Multi-parametric MRI-based radiomics signature for preoperative prediction of Ki-67 proliferation status in sinonasal malignancies: a two-centre study

Development and validation of a prediction model for malignant sinonasal tumors based on MR radiomics and machine learning

OBJECTIVES

This study aimed to utilize MR radiomics-based machine learning classifiers on a large-sample, multicenter dataset to develop an optimal model for predicting malignant sinonasal tumors and tumor-like lesions.

METHODS

This study included 1711 adult patients (875 benign and 836 malignant) with sinonasal tumors or tumor-like lesions from three institutions. Patients from institution 1 (n = 1367) constituted both the training and validation cohorts, while those from institution 2 and 3 (n = 158/186) made up the test cohorts. Manual segmentation of the region of interest of the tumor was performed on T1WI, T2WI, and contrast-enhanced T1WI (CE-T1WI). Data normalization, dimensional reductions, feature selection, and classifications were performed using ten machine-learning classifiers. Four fusion models, namely T1WI + T2WI, T1WI + CE-T1WI, T2WI + CE-T1WI, and T1WI + T2WI + CE-T1WI, were constructed using the top ten features with the highest contribution in feature selection in the optimal models of T1WI, T2WI, and CE-T1WI. The Delong test compared areas under the curve (AUC) between models.

RESULTS

The AUCs of training/validation/test1/test2 datasets for T1WI, T2WI, and CE-T1WI were 0.900/0.842/0.872/0.839, 0.876/0.789/0.842/0.863, and 0.899/0.824/0.831/0.707, respectively. The fusion model from T1WI + T2WI + CE-T1WI had the highest AUC. The AUCs of training/validation/test1/test2 datasets were 0.947/0.849/0.871/0.887. The T1WI + T2WI + CE-T1WI model demonstrated a significantly higher AUC than the T2WI + CE-T1WI model in both cohorts (p < 0.05) and outperformed the T2WI model in test 1 (p = 0.008) and the T1WI model in test 2 (p = 0.006).

CONCLUSIONS

This fusion model based on radiomics from T1WI + T2WI + CE-T1WI images and machine learning can improve the power in predicting malignant sinonasal tumors with high accuracy, resilience, and robustness.

CLINICAL RELEVANCE STATEMENT

Our study proposes a radiomics-based machine learning fusion model from T1- and T2-weighted images and contrast-enhanced T1-weighted images, which can non-invasively identify the nature of sinonasal tumors and improve the performance in predicting malignant sinonasal tumors.

KEY POINTS

Differentiating benign and malignant sinonasal tumors is difficult due to similar clinical presentations. A radiomics model from T1 + T2 + contrast-enhanced T1 images can identify the nature of sinonasal tumors. This model can help distinguish benign and malignant sinonasal tumors.

Sinonasal adenoid cystic carcinoma: preoperative apparent diffusion coefficient histogram analysis in prediction of prognosis and Ki-67 proliferation status

PURPOSE

To investigate the value of preoperative apparent diffusion coefficient (ADC) histogram analysis in predicting the prognosis of patients with sinonasal adenoid cystic carcinoma (ACC) and the correlation between ADC histogram parameters and Ki-67 labeling index (LI).

MATERIALS AND METHODS

The study enrolled 66 patients with sinonasal ACC who were surgically resected and confirmed by histopathology. The disease-free survival (DFS) was evaluated with clinical-pathologic and radiologic characteristics using the Cox proportion hazard model. Spearman correlation analysis was used to evaluate the correlation between ADC histogram parameters and Ki-67 LI. The predictive performance of ADC histogram parameters for Ki-67 LI was assessed using the receiver operating characteristic (ROC) curve.

RESULTS

Multivariable analysis showed Ki-67 LI (hazard ratio: 9.279; 95% confidence interval 1.099-78.338; P = 0.041) and ADCskewness (hazard ratio: 5.942; 95% confidence interval 1.832-19.268; P = 0.003) were significant independent predictors of DFS. The combination of these two variables achieved the predictive ability with a C-index of 0.717 (95% confidence interval 0.607-0.826). ADCmean and all ADC percentiles (10th, 50th, and 90th) significantly and inversely correlated with Ki-67 LI of ACC (Correlation coefficients = - 0.574 to - 0.591, Ps < 0.001). Among the ADC histogram parameters, the ADC50th showed superior performance for the differentiation of the high from low Ki-67 LI groups with an area under the curve (AUC) of 0.834 and an accuracy of 80.30%.

CONCLUSION

ADC histogram analysis had predictive value for DFS and Ki-67 LI, which may be a valuable biomarker for prognosis and proliferation status for ACC in clinical practice.

Radiomics model building from multiparametric MRI to predict Ki-67 expression in patients with primary central nervous system lymphomas: a multicenter study

OBJECTIVES

To examine the correlation of apparent diffusion coefficient (ADC), diffusion weighted imaging (DWI), and T1 contrast enhanced (T1-CE) with Ki-67 in primary central nervous system lymphomas (PCNSL). And to assess the diagnostic performance of MRI radiomics-based machine-learning algorithms in differentiating the high proliferation and low proliferation groups of PCNSL.

METHODS

83 patients with PCNSL were included in this retrospective study. ADC, DWI and T1-CE sequences were collected and their correlation with Ki-67 was examined using Spearman's correlation analysis. The Kaplan-Meier method and log-rank test were used to compare the survival rates of the high proliferation and low proliferation groups. The radiomics features were extracted respectively, and the features were screened by machine learning algorithm and statistical method. Radiomics models of seven different sequence permutations were constructed. The area under the receiver operating characteristic curve (ROC AUC) was used to evaluate the predictive performance of all models. DeLong test was utilized to compare the differences of models.

RESULTS

Relative mean apparent diffusion coefficient (rADCmean) (ρ=-0.354, p = 0.019), relative mean diffusion weighted imaging (rDWImean) (b = 1000) (ρ = 0.273, p = 0.013) and relative mean T1 contrast enhancement (rT1-CEmean) (ρ = 0.385, p = 0.001) was significantly correlated with Ki-67. Interobserver agreements between the two radiologists were almost perfect for all parameters (rADCmean ICC = 0.978, 95%CI 0.966-0.986; rDWImean (b = 1000) ICC = 0.931, 95% CI 0.895-0.955; rT1-CEmean ICC = 0.969, 95% CI 0.953-0.980). The differences in PFS (p = 0.016) and OS (p = 0.014) between the low and high proliferation groups were statistically significant. The best prediction model in our study used a combination of ADC, DWI, and T1-CE achieving the highest AUC of 0.869, while the second ranked model used ADC and DWI, achieving an AUC of 0.828.

CONCLUSION

rDWImean, rADCmean and rT1-CEmean were correlated with Ki-67. The radiomics model based on MRI sequences combined is promising to distinguish low proliferation PCNSL from high proliferation PCNSL.

Multicenter study on predicting postoperative upper limb muscle strength improvement in cervical spinal cord injury patients using radiomics and deep learning

Cervical spinal cord injury is often catastrophic, frequently leading to irreversible impairment. MRI has become the gold standard for evaluating spinal cord injuries (SCI). Our study aimed to assess the accuracy of a radiomics approach, based on machine learning and utilizing conventional MRI, in predicting the prognosis of patients with SCI. In a retrospective analysis of 82 SCI patients from three hospitals, we categorized them into good (n = 49) and poor (n = 33) prognosis groups. Preoperative T2-weighted MRI images were segmented using 3D-Region of Interest (ROI) techniques, and both radiomic and deep transfer learning features were extracted. These features were normalized using Z-score and harmonized via ComBat. Feature selection was performed using a greedy algorithm and Least absolute shrinkage and selection operator (LASSO), and others, followed by the calculation of radiomics scores through linear regression. Machine learning was then used to identify the most predictive radiomic features. Model performance was evaluated by analyzing the area under the curve (AUC) and other indicators.Univariate analysis indicated that the demographic characteristics of cervical spinal cord injury were not statistically significant. In the test dataset, the random forest (RF) combined with radiomics and ResNet34 demonstrated better performance, with an accuracy of 0.800 and an AUC of 0.893.Using MRI, deep learning-based radiomics signals show promise in evaluating and predicting the postoperative prognosis of these patients.

An MRI-based radiomics nomogram for preoperative prediction of Ki-67 index in nasopharyngeal carcinoma: a two-center study

Background

The expression level of Ki-67 in nasopharyngeal carcinoma (NPC) affects the prognosis and treatment options of patients. Our study developed and validated an MRI-based radiomics nomogram for preoperative evaluation of Ki-67 expression levels in nasopharyngeal carcinoma (NPC).

Methods

In all, 133 patients with pathologically-confirmed (post-operatively) NPC who underwent MRI examination in one of two medical centers. Data from one medical center (n=105; Ki-67: ≥50% [n=57], <50% [n=48]) formed the training set, while data from another medical center (n=28; Ki-67: ≥50% [n=15], <50% [n=13]) formed the test set. Clinical data and routine MRI results were reviewed to determine significant predictive factors. The minimum absolute shrinkage and selection operator method was used to select key radiomics features to form a radiomics signatures from resonance imaging (MRI), and a radiomics score (Rad-score) was calculated. Subsequently, a radiomics nomogram was established using a logistic regression (LR) algorithm. The predictive performance of the nomogram was evaluated using operating characteristics curve (ROC), decision curve analysis (DCA), and the area under the curve (AUC).

Results

Five radiomics features were selected to build the radiomics signature. The radiomics nomogram incorporating the clinical factors and radiomics signature showed favorable predictive value for expression level of Ki-67, with AUC 0.841 (95% confidence intervals: 0.654 -0.951) for the test set. Decision curve analysis showed that the nomogram outperformed a clinical model in terms of clinical usefulness.

Conclusions

The radiomics nomogram based on MRI effectively predicted the pre-surgical expression level of Ki-67.

Prognostic value of FDX1, the cuprotosis key gene, and its prediction models across imaging modalities and histology

BACKGROUND

Cuprotosis has been identified as a novel way of cell death. The key regulator ferredoxin 1 (FDX1) was explored via pan-cancer analysis, and its prediction models were proposed across seven malignancies and two imaging modalities.

METHODS

The prognostic value of FDX1 was explored via 1654 cases of 33 types of cancer in the Cancer Genome Atlas database. The MRI cohort of hepatocellular carcinoma in the First Affiliated Hospital of Fujian Medical University, and CT and MRI images from the Cancer Imaging Archive, REMBRANDT and Duke databases were exploited to formulate radiomic models to predict FDX1 expression. After segmentation of volumes of interest and feature extraction, the recursive feature elimination algorithm was used to screen features, logistic regression was used to model features, immunohistochemistry staining with FDX1 antibody was performed to test the radiomic model.

RESULTS

FDX1 was found to be prognostic in various types of cancer. The area under the receiver operating characteristic curve of radiomic models to predict FDX1 expression reached 0.825 (95% CI = 0.739-0.911). Cross-tissue compatibility was confirmed in pan-cancer validation and test cohorts. Mechanistically, the radiomic score was significantly correlated with various immunosuppressive genes and gene mutations. The radiomic score was also found to be an independent prognostic factor, making it a potentially actionable biomarker in the clinical setting.

CONCLUSIONS

The expression of FDX1 could be non-invasively predicted via radiomics. The radiomic patterns with biological and clinical relevance across histology and modalities could have a broad impact on a larger population of patients.

MRI-based radiomics signatures for preoperative prediction of Ki-67 index in primary central nervous system lymphoma

PURPOSE

The aim of this study was to develop and validate radiomics signatures based on MRI for preoperative prediction of Ki-67 proliferative index (PI) expression in primary central nervous system lymphoma (PCNSL).

METHODS

A total of 341 patients with PCNSL were retrospectively analyzed, including 286 patients in one center as the training set and 55 patients in another two centers as the external validation set. Radiomics features were extracted and selected from preoperative contrast-enhanced T1-weighted images, fluid attenuation inversion recovery to build radiomics signatures according to the Ki-67 PI. The predictive performances of the radiomics model were evaluated using four classifiers including random forest, K-Nearest Neighbors, Neural Network and Decision Tree. A combined model was built by incorporating radiomics signature, clinical variables and MRI radiological characteristics using multivariate logistic regression analysis, and a nomogram was established to predict the expression of Ki-67 individually. The predictive performances of the models were evaluated using area under receiver operating characteristic curve (AUC) and decision curve analysis (DCA).

RESULTS

Radiomics signatures were independent predictors of the expression level of Ki-67 (OR: 2.523, P < 0.001). RF radiomics models had the highest accuracy (0.934 in the training set and 0.811 in the external validation set) and F1 Score (0.920 in the training set and 0.836 in the external validation set). The clinic-radiologic-radiomics nomogram showed better predictive performance with AUCs of 0.877(95 % CI: 0.837-0.918) in the training set and 0.866(95 % CI: 0.774-0.957) in the external validation set. The calibration curve and DCA demonstrated goodness-of-fit and improved benefits in clinical practice of the nomogram.

CONCLUSIONS

Nomograms integrating MRI-based radiomics and clinical-radiological characteristics could effectively predict Ki-67 PI in primary PCNSL.

Radiomics nomogram based on CT radiomics features and clinical factors for prediction of Ki-67 expression and prognosis in clear cell renal cell carcinoma: a two-center study

OBJECTIVES

To develop and validate a radiomics nomogram combining radiomics features and clinical factors for preoperative evaluation of Ki-67 expression status and prognostic prediction in clear cell renal cell carcinoma (ccRCC).

METHODS

Two medical centers of 185 ccRCC patients were included, and each of them formed a training group (n = 130) and a validation group (n = 55). The independent predictor of Ki-67 expression status was identified by univariate and multivariate regression, and radiomics features were extracted from the preoperative CT images. The maximum relevance minimum redundancy (mRMR) and the least absolute shrinkage and selection operator algorithm (LASSO) were used to identify the radiomics features that were most relevant for high Ki-67 expression. Subsequently, clinical model, radiomics signature (RS), and radiomics nomogram were established. The performance for prediction of Ki-67 expression status was validated using area under curve (AUC), calibration curve, Delong test, decision curve analysis (DCA). Prognostic prediction was assessed by survival curve and concordance index (C-index).

RESULTS

Tumour size was the only independent predictor of Ki-67 expression status. Five radiomics features were finally identified to construct the RS (AUC: training group, 0.821; validation group, 0.799). The radiomics nomogram achieved a higher AUC (training group, 0.841; validation group, 0.814) and clinical net benefit. Besides, the radiomics nomogram provided a highest C-index (training group, 0.841; validation group, 0.820) in predicting prognosis for ccRCC patients.

CONCLUSIONS

The radiomics nomogram can accurately predict the Ki-67 expression status and exhibit a great capacity for prognostic prediction in patients with ccRCC and may provide value for tailoring personalized treatment strategies and facilitating comprehensive clinical monitoring for ccRCC patients.

Multiparametric MRI-based radiomics approach with deep transfer learning for preoperative prediction of Ki-67 status in sinonasal squamous cell carcinoma

Purpose

Based on comparison of different machine learning (ML) models, we developed the model that integrates traditional hand-crafted (HC) features and ResNet50 network-based deep transfer learning (DTL) features from multiparametric MRI to predict Ki-67 status in sinonasal squamous cell carcinoma (SNSCC).

Methods

Two hundred thirty-one SNSCC patients were retrospectively reviewed [training cohort (n = 185), test cohort (n = 46)]. Pathological grade, clinical, and MRI characteristics were analyzed to choose the independent predictor. HC and DTL radiomics features were extracted from fat-saturated T2-weighted imaging, contrast-enhanced T1-weighted imaging, and apparent diffusion coefficient map. Then, HC and DTL features were fused to formulate the deep learning-based radiomics (DLR) features. After feature selection and radiomics signature (RS) building, we compared the predictive ability of RS-HC, RS-DTL, and RS-DLR.

Results

No independent predictors were found based on pathological, clinical, and MRI characteristics. After feature selection, 42 HC and 10 DTL radiomics features were retained. The support vector machine (SVM), LightGBM, and ExtraTrees (ET) were the best classifier for RS-HC, RS-DTL, and RS-DLR. In the training cohort, the predictive ability of RS-DLR was significantly better than those of RS-DTL and RS-HC (p< 0.050); in the test set, the area under curve (AUC) of RS-DLR (AUC = 0.817) was also the highest, but there was no significant difference of the performance between DLR-RS and HC-RS.

Conclusions

Both the HC and DLR model showed favorable predictive efficacy for Ki-67 expression in patients with SNSCC. Especially, the RS-DLR model represented an opportunity to advance the prediction ability.

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.

Prediction of the Ki-67 expression level in head and neck squamous cell carcinoma with machine learning-based multiparametric MRI radiomics: a multicenter study

BACKGROUND

This study aimed to develop and validate a machine learning (ML)-based fusion model to preoperatively predict Ki-67 expression levels in patients with head and neck squamous cell carcinoma (HNSCC) using multiparametric magnetic resonance imaging (MRI).

METHODS

A total of 351 patients with pathologically proven HNSCC from two medical centers were retrospectively enrolled in the study and divided into training (n = 196), internal validation (n = 84), and external validation (n = 71) cohorts. Radiomics features were extracted from T2-weighted images and contrast-enhanced T1-weighted images and screened. Seven ML classifiers, including k-nearest neighbors (KNN), support vector machine (SVM), logistic regression (LR), random forest (RF), linear discriminant analysis (LDA), naive Bayes (NB), and eXtreme Gradient Boosting (XGBoost) were trained. The best classifier was used to calculate radiomics (Rad)-scores and combine clinical factors to construct a fusion model. Performance was evaluated based on calibration, discrimination, reclassification, and clinical utility.

RESULTS

Thirteen features combining multiparametric MRI were finally selected. The SVM classifier showed the best performance, with the highest average area under the curve (AUC) of 0.851 in the validation cohorts. The fusion model incorporating SVM-based Rad-scores with clinical T stage and MR-reported lymph node status achieved encouraging predictive performance in the training (AUC = 0.916), internal validation (AUC = 0.903), and external validation (AUC = 0.885) cohorts. Furthermore, the fusion model showed better clinical benefit and higher classification accuracy than the clinical model.

CONCLUSIONS

The ML-based fusion model based on multiparametric MRI exhibited promise for predicting Ki-67 expression levels in HNSCC patients, which might be helpful for prognosis evaluation and clinical decision-making.

Reproducibility of radiomics quality score: an intra- and inter-rater reliability study

OBJECTIVES

To investigate the intra- and inter-rater reliability of the total radiomics quality score (RQS) and the reproducibility of individual RQS items' score in a large multireader study.

METHODS

Nine raters with different backgrounds were randomly assigned to three groups based on their proficiency with RQS utilization: Groups 1 and 2 represented the inter-rater reliability groups with or without prior training in RQS, respectively; group 3 represented the intra-rater reliability group. Thirty-three original research papers on radiomics were evaluated by raters of groups 1 and 2. Of the 33 papers, 17 were evaluated twice with an interval of 1 month by raters of group 3. Intraclass coefficient (ICC) for continuous variables, and Fleiss' and Cohen's kappa (k) statistics for categorical variables were used.

RESULTS

The inter-rater reliability was poor to moderate for total RQS (ICC 0.30-055, p < 0.001) and very low to good for item's reproducibility (k - 0.12 to 0.75) within groups 1 and 2 for both inexperienced and experienced raters. The intra-rater reliability for total RQS was moderate for the less experienced rater (ICC 0.522, p = 0.009), whereas experienced raters showed excellent intra-rater reliability (ICC 0.91-0.99, p < 0.001) between the first and second read. Intra-rater reliability on RQS items' score reproducibility was higher and most of the items had moderate to good intra-rater reliability (k - 0.40 to 1).

CONCLUSIONS

Reproducibility of the total RQS and the score of individual RQS items is low. There is a need for a robust and reproducible assessment method to assess the quality of radiomics research.

CLINICAL RELEVANCE STATEMENT

There is a need for reproducible scoring systems to improve quality of radiomics research and consecutively close the translational gap between research and clinical implementation.

KEY POINTS

• Radiomics quality score has been widely used for the evaluation of radiomics studies. • Although the intra-rater reliability was moderate to excellent, intra- and inter-rater reliability of total score and point-by-point scores were low with radiomics quality score. • A robust, easy-to-use scoring system is needed for the evaluation of radiomics research.

Contrast-enhanced CT radiomics for preoperative prediction of stage in epithelial ovarian cancer: a multicenter study

BACKGROUND

Preoperative prediction of International Federation of Gynecology and Obstetrics (FIGO) stage in patients with epithelial ovarian cancer (EOC) is crucial for determining appropriate treatment strategy. This study aimed to explore the value of contrast-enhanced CT (CECT) radiomics in predicting preoperative FIGO staging of EOC, and to validate the stability of the model through an independent external dataset.

METHODS

A total of 201 EOC patients from three centers, divided into a training cohort (n = 106), internal (n = 46) and external (n = 49) validation cohorts. The least absolute shrinkage and selection operator (LASSO) regression algorithm was used for screening radiomics features. Five machine learning algorithms, namely logistic regression, support vector machine, random forest, light gradient boosting machine (LightGBM), and decision tree, were utilized in developing the radiomics model. The optimal performing algorithm was selected to establish the radiomics model, clinical model, and the combined model. The diagnostic performances of the models were evaluated through receiver operating characteristic analysis, and the comparison of the area under curves (AUCs) were conducted using the Delong test or F-test.

RESULTS

Seven optimal radiomics features were retained by the LASSO algorithm. The five radiomics models demonstrate that the LightGBM model exhibits notable prediction efficiency and robustness, as evidenced by AUCs of 0.83 in the training cohort, 0.80 in the internal validation cohort, and 0.68 in the external validation cohort. The multivariate logistic regression analysis indicated that carcinoma antigen 125 and tumor location were identified as independent predictors for the FIGO staging of EOC. The combined model exhibited best diagnostic efficiency, with AUCs of 0.95 in the training cohort, 0.83 in the internal validation cohort, and 0.79 in the external validation cohort. The F-test indicated that the combined model exhibited a significantly superior AUC value compared to the radiomics model in the training cohort (P < 0.001).

CONCLUSIONS

The combined model integrating clinical characteristics and radiomics features shows potential as a non-invasive adjunctive diagnostic modality for preoperative evaluation of the FIGO staging status of EOC, thereby facilitating clinical decision-making and enhancing patient outcomes.

MRI-based radiomics nomogram for distinguishing solitary fibrous tumor from schwannoma in the orbit: a two-center study

OBJECTIVES

To investigate the value of magnetic resonance imaging (MRI) radiomics for distinguishing solitary fibrous tumor (SFT) from schwannoma in the orbit.

MATERIALS AND METHODS

A total of 140 patients from two institutions were retrospectively included. All patients from institution 1 were randomized into a training cohort (n = 69) and a validation cohort (n = 35), and patients from institution 2 were used as an external testing cohort (n = 36). One hundred and six features were extracted from T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted imaging (CET1WI). A radiomics model was built for each sequence using least absolute shrinkage and selection operator logistic regression, and radiomics scores were calculated. A combined model was constructed and displayed as a radiomics nomogram. Two radiologists jointly assessed tumor category based on MRI findings. The performances of the radiomics models and visual assessment were compared via area under the curve (AUC).

RESULTS

The performances of the radiomics nomogram combining T2WI and CET1WI radiomics scores were superior to those of the pooled readers in the training (AUC 0.986 vs. 0.807, p < 0.001), validation (AUC 0.989 vs. 0.788, p = 0.009), and the testing (AUC 0.903 vs. 0.792, p = 0.093), although significant difference was not found in the testing cohort. Decision curve analysis demonstrated that the radiomics nomogram had better clinical utility than visual assessment.

CONCLUSION

MRI radiomics nomogram can be used for distinguishing between orbital SFT and schwannoma, which may help tumor management by clinicians.

CLINICAL RELEVANCE STATEMENT

It is of great importance and challenging for distinguishing solitary fibrous tumor from schwannoma in the orbit. In the present study, an MRI-based radiomics nomogram were developed and independently validated, which could help the discrimination of the two entities.

KEY POINTS

• It is challenging to differentiate solitary fibrous tumor from schwannoma in the orbit due to similar clinical and image features. • A radiomics nomogram based on T2-weighted imaging and contrast-enhanced T1-weighted imaging has advantages over radiologists. • Radiomics can provide a non-invasive diagnostic tool for differentiating between the two entities.