A nomogram based on dual-layer detector spectral computed tomography quantitative parameters and morphological quantitative indicator for distinguishing metastatic and nonmetastatic regional lymph nodes in pancreatic ductal adenocarcinoma.

Background: There is no unified scope for regional lymph node (LN) dissection in patients with pancreatic ductal adenocarcinoma (PDAC). Incomplete regional LN dissection can lead to postoperative recurrence, while blind expansion of the scope of regional LN dissection significantly increases the perioperative risk without significantly prolonging overall survival. We aimed to establish a noninvasive visualization tool based on dual-layer detector spectral computed tomography (DLCT) to predict the probability of regional LN metastasis in patients with PDAC. Methods: A total of 163 regional LNs were reviewed and divided into a metastatic cohort (n=58 LNs) and nonmetastatic cohort (n=105 LNs). The DLCT quantitative parameters and the nodal ratio of the longest axis to the shortest axis (L/S) of the regional LNs were compared between the two cohorts. The DLCT quantitative parameters included the iodine concentration in the arterial phase (APIC), normalized iodine concentration in the arterial phase (APNIC), effective atomic number in the arterial phase (APZeff), normalized effective atomic number in the arterial phase (APNZeff), slope of the spectral attenuation curves in the arterial phase (APλHU), iodine concentration in the portal venous phase (PVPIC), normalized iodine concentration in the portal venous phase (PVPNIC), effective atomic number in the portal venous phase (PVPZeff), normalized effective atomic number in the portal venous phase (PVPNZeff), and slope of the spectral attenuation curves in the portal venous phase (PVPλHU). Logistic regression analysis based on area under the curve (AUC) was used to analyze the diagnostic performance of significant DLCT quantitative parameters, L/S, and the models combining significant DLCT quantitative parameters and L/S. A nomogram based on the models with highest diagnostic performance was developed as a predictor. The goodness of fit and clinical applicability of the nomogram were assessed through calibration curve and decision curve analysis (DCA). Results: The combined model of APNIC + L/S (APNIC + L/S) had the highest diagnostic performance among all models, yielding an AUC, sensitivity, and specificity of 0.878 [95% confidence interval (CI): 0.825-0.931], 0.707, and 0.886, respectively. The calibration curve indicated that the APNIC-L/S nomogram had good agreement between the predicted probability and the actual probability. Meanwhile, the decision curve indicated that the APNIC-L/S nomogram could produce a greater net benefit than could the all- or-no-intervention strategy, with threshold probabilities ranging from 0.0 to 0.75. Conclusions: As a valid and visual noninvasive prediction tool, the APNIC-L/S nomogram demonstrated favorable predictive efficacy for identifying metastatic LNs in patients with PDAC.

Differentiation of benign and malignant lesions in Bethesda III and IV thyroid nodules via dual-energy computed tomography quantitative parameters and morphologic features.

Background: Thyroid nodules (TNs) cytologically defined as category Bethesda III and IV pose a major diagnostic challenge before surgery, demanding new methods to reduce unnecessary diagnostic thyroid lobectomies for patients with benign TNs. This study aimed to assess whether a model combining dual-energy computed tomography (DECT) quantitative parameters with morphologic features could reliably differentiate between benign and malignant lesions in Bethesda III and IV TNs. Methods: Data from 77 patients scheduled for thyroid surgery for Bethesda III and IV TNs (malignant =48; benign =29) who underwent DECT scans were reviewed. DECT quantitative parameters including normalized iodine concentration (NIC), attenuation on the slope of spectral Hounsfield unit (HU) curve, and normalized effective atomic number (Zeff) were measured in the arterial phase (AP) and venous phase (VP). DECT quantitative parameters and morphologic features were compared between the malignant and benign cohorts. The receiver operating characteristic curve was performed to compare the performances of significant DECT quantitative parameters, morphologic features, or the models combining the DECT parameters, respectively, with morphologic features. A nomogram was constructed from the optimal performance model, and the performance was evaluated via the calibration curve and decision curve analysis. Results: The areas under the receiver operating characteristic curve with 95% confidence interval (CI) of the NIC in the AP (AP-NIC), slope of spectral HU curve in the AP, and NZeff in the AP were 0.749 (95% CI: 0.641-0.857), 0.654 (95% CI: 0.530-0.778), and 0.722 (95% CI: 0.602-0.842), respectively. The model combining AP-NIC with enhanced blurring showed the highest diagnostic performance, with an area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of 0.808, 0.854, and 0.655, respectively; it was then used to construct a nomogram. The calibration curve showed that the discrepancy between the prediction of the nomogram and actual observations was less than 5%. The decision curve analysis indicated the nomogram had a positive net benefit in threshold risk ranges of 14% to 58% or 60% to 91% for malignant Bethesda III and IV TNs. Conclusions: The model combining AP-NIC with enhanced blurring could reliably differentiate between benign and malignant lesions in Bethesda III and IV TNs.

Detection of malignant lesions in cytologically indeterminate thyroid nodules using a dual-layer spectral detector CT-clinical nomogram.

Purpose: To evaluate the capability of dual-layer detector spectral CT (DLCT) quantitative parameters in conjunction with clinical variables to detect malignant lesions in cytologically indeterminate thyroid nodules (TNs). Materials and methods: Data from 107 patients with cytologically indeterminate TNs who underwent DLCT scans were retrospectively reviewed and randomly divided into training and validation sets (7:3 ratio). DLCT quantitative parameters (iodine concentration (IC), NICP (IC nodule/IC thyroid parenchyma), NICA (IC nodule/IC ipsilateral carotid artery), attenuation on the slope of spectral HU curve and effective atomic number), along with clinical variables, were compared between benign and malignant cohorts through univariate analysis. Multivariable logistic regression analysis was employed to identify independent predictors which were used to construct the clinical model, DLCT model, and combined model. A nomogram was formulated based on optimal performing model, and its performance was assessed using receiver operating characteristic curve, calibration curve, and decision curve analysis. The nomogram was subsequently tested in the validation set. Results: Independent predictors associated with malignant TNs with indeterminate cytology included NICP in the arterial phase, Hashimoto's Thyroiditis (HT), and BRAF V600E (all p < 0.05). The DLCT-clinical nomogram, incorporating the aforementioned variables, exhibited superior performance than the clinical model or DLCT model in both training set (AUC: 0.875 vs 0.792 vs 0.824) and validation set (AUC: 0.874 vs 0.792 vs 0.779). The DLCT-clinical nomogram demonstrated satisfactory calibration and clinical utility in both training set and validation set. Conclusion: The DLCT-clinical nomogram emerges as an effective tool to detect malignant lesions in cytologically indeterminate TNs.

Development and validation of a model for predicting the expression of Ki-67 in pancreatic ductal adenocarcinoma with radiological features and dual-energy computed tomography quantitative parameters.

OBJECTIVE: To construct and validate a model based on the dual-energy computed tomography (DECT) quantitative parameters and radiological features to predict Ki-67 expression levels in pancreatic ductal adenocarcinoma (PDAC). MATERIALS AND METHODS: Data from 143 PDAC patients were analysed. The variables of clinic, radiology and DECT were evaluated. In the arterial phase and portal venous phase (PVP), the normalized iodine concentration (NIC), normalized effective atomic number and slope of the spectral attenuation curves were measured. The extracellular volume fraction (ECVf) was measured in the equilibrium phase. Univariate analysis was used to screen independent risk factors to predict Ki-67 expression. The Radiology, DECT and DECT-Radiology models were constructed, and their diagnostic effectiveness and clinical applicability were obtained through area under the curve (AUC) and decision curve analysis, respectively. The nomogram was established based on the optimal model, and its goodness-of-fit was assessed by a calibration curve. RESULTS: Computed tomography reported regional lymph node status, NIC of PVP, and ECVf were independent predictors for Ki-67 expression prediction. The AUCs of the Radiology, DECT, and DECT-Radiology models were 0.705, 0.884, and 0.905, respectively, in the training cohort, and 0.669, 0.835, and 0.865, respectively, in the validation cohort. The DECT-Radiology nomogram was established based on the DECT-Radiology model, which showed the highest net benefit and satisfactory consistency. CONCLUSIONS: The DECT-Radiology model shows favourable predictive efficacy for Ki-67 expression, which may be of value for clinical decision-making in PDAC patients. CRITICAL RELEVANCE STATEMENT: The DECT-Radiology model could contribute to the preoperative and non-invasive assessment of Ki-67 expression of PDAC, which may help clinicians to screen out PDAC patients with high Ki-67 expression. KEY POINTS: • Dual-energy computed tomography (DECT) can predict Ki-67 in pancreatic ductal adenocarcinoma (PDAC). • The DECT-Radiology model facilitates preoperative and non-invasive assessment of PDAC Ki-67 expression. • The nomogram may help screen out PDAC patients with high Ki-67 expression.

Development of a CT radiomics nomogram for preoperative prediction of Ki-67 index in pancreatic ductal adenocarcinoma: a two-center retrospective study.

OBJECTIVES: To develop and validate a contrast-enhanced computed tomography (CECT)-based radiomics nomogram for the preoperative evaluation of Ki-67 proliferation status in pancreatic ductal adenocarcinoma (PDAC). METHODS: In this two-center retrospective study, a total of 181 patients (95 in the training cohort; 42 in the testing cohort, and 44 in the external validation cohort) with PDAC who underwent CECT examination were included. Radiomic features were extracted from portal venous phase images. The radiomics signatures were built by using two feature-selecting methods (relief and recursive feature elimination) and four classifiers (support vector machine, naive Bayes, linear discriminant analysis (LDA), and logistic regression (LR)). Multivariate LR was used to build a clinical model and radiomics-clinical nomogram. The predictive performances of the models were evaluated using area under receiver operating characteristic curve (AUC) and decision curve analysis (DCA). RESULTS: The relief selector and LDA classifier using twelve features built the optimal radiomics signature, with AUCs of 0.948, 0.927, and 0.824 in the training, testing, and external validation cohorts, respectively. The radiomics-clinical nomogram incorporating the optimal radiomics signature, CT-reported lymph node status, and CA19-9 showed better predictive performance with AUCs of 0.976, 0.955, and 0.882 in the training, testing, and external validation cohorts, respectively. The calibration curve and DCA demonstrated goodness-of-fit and improved benefits in clinical practice of the nomogram. CONCLUSIONS: The radiomics-clinical nomogram is an effective and non-invasive computer-aided tool to predict the Ki-67 expression status in patients with PDAC. CLINICAL RELEVANCE STATEMENT: The radiomics-clinical nomogram is an effective and non-invasive computer-aided tool to predict the Ki-67 expression status in patients with pancreatic ductal adenocarcinoma. KEY POINTS: The radiomics analysis could be helpful to predict Ki-67 expression status in patients with pancreatic ductal adenocarcinoma (PDAC). The radiomics-clinical nomogram integrated with the radiomics signature, clinical data, and CT radiological features could significantly improve the differential diagnosis of Ki-67 expression status. The radiomics-clinical nomogram showed satisfactory calibration and net benefit for discriminating high and low Ki-67 expression status in PDAC.

Diagnostic accuracy of dual-energy computed tomography-based nomogram for differentiating papillary thyroid microcarcinomas from micronodular goiters.

Background: The misdiagnosis of papillary thyroid microcarcinoma (PTMC) and micronodular goiter (MNG) may lead to overtreatment and unnecessary medical expenditure by patients. This study developed and validated a dual-energy computed tomography (DECT)-based nomogram for the preoperative differential diagnosis of PTMC and MNG. Methods: This retrospective study analyzed the data of 366 pathologically confirmed thyroid micronodules, of which 183 were PTMCs and 183 were MNGs, from 326 patients who underwent DECT examinations. The cohort was divided into the training (n=256) and validation cohorts (n=110). The conventional radiological features and DECT quantitative parameters were analyzed. The iodine concentration (IC), normalized iodine concentration (NIC), effective atomic number, normalized effective atomic number, and slope of the spectral attenuation curves in the arterial phase (AP) and venous phase (VP) were measured. A univariate analysis and stepwise logistic regression analysis were performed to screen the independent indicators for PTMC. A radiological model, DECT model, and DECT-radiological nomogram were constructed, and the performances of the 3 models were assessed using the receiver operating characteristic curve, DeLong test, and a decision curve analysis (DCA). Results: The IC in the AP [odds ratio (OR) =0.172], NIC in the AP (OR =0.003), punctate calcification (OR =2.163), and enhanced blurring (OR =3.188) were identified as independent predictors in the stepwise-logistic regression. The areas under the curve with 95% confidence intervals (CIs) of the radiological model, DECT model, and DECT-radiological nomogram were 0.661 (95% CI: 0.595-0.728), 0.856 (95% CI: 0.810-0.902), and 0.880 (95% CI: 0.839-0.921), respectively, in the training cohort; and 0.701 (95% CI: 0.601-0.800), 0.791 (95% CI: 0.704-0.877), and 0.836 (95% CI: 0.760-0.911), respectively, in the validation cohort. The diagnostic performance of the DECT-radiological nomogram was better than that of the radiological model (P<0.05). The DECT-radiological nomogram was found to be well calibrated and had a good net benefit. Conclusions: DECT provides valuable information for differentiating between PTMC and MNG. The DECT-radiological nomogram could serve as an easy-to-use, noninvasive, and effective method for differentiating between PTMC and MNG and help clinicians in decision-making.

Nomogram based on spectral CT quantitative parameters and typical radiological features for distinguishing benign from malignant thyroid micro-nodules.

PURPOSE: To analyse the predictive effect of a nomogram combining dual-layer spectral computed tomography (DSCT) quantitative parameters with typical radiological features in distinguishing benign micro-nodule from thyroid microcarcinoma (TMC). METHODS: Data from 342 instances with thyroid micro-nodules (≤1 cm) who underwent DSCT (benign group: n = 170; malignant group: n = 172) were reviewed. Typical radiological features including micro-calcification and enhanced blurring, and DSCT quantitative parameters including attenuation on virtual monoenergetic images (40 keV, 70 keV and 100 keV), the slope of the spectral HU curve (λHU), normalized iodine concentration (NIC), and normalized effective atomic number (NZeff) in the arterial phase (AP) and venous phase (VP), were measured and compared between the benign and malignant groups. The receiver operating characteristic (ROC) curve was used to assess the diagnostic performance of significant quantitative DSCT parameters or the models combining DSCT parameters respectively and typical radiological features based on multivariate logistic regression (LR) analysis. A nomogram was developed using predictors with the highest diagnostic performance in the above model, as determined by multivariate LR analysis. RESULTS: The DSCT parameter APλHU showed the greatest diagnostic efficiency in identifying patients with TMC, with an area under the ROC curve (AUC) of 0.829, a sensitivity and specificity of 0.738 and 0.753, respectively. Then, APλHU was combined with the two radiological features to construct the DSCT-Radiological nomogram, which had an AUC of 0.858, a sensitivity of 0.791 and a specificity of 0.800. The calibration curve of the nomogram demonstrated that the prediction result was in good agreement with the actual observation. The decision curve revealed that the nomogram can result in a greater net benefit than the all/none-intervention strategy for all threshold probabilities. CONCLUSION: As a valid and visual noninvasive prediction tool, the DSCT-Radiological nomogram incorporating DSCT quantitative parameters and radiological features shows favourable predictive efficiency for identifying benign and malignant thyroid micro-nodules.

The application of 3-dimensional magnetic resonance imaging in nasopharyngeal carcinoma with pterygopalatine fossa invasion.

OBJECT: The pterygopalatine fossa (PPF) is a covert neurovascular pathway in the skull base and connects with numerous intracranial and extracranial spaces. The aim of this study was to explore the magnetic resonance imaging (MRI) features of PPF invasion in patients with nasopharyngeal carcinoma (NPC). MATERIAL AND METHODS: The medical records of 88 patients with stage T3 or T4 NPC were retrospectively analyzed. The 3-Dimensional (3D) volumetric images of MRI were reconstructed for the tiny connecting conduits of the invaded PPFs in the NPC patients. The infiltration incidence of conduits and connected further structures were calculated. RESULTS: Forty-six PPFs from 37 patients were invaded by NPC. The proportions of stage T4 NPC and intracranial extension were higher in patients with PPF invasion than that without PPF invasion (P < 0.05). Each connecting conduit of the PPF had corresponding optimal reconstructed orientation based on 3D volumetric MRI images. The first three most common infiltrated conduits were palatovaginal canal, vidian canal and sphenopalatine foramen, which were adjacent to the nasopharynx. Among the conduits connecting with further structures, the most common infiltrated conduit was pterygomaxillary fissure, followed by foramen rotundum and inferior orbital fissure. Furthermore, The NPC lesions involved stage T4 structures via the conduits from 19.6% of the invaded PPFs. CONCLUSIONS: The application of high-quality reconstruction images based on 3D sequence of MRI in NPC patients proved to be feasible and beneficial for the manifestation of the invaded PPFs and connecting conduits.

Feasibility of a CT-based lymph node radiomics nomogram in detecting lymph node metastasis in PDAC patients.

Objectives: To investigate the potential value of a contrast enhanced computed tomography (CECT)-based radiological-radiomics nomogram combining a lymph node (LN) radiomics signature and LNs' radiological features for preoperative detection of LN metastasis in patients with pancreatic ductal adenocarcinoma (PDAC). Materials and methods: In this retrospective study, 196 LNs in 61 PDAC patients were enrolled and divided into the training (137 LNs) and validation (59 LNs) cohorts. Radiomic features were extracted from portal venous phase images of LNs. The least absolute shrinkage and selection operator (LASSO) regression algorithm with 10-fold cross-validation was used to select optimal features to determine the radiomics score (Rad-score). The radiological-radiomics nomogram was developed by using significant predictors of LN metastasis by multivariate logistic regression (LR) analysis in the training cohort and validated in the validation cohort independently. Its diagnostic performance was assessed by receiver operating characteristic curve (ROC), decision curve (DCA) and calibration curve analyses. Results: The radiological model, including LN size, and margin and enhancement pattern (three significant predictors), exhibited areas under the curves (AUCs) of 0.831 and 0.756 in the training and validation cohorts, respectively. Nine radiomic features were used to construct a radiomics model, which showed AUCs of 0.879 and 0.804 in the training and validation cohorts, respectively. The radiological-radiomics nomogram, which incorporated the LN Rad-score and the three LNs' radiological features, performed better than the Rad-score and radiological models individually, with AUCs of 0.937 and 0.851 in the training and validation cohorts, respectively. Calibration curve analysis and DCA revealed that the radiological-radiomics nomogram showed satisfactory consistency and the highest net benefit for preoperative diagnosis of LN metastasis. Conclusions: The CT-based LN radiological-radiomics nomogram may serve as a valid and convenient computer-aided tool for personalized risk assessment of LN metastasis and help clinicians make appropriate clinical decisions for PADC patients.

Corrigendum: Location-Specific Radiomics Score: Novel Imaging Marker for Predicting Poor Outcome of Deep and Lobar Spontaneous Intracerebral Hemorrhage.

[This corrects the article DOI: 10.3389/fnins.2021.766228.].

A Novel CT-based Radiomics-Clinical Nomogram for the Prediction of Short-Term Prognosis in Deep Intracerebral Hemorrhage.

OBJECTIVE: To develop and validate a radiomics-clinical nomogram for the prediction of short-term prognosis in patients with deep intracerebral hemorrhage (DICH) on admission. METHODS: A total of 326 patients with DICH (development cohort = 187; testing cohort = 81; validation cohort = 58) were retrospectively included. Radiomics features were extracted from computed tomography (CT) images and optimal features were selected using least absolute shrinkage and selection operator regression. A radiomics score (R-score) was developed using the optimal features. Univariate and multivariate analyses were used to determine independent risk factors for poor outcomes at 30 days. A radiomics-clinical (R-C) nomogram was developed and validated in the three cohorts. Receiver operating characteristic curve (ROC), calibration curve and decision curve analyses were conducted to evaluate the performances of the R-C nomogram. RESULTS: Only 4 of 396 radiomics features were selected to develop R-scores. Age, onset-to-CT time, Glasgow Coma Scale score, midline shift, and R-score were detected as independent predictors of poor prognosis of DICH. The R-C nomogram was developed by the independent predictors and showed acceptable discrimination with areas under ROCs of 0.80, 0.79, and 0.70 in the development, testing and validation cohorts, respectively. The R-C nomogram showed good agreement between the predicted probability and the actual probability (all P > 0.05) and clinical applicability in each cohort. CONCLUSIONS: The R-C nomogram is a stable and effective tool for predicting the short-term prognosis of DICH, which may help clinicians perform individual risk assessments and make decisions for patients with DICH.

Location-Specific Radiomics Score: Novel Imaging Marker for Predicting Poor Outcome of Deep and Lobar Spontaneous Intracerebral Hemorrhage.

Objective: To derive and validate a location-specific radiomics score (Rad-score) based on noncontrast computed tomography for predicting poor deep and lobar spontaneous intracerebral hemorrhage (SICH) outcome. Methods: In total, 494 SICH patients from multiple centers were retrospectively reviewed. Poor outcome was considered mRS 3-6 at 6 months. The Rad-score was derived using optimal radiomics features. The optimal location-specific Rad-score cut-offs for poor deep and lobar SICH outcomes were identified using receiver operating characteristic curve analysis. Univariable and multivariable analyses were used to determine independent poor outcome predictors. The combined models for deep and lobar SICH were constructed using independent predictors of poor outcomes, including dichotomized Rad-score in the derivation cohort, which was validated in the validation cohort. Results: Of 494 SICH patients, 392 (79%) had deep SICH, and 373 (76%) had poor outcomes. The Glasgow Coma Scale score, haematoma enlargement, haematoma location, haematoma volume and Rad-score were independent predictors of poor outcomes (all P < 0.05). Cut-offs of Rad-score, 82.90 (AUC = 0.794) in deep SICH and 80.77 (AUC = 0.823) in lobar SICH, were identified for predicting poor outcomes. For deep SICH, the AUCs of the combined model were 0.856 and 0.831 in the derivation and validation cohorts, respectively. For lobar SICH, the combined model AUCs were 0.866 and 0.843 in the derivation and validation cohorts, respectively. Conclusion: Location-specific Rad-scores and combined models can identify subjects at high risk of poor deep and lobar SICH outcomes, which could improve clinical trial design by screening target patients.

Machine learning-based modified BAT score in predicting hematoma enlargement after spontaneous intracerebral hemorrhage.

OBJECTIVE: The BAT score is an easy-to-use prediction tool to detect hematoma enlargement after spontaneous intracerebral hemorrhage. Machine learning technique has potential predictive gains in accuracy over regression models. We sought to use machine learning technique to improve the BAT score for the prediction of hematoma enlargement. METHODS: Totally 232 patients with spontaneous intracerebral hemorrhage were enrolled from our hospital between 2015 and 2020. The BAT score was calculated to identify high-risk patients with hematoma enlargement. Using the same variables of the original BAT score and 5 common machine learning algorithms, the modified BAT scores were constructed in the training subset (n = 162) and validated in the testing subset (n = 70). Receiver operating characteristic curves were performed to evaluate the discriminative ability of all BAT scores. RESULTS: Among 5 modified BAT scores, the modified BAT score based on Naive Bayes algorithm performed best, with the area under the receiver operating characteristic curve (AUC) of 0.83 in the training subset and 0.77 in the testing subset. The DeLong test showed that the performances of the modified BAT score based on Naive Bayes algorithm were significantly better than that of the BAT score (AUC = 0.57) in the training and testing subsets (both P < 0.001). CONCLUSIONS: Machine learning technique could improve the identification performance of the original BAT score using the same variables. The modified BAT score based on Naive Bayes algorithm could be used as an effective prediction tool for identifying high-risk patients with hematoma enlargement.

A clinical-radiomics nomogram may provide a personalized 90-day functional outcome assessment for spontaneous intracerebral hemorrhage.

OBJECTIVES: To develop and validate a noncontrast computed tomography (NCCT)-based clinical-radiomics nomogram to identify spontaneous intracerebral hemorrhage (sICH) patients with a poor 90-day prognosis on admission. METHODS: In this double-center retrospective study, data from 435 patients with sICH (training cohort: n = 244; internal validation cohort: n = 104; external validation cohort: n = 87) were reviewed. The radiomics score (Rad-score) was calculated based on the coefficients of the selected radiomics features. A clinical-radiomics nomogram was developed by using independent predictors of poor outcome at 90 days through multivariate logistic regression analysis in the training cohort and was validated in the internal and external cohorts. RESULTS: At 90 days, 200 of 435 (46.0%) patients had a poor prognosis. The clinical-radiomics nomogram was developed by six independent predictors namely midline shift, NCCT time from sICH onset, Glasgow Coma Scale score, serum glucose, uric acid, and Rad-score. In identifying patients with poor prognosis, the clinical-radiomics nomogram showed an area under the receiver operating characteristic curve (AUC) of 0.81 in the training cohort, an AUC of 0.78 in the internal validation cohort, and an AUC of 0.73 in the external validation cohort. The calibration curve revealed that the clinical-radiomics nomogram showed satisfactory calibration in the training and internal validation cohorts (both p > 0.05), but slightly poor agreement in the external validation cohort (p < 0.05). CONCLUSIONS: The clinical-radiomics nomogram is a valid computer-aided tool that may provide personalized risk assessment of 90-day functional outcome for sICH patients. KEY POINTS: • The proposed Rad-score was significantly associated with 90-day poor functional outcome in patients with sICH. • The clinical-radiomics nomogram showed satisfactory calibration and the most net benefit for discriminating 90-day poor outcome. • The clinical-radiomics nomogram may provide personalized risk assessment of 90-day functional outcome for sICH patients.

Noncontrast Computed Tomography-Based Radiomics Analysis in Discriminating Early Hematoma Expansion after Spontaneous Intracerebral Hemorrhage.

OBJECTIVE: To determine whether noncontrast computed tomography (NCCT) models based on multivariable, radiomics features, and machine learning (ML) algorithms could further improve the discrimination of early hematoma expansion (HE) in patients with spontaneous intracerebral hemorrhage (sICH). MATERIALS AND METHODS: We retrospectively reviewed 261 patients with sICH who underwent initial NCCT within 6 hours of ictus and follow-up CT within 24 hours after initial NCCT, between April 2011 and March 2019. The clinical characteristics, imaging signs and radiomics features extracted from the initial NCCT images were used to construct models to discriminate early HE. A clinical-radiologic model was constructed using a multivariate logistic regression (LR) analysis. Radiomics models, a radiomics-radiologic model, and a combined model were constructed in the training cohort (n = 182) and independently verified in the validation cohort (n = 79). Receiver operating characteristic analysis and the area under the curve (AUC) were used to evaluate the discriminative power. RESULTS: The AUC of the clinical-radiologic model for discriminating early HE was 0.766. The AUCs of the radiomics model for discriminating early HE built using the LR algorithm in the training and validation cohorts were 0.926 and 0.850, respectively. The AUCs of the radiomics-radiologic model in the training and validation cohorts were 0.946 and 0.867, respectively. The AUCs of the combined model in the training and validation cohorts were 0.960 and 0.867, respectively. CONCLUSION: NCCT models based on multivariable, radiomics features and ML algorithm could improve the discrimination of early HE. The combined model was the best recommended model to identify sICH patients at risk of early HE.