivity, specificity, and accuracy achieving 0.831 (95% CI, 0.690-0.926), 86.7%, 73.3%, and 82.2%, respectively. CONCLUSION: The DL model integrating full ROIs from serial CT images shows improved predictive performance in differentiating noninvasive from invasive GGNs than the model using only baseline CT images, which could benefit the clinical management of GGNs.

Development and validation of MRI-based deep learning models for prediction of microsatellite instability in rectal cancer.

BACKGROUND: Microsatellite instability (MSI) predetermines responses to adjuvant 5-fluorouracil and immunotherapy in rectal cancer and serves as a prognostic biomarker for clinical outcomes. Our objective was to develop and validate a deep learning model that could preoperatively predict the MSI status of rectal cancer based on magnetic resonance images. METHODS: This single-center retrospective study included 491 rectal cancer patients with pathologically proven microsatellite status. Patients were randomly divided into the training/validation cohort (n = 395) and the testing cohort (n = 96). A clinical model using logistic regression was constructed to discriminate MSI status using only clinical factors. Based on a modified MobileNetV2 architecture, deep learning models were tested for the predictive ability of MSI status from magnetic resonance images, with or without integrating clinical factors. RESULTS: The clinical model correctly classified 37.5% of MSI status in the testing cohort, with an AUC value of 0.573 (95% confidence interval [CI], 0.468 ~ 0.674). The pure imaging-based model and the combined model correctly classified 75.0% and 85.4% of MSI status in the testing cohort, with AUC values of 0.820 (95% CI, 0.718 ~ 0.884) and 0.868 (95% CI, 0.784 ~ 0.929), respectively. Both deep learning models performed better than the clinical model (p < 0.05). There was no statistically significant difference between the deep learning models with or without integrating clinical factors. CONCLUSIONS: Deep learning based on high-resolution T2-weighted magnetic resonance images showed a good predictive performance for MSI status in rectal cancer patients. The proposed model may help to identify patients who would benefit from chemotherapy or immunotherapy and determine individualized therapeutic strategies for these patients.