Contrast enhanced magnetic resonance imaging-based radiomics nomogram for preoperatively predicting expression status of Ki-67 in meningioma: a two-center study

Abstract

Background: The aim of this study was to develop and validate a radiomics nomogram for preoperative prediction of Ki-67 proliferative index (Ki-67 PI) expression in patients with meningioma.

Methods: A total of 280 patients from 2 independent hospital centers were enrolled. Patients from center I were randomly divided into a training cohort of 168 patients and a test cohort of 72 patients, and 40 patients from center II served as an external validation cohort. Interoperator reproducibility test, Z-score standardization, analysis of variance (ANOVA), and least absolute shrinkage and selection operator (LASSO) binary logistic regression were used to select radiomics features, which were extracted from contrast-enhanced T1-weighted imaging (CE-T1WI) imaging. The radiomics signature for predicting Ki-67 PI expression was developed and validated using 4 classifiers including logistic regression (LR), decision tree (DT), support vector machine (SVM), and adaptive boost (AdaBoost). Finally, combined radiological characteristics with radiomics signature were used to establish the nomogram to predict the risk of high Ki-67 PI expression in patients with meningioma.

Results: Fourteen radiomics features were used to construct the radiomics signature. The radiomics nomogram that incorporated the radiomics signature and radiological characteristics showed excellent discrimination in the training, test, and validation cohorts with areas under the curve of 0.817 (95% CI: 0.753-0.881), 0.822 (95% CI: 0.727-0.916), and 0.845 (95% CI: 0.708-0.982), respectively. In addition, the calibration curve for the nomogram demonstrated good agreement between prediction and actual observation.

Conclusions: The proposed contrast enhanced magnetic resonance imaging (MRI)-based radiomics nomogram could be an effective tool to predict the risk of Ki-67 high expression in patients with meningioma.

Keywords: Ki-67; Meningioma; magnetic resonance imaging; radiomics.

Figure 1

Study flowchart. (A) The ROI of Meningioma was delineated in contrast-enhanced T1-weighted imaging. (B) After image preprocessing, 3 categories of radiomics features were extracted from the ROI. (C) Four selection steps were applied to all extracted radiomics features. (D) A final prediction model was constructed by incorporating both a radiomics and a radiological signature; a receiver operating curve and decision curve were generated for further statistical analyses, and a combined nomogram was adopted to present the outcomes of prediction to achieve clinical usefulness. ROI, region of interest; ICC, intraclass correlation coefficient; LR, logistic regression; DT, decision tree; SVM, support vector machine; AdaBoost, adaptive boost; ANOVA, analysis of variance; ROC, receiver operating characteristic; LASSO, least absolute shrinkage and selection operator; DC, decision curve.

Figure 2

LASSO regression model. (A) Tuning parameter log (λ) selection in the LASSO regression used 10-fold cross-validation via the minimum criterion. Dotted vertical lines were drawn at the optimal values by using the minimum criterion and the 1 standard error of the minimum criterion, and a value λ of 0.063 was chosen. (B) LASSO coefficient profiles, displaying 622 radiomics features of contrast-enhanced T1-weighted imaging. A coefficient profile plot was produced against the log (λ) sequence. Finally, 14 radiomics features with nonzero coefficients were selected. LASSO, least absolute shrinkage and selection operator.

Figure 3

ROCs of the radiology signature (blue line), the radiomics signature (red line), and the combined model (green line) in the training (A), test (B), and validation (C) cohorts, respectively. (D) Decision curve for 3 models in the training cohort. The y-axis indicates the net benefit; the x-axis indicates the threshold probability. For differentiating the high expression of Ki-67 from low expression in meningioma, the combined model (green line) had a higher overall net benefit than did the radiology model (blue line) and simple diagnoses such as all Ki-67≤4% patients (gray line) or all Ki-67>4% patients (black line) across the full range of conceivable threshold probabilities. AUC, area under curve; ROC, receiver operating characteristic curve.

Figure 4

A t-test showed significant differences in the radiomics scores between the Ki-67+ and Ki-67− groups in the training, test, and validation cohorts (training cohort: 0.630±1.097, vs. −0.634±1.175, P<0.001; test cohort: 0.722±1.267 vs. −0.484±0.972, P<0.001; validation cohort: 1.232±2.270 vs. −0.763±1.867, P=0.004); considering all the data, the radiomics scores were significantly higher in the Ki-67+ group than in the Ki-67− group; **, P<0.01; ***, P<0.001.

Figure 5

Development and validation of the nomogram. (A) The radiomics nomogram based on the training cohort. Calibration curves for the radiomics nomogram in the training (B), test (C), and validation (D) cohorts. The solid line represents the ideal reference line that predicts Ki-67 expression status and corresponds to the actual outcome, the short-dashed line represents the apparent prediction of the nomogram, and the 45° diagonal dotted line represents an ideal evaluation. All 3 calibration curves showed similar trends, and the actual probability corresponded closely to the predictive probability of the radiomics nomogram.

Figure 6

Examples of using the nomogram to predict the individual risk of high Ki-67 expression in meningioma by manually placing straight lines across the diagram. (A-C) A 70-year-old female with cerebellum meningioma. CE-T1WI showed a tumor with less homogeneous enhancement (green line on the “Points” scale =12 points); the radiomics score was –2.7 (red line on the “Points” scale =5 points). The above 2 “Points” were added to obtain total points (12+5=17 points). The graph revealed that the risk of Ki-67 high expression turned out to be about 8% (low risk) via the drawing of a black line on the “Total points” scale. (B) Postoperative immunohistochemistry proved the Ki-67 PI of this patient was less than 1% (HE, ×100). (D-F) A 50-year-old female with right frontal meningioma. The tumor exhibits homogeneous enhancement in CE-T1WI (green line on the “Points” scale =0 point); the radiomics score was 2.6 (red line on the “Points” scale =75 points). A total of 75 points was obtained. The risk of high Ki-67 expression was about 88% (extremely high risk). (E) Postoperative immunohistochemistry proved the Ki-67 PI of this patient was about 5% (HE, ×100); CE-T1WI, contrast-enhanced T1-weighted imaging; HE, hematoxylin and eosin staining.