Development and Validation of Multi-Omics Thymoma Risk Classification Model Based on Transfer Learning.

This study aimed to construct a machine learning model using clinical variables and ultrasound radiomics features for the prediction of the benign or malignant nature of pancreatic tumors. 242 pancreatic tumor patients who were hospitalized at the First Affiliated Hospital of Guangxi Medical University between January 2020 and June 2023 were included in this retrospective study. The patients were randomly divided into a training cohort (n=169) and a test cohort (n=73). We collected 28 clinical features from the patients. Concurrently, 306 radiomics features were extracted from the ultrasound images of the patients' tumors. Initially, a clinical model was constructed using the logistic regression algorithm. Subsequently, radiomics models were built using SVM, random forest, XGBoost, and KNN algorithms. Finally, we combined clinical features with a new feature RAD prob calculated by applying radiomics model to construct a fusion model, and developed a nomogram based on the fusion model. The performance of the fusion model surpassed that of both the clinical and radiomics models. In the training cohort, the fusion model achieved an AUC of 0.978 (95% CI: 0.96-0.99) during 5-fold cross-validation and an AUC of 0.925 (95% CI: 0.86-0.98) in the test cohort. Calibration curve and decision curve analyses demonstrated that the nomogram constructed from the fusion model has high accuracy and clinical utility. The fusion model containing clinical and ultrasound radiomics features showed excellent performance in predicting the benign or malignant nature of pancreatic tumors.

ObjectivesThe aim was to determine whether the dual-energy CT radiomics model derived from an iodine map (IM) has incremental diagnostic value for the model based on 120-kV equivalent mixed images (120 kVp) in preoperative restaging of serosal invasion with locally advanced gastric cancer (LAGC) after neoadjuvant chemotherapy (NAC).MethodsA total of 155 patients (110 in the training cohort and 45 in the testing cohort) with LAGC who had standard NAC before surgery were retrospectively enrolled. All CT images were analyzed by two radiologists for manual classification. Volumes of interests (VOIs) were delineated semi-automatically, and 1,226 radiomics features were extracted from every segmented lesion in both IM and 120 kVp images, respectively. Spearman’s correlation analysis and the least absolute shrinkage and selection operator (LASSO) penalized logistic regression were implemented for filtering unstable and redundant features and screening out vital features. Two predictive models (120 kVp and IM-120 kVp) based on 120 kVp selected features only and 120 kVp combined with IM selected features were established by multivariate logistic regression analysis. We then build a combination model (ComModel) developed with IM-120 kVp signature and ycT. The performance of these three models and manual classification were evaluated and compared.ResultThree radiomics models showed great predictive accuracy and performance in both the training and testing cohorts (ComModel: AUC: training, 0.953, testing, 0.914; IM-120 kVp: AUC: training, 0.953, testing, 0.879; 120 kVp: AUC: training, 0.940, testing, 0.831). All these models showed higher diagnostic accuracy (ComModel: 88.9%, IM-120 kVp: 84.4%, 120 kVp: 80.0%) than manual classification (68.9%) in the testing group. ComModel and IM-120 kVp model had better performances than manual classification both in the training (both p<0.001) and testing cohorts (p<0.001 and p=0.034, respectively).ConclusionsDual-energy CT-based radiomics models demonstrated convincible diagnostic performance in differentiating serosal invasion in preoperative restaging for LAGC. The radiomics features derived from IM showed great potential for improving the diagnostic capability.

To explore a novel approach for the early prediction of malignant cerebral edema (MCE) in stroke patients using radiomics features extracted from dual-energy computed tomography angiography (DE-CTA) reconstructed images of infarcted brain tissue. This retrospective study enrolled 398 stroke patients who underwent DE-CTA between April 2016 and November 2022 from three medical centers. Patients were allocated into a training cohort (n = 227) and a test cohort (n = 171) based on their source institution. Radiomics features were extracted from DE-CTA reconstructions of the infarct tissue. Radiomics models were built for each reconstruction scan images and for a combined model using features from all images. Additionally, a nomogram model integrating significant radiomics features and clinical characteristics was developed. The diagnostic performance of all models was evaluated using receiver operating characteristic (ROC) curve analysis, calculating the area under the curve (AUC). Radiomics models based on each individual DE-CTA reconstruction demonstrated robust predictive performance. In the test cohort, the combined radiomics model integrating features from all DE-CTA reconstructions significantly superior predictive performance compared to most single-image models. In the test cohort, both the combined radiomics model (AUC = 0.950) and the nomogram model (AUC = 0.935) significantly outperformed the clinical model (AUC = 0.574), both p < 0.001. No significant difference in performance was observed between the nomogram and the radiomics model (p = 0.30). Radiomics models derived from DE-CTA reconstructed images of infarcted tissue provide a reliable method for the early prediction of MCE development in stroke patients.

We aimed at comparing intratumoral and peritumoral deep learning, radiomics, and fusion models in predicting KRAS mutations in rectal cancer using endorectal ultrasound imaging. This study included 304 patients with rectal cancer from Fujian Medical University Union Hospital. The patients were randomly divided into a training group (213 patients) and a test group (91 patients) at a 7:3 ratio. Radiomics and deep learning models were established using primary tumor and peritumoral images. In the optimally performing regions-of-interest, two fusion strategies, a feature-based and a decision-based model, were employed to build the fusion models. The Shapley additive explanation (SHAP) method was used to evaluate the significance of features in the optimal radiomics, deep learning, and fusion models. The performance of each model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). In the test cohort, both the radiomics and deep learning models exhibited optimal performance with a 10-pixel patch extension, yielding AUC values of 0.824 and 0.856, respectively. The feature-based DLRexpand10_FB model attained the highest AUC (0.896) across all study sets. In addition, the DLRexpand10_FB model demonstrated excellent sensitivity, specificity, and DCA. SHAP analysis underscored the deep learning feature (DL_1) as the most significant factor in the hybrid model. The feature-based fusion model DLRexpand10_FB can be employed to predict KRAS gene mutations based on pretreatment endorectal ultrasound images of rectal cancer. The integration of peritumoral regions enhanced the predictive performance of both the radiomics and deep learning models.

A Novel Machine-Learning Model to Predict Early Relapse in Mantle Cell Lymphoma (MCL)

US-based radiomics analysis of different machine learning models for differentiating benign and malignant BI-RADS 4A breast lesions

Machine Learning-Based CT Radiomics Model to Predict the Risk of Hip Fragility Fracture.

Rectal cancer is a common malignant tumor, and the presence of distant metastasis is critically important for determining treatment strategies. This study aimed to develop a deep learning radiomics model based on contrast-enhanced transrectal ultrasound (CETRUS) imaging to predict distant metastasis in patients with rectal cancer. We retrospectively analyzed the clinical data and CETRUS imaging of 878 patients with rectal cancer treated at The First Affiliated Hospital of Guangxi Medical University. Univariate and multivariate logistic regression analyses were performed to identify relevant clinical variables. Deep learning radiomics features were extracted using a pretrained DenseNet201 model and subsequently selected via the Mann-Whitney U test, Spearman correlation analysis, and least absolute shrinkage and selection operator regression. Separate models were constructed based on clinical data, two-dimensional ultrasound (TDUS), color Doppler ultrasound (CDUS), and contrast-enhanced ultrasound (CEUS) imaging. The optimal deep learning radiomics model was then combined with the clinical model to develop an integrated predictive model. The clinical prediction model achieved area under the curve (AUC) values of 0.631 and 0.604 in the training and test cohorts, respectively. Among the three deep learning radiomics models, the CEUS model demonstrated the best performance, with AUC of 0.950 and 0.740 in the training and test cohorts, respectively. The TDUS model achieved AUC of 0.935 and 0.586, while the CDUS model yielded AUC of 0.805 and 0.521. The integrated model combining the clinical and contrast-enhanced ultrasound radiomics models achieved AUC of 0.947 and 0.749 in the training and test cohorts, respectively. The clinical-deep learning radiomics model based on CETRUS showed promising predictive performance in assessing distant metastasis in rectal cancer patients. This approach has the potential to assist clinicians in developing personalized patient management strategies, pending further validation to confirm its clinical applicability.

This study aimed to investigate whether establishing a machine learning (ML) model based on contrast-enhanced cone-beam breast computed tomography (CE-CBBCT) radiomic features could predict human epidermal growth factor receptor 2-positive breast cancer (BC). Eighty-eight patients diagnosed with invasive BC who underwent preoperative CE-CBBCT were retrospectively enrolled. Patients were randomly assigned to the training and testing cohorts at a ratio of approximately 7:3. A total of 1046 quantitative radiomics features were extracted from the CE-CBBCT images using PyRadiomics. Z-score normalization was used to standardize the radiomics features, and Pearson correlation coefficient and one-way analysis of variance were used to explore the significant features. Six ML algorithms (support vector machine, random forest [RF], logistic regression, adaboost, linear discriminant analysis, and decision tree) were used to construct optimal predictive models. Receiver operating characteristic curves were constructed and the area under the curve (AUC) was calculated. Four top-performing radiomic models were selected to develop the 6 predictive features. The AUC values for support vector machine, linear discriminant analysis, RF, logistic regression, adaboost, and decision tree were 0.741, 0.753, 1.000, 0.752, 1.000, and 1.000, respectively, in the training cohort, and 0.700, 0.671, 0.806, 0.665, 0.706, and 0.712, respectively, in the testing cohort. Notably, the RF model exhibited the highest predictive ability with an AUC of 0.806 in the testing cohort. For the RF model, the DeLong test showed statistically significant differences in the AUC between the training and testing cohorts (Z = 2.105, P = .035). The ML model based on CE-CBBCT radiomics features showed promising predictive ability for human epidermal growth factor receptor 2-positive BC, with the RF model demonstrating the best diagnostic performance.

ObjectivesMultidrug-resistant TB (MDR-TB) is a severe burden and public health threat worldwide. This study aimed to develop a radiomics model based on the tree-in-bud (TIB) sign and nodules and validate its predictive performance for MDR-TB.MethodsWe retrospectively recruited 454 patients with proven active TB from two hospitals and classified them into three training and testing cohorts: TIB (n = 295, 102), nodules (n = 302, 97), and their combination (n = 261, 81). Radiomics features relating to TIB and nodules were separately extracted. The maximal information coefficient and recursive feature elimination were used to select informative features per the two signs. Two radiomics models were constructed to predict MDR-TB using a random forest classifier. Then, a combined model was built incorporating radiomics features based on these two signs. The capability of the models in the combined training and testing cohorts was validated with ROC curves.ResultsSixteen features were extracted from TIB and 15 from nodules. The AUCs of the combined model were slightly higher than those of the TIB model in the combined training cohort (0.911 versus 0.877, p > 0.05) and testing cohort (0.820 versus 0.786, p < 0.05) and similar to the performance of the nodules model in the combined training cohort (0.911 versus 0.933, p > 0.05) and testing cohort (0.820 versus 0.855, p > 0.05).ConclusionsThe CT-based radiomics models hold promise for use as a non-invasive tool in the prediction of MDR-TB.Clinical relevance statementOur study revealed that complementary information regarding MDR-TB can be provided by radiomics based on the TIB sign and nodules. The proposed radiomics models may be new markers to predict MDR in active TB patients.Key Points• This is the first study to build, validate, and apply radiomics based on tree-in-bud sign and nodules for the prediction of MDR-TB.• The radiomics model showed a favorable performance for the identification of MDR-TB.• The combined model holds potential to be used as a diagnostic tool in routine clinical practice.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00330-023-09589-x.

Computed tomography-based deep-learning prediction of neoadjuvant chemoradiotherapy treatment response in esophageal squamous cell carcinoma

Optimizing the radiomics-machine-learning model based on non-contrast enhanced CT for the simplified risk categorization of thymic epithelial tumors: A large cohort retrospective study

To evaluate the value of radiomics based on ultrasound in differentiating uterine fibroids (UFs) and uterine adenomyomas (AMs) and construct a noninvasive tool for this differentiation. The clinical data and ultrasound images of 659 patients diagnosed with UFs or AMs postoperatively were retrospectively analyzed at 3 institutions from January 2020 to December 2024. The cohort comprised patients with 422 UFs and 237 uterine AMs, divided into training (Institution 1 and Institution 2) and external test cohort (Institution 3). Radiomics features extracted from ultrasound images were used to develop models based on different machine learning classifiers. A combined model was developed combining radiomics features with clinical characteristics and a nomogram was depicted. The performance of the models was assessed by area under the receiver operating characteristic curve (AUC), calibration curve, and decision curve. The clinical model demonstrated moderate performance with AUCs of 0.767 (95% CI: 0.7259-0.8082) in the training cohort and 0.747 (95% CI: 0.6003-0.8933) in the test cohort. The radiomics model based on the support vector machine (SVM) showed the best performance in discriminating UFs and AMs, with AUCs of 0.899 (95% CI: 0.8721-0.9256) in the training cohort and 0.823 (95% CI: 0.7297-0.9159) in the external test cohort, respectively. The combined model presented better efficacy compared with the clinical model and the radiomics model, with AUCs of 0.923 (95% CI: 0.9015-0.9451) and 0.894 (95% CI: 0.8285-0.9596) in the training and external test cohorts, respectively. The calibration curves suggested good consistency and decision curves showed the highest overall clinical benefit for the combined model. Ultrasound radiomics model based on SVM is feasible to differentiate UFs and AMs, and the combined model is a reliable and effective noninvasive tool for differentiation between UFs and AMs, which can assist clinicians in preoperative planning and fertility preservation strategies.

ObjectivesThis study aimed to develop and validate a deep-learning radiomics model using CT, T2, and DWI images for predicting pathological complete response (pCR) in patients with esophageal squamous cell carcinoma (ESCC) undergoing neoadjuvant chemoradiotherapy (nCRT).Materials and methodsPatients with ESCC undergoing nCRT followed by surgery were retrospectively enrolled from three institutions and divided into training and testing cohorts. Both traditional and deep-learning radiomics features were extracted from pre-treatment CT, T2, and DWI. Multiple radiomics models were developed, both single modality and integrated, using machine learning algorithms. The models’ performance was assessed using receiver operating characteristic curve analysis, with the area under the curve (AUC) as a primary metric, alongside sensitivity and specificity from the cut-off analysis.ResultsThe study involved 151 patients, among whom 63 achieved pCR. The training cohort consisted of 89 patients from Institution 1 (median age 62, 73 males) and the testing cohort included 52 patients from Institution 2 (median age 62, 41 males), and 10 in a clinical trial from Institution 3 (median age 69, 9 males). The integrated model, combining traditional and deep learning radiomics features from CT, T2, and DWI, demonstrated the best performance with an AUC of 0.868 (95% CI: 0.766–0.959), sensitivity of 88% (95% CI: 73.9–100), and specificity of 78.4% (95% CI: 63.6–90.2) in the testing cohort. This model outperformed single-modality models and the clinical model.ConclusionA multimodality deep learning radiomics model, utilizing CT, T2, and DWI images, was developed and validated for accurately predicting pCR of ESCC following nCRT.Critical relevance statementOur research demonstrates the satisfactory predictive value of multimodality deep learning radiomics for the response of nCRT in ESCC and provides a potentially helpful tool for personalized treatment including organ preservation strategy.Key PointsAfter neoadjuvant chemoradiotherapy, patients with ESCC have pCR rates of about 40%.The multimodality deep learning radiomics model, could predict pCR after nCRT with high accuracy.The multimodality radiomics can be helpful in personalized treatment of esophageal cancer.Graphical

Blood serum MRS metabolomics screening of human lung cancer prior to disease diagnosis Lung cancer (LuCa), the leading cause of cancer deaths, is often diagnosed late. Low-dose spiral CT can detect small and early stage LuCa lesions, but cannot be used widely as an annual LuCa screening tool. Metabolomics detects global metabolite variations under physiology and pathology. Metabolomic profiles measured from blood may reveal LuCa at early stages as a screening tool to triage suspicious patients to CT tests. Blood sera obtained from LuCa patients prior to their diagnosis were studied with MRS to establish LuCa screening metabolomic profiles.Samples. Sera from non-small cell LuCa (NSCLC) patients and their age, gender and smoking habit matched healthy controls were grouped according to the design of training-testing-validation cohorts in this study. The training cohort included 25 NSCLC sera from patients at the time of diagnosis (TOD) and controls (Ctrl); the testing cohort consisted 25 sera collected 0.5 to 5 yrs prior to diagnosis (PTD) from the 25 NSCLC patients in the training cohort; and the validation cohort recruited sera collected less than 2 yrs PTD from additional 54 NSCLC patients and controls. MR Spectroscopy. High resolution magic angle spinning (HRMAS) MRS analysis of serum samples are performed at 4°C by a 600MHz Bruker spectrometer at 3,600Hz spinning rate. Spectra are analyzed with a MatLab-based curve fitting program. Data Analysis. 57 spectral regions were selected based on the training and testing cohorts. Following selections of these regions, all data analytical procedures, including principal component and canonical analyses, were performed on the training cohort and followed by the testing and validation cohorts. Analyses of native sera HRMAS MRS of the training cohort selected 57 spectral regions for statistical analyses, and PCA and canonical analysis were conducted to differentiate TOD from Ctrl groups, with the testing and validation cohorts passively followed the calculations to produce scores within each case.Using the mean plus one standard error (M+SE) as the threshold, calculated from the canonical score differences between TOD and PTD for each case, i.e. the difference of the two scores for each patient, a Kaplan-Meier survival analysis indicates better survival rate from their time of NSCLC diagnoses for patients with score differences higher than the threshold (p=0.031). Furthermore, for Stage I and IIA patients, survivals can be predicted (p=0.027) from their PTD samples, if their score values are higher than the M+SE threshold. Since neither testing nor validation cohorts were involved in the determination of values of the canonical score, by collectively examining all Stage I and IIA in both cohorts, the resulting Kaplan-Meier survival predicting capability by the threshold (established by the testing cohort) was enhanced significantly (p=0.0044). For 5-year survival, our data showed Sen=0.63, Spe=1.00, PPV=1.00, NPV=0.21, ACC=0.67, and F1=0.78. Thus, serum metabolomics may be a screening candidate for early detections with the potential to further probe into all the related metabolic pathways for better understanding of disease. Citation Format: Tjada Schult, Mara Lauer, Yannick Berker, Marcella Cardoso, Lindsey Vandergrift, Piet Habbel, Johannes Nowak, Martin Aryee, Mari Mino-Kenudson, David Christiani, Leo Cheng. Blood serum MRS metabolomics screening of human lung cancer prior to disease diagnosis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr LB209.