A Radiomics Approach to Assess High Risk Carotid Plaques: A Non-invasive Imaging Biomarker, Retrospective Study

Abstract

Objective: This study aimed to construct a radiomics-based MRI sequence from high-resolution magnetic resonance imaging (HRMRI), combined with clinical high-risk factors for non-invasive differentiation of the plaque of symptomatic patients from asyptomatic patients.

Methods: A total of 115 patients were retrospectively recruited. HRMRI was performed, and patients were diagnosed with symptomatic plaques (SPs) and asymptomatic plaques (ASPs). Patients were randomly divided into training and test groups in the ratio of 7:3. T2WI was used for segmentation and extraction of the texture features. Max-Relevance and Min-Redundancy (mRMR) and least absolute shrinkage and selection operator (LASSO) were employed for the optimized model. Radscore was applied to construct a diagnostic model considering the T2WI texture features and patient demography to assess the power in differentiating SPs and ASPs.

Results: SPs and ASPs were seen in 75 and 40 patients, respectively. Thirty texture features were selected by mRMR, and LASSO identified a radscore of 16 radiomics features as being related to plaque vulnerability. The radscore, consisting of eight texture features, showed a better diagnostic performance than clinical information, both in the training (area under the curve [AUC], 0.923 vs. 0.713) and test groups (AUC, 0.989 vs. 0.735). The combination model of texture and clinical information had the best performance in assessing lesion vulnerability in both the training (AUC, 0.926) and test groups (AUC, 0.898).

Conclusion: This study demonstrated that HRMRI texture features provide incremental value for carotid atherosclerotic risk assessment.

Keywords: HRMRI; asymptomatic; carotid atherosclerosis; radiomics; symptomatic.

Figure 1

Study flowchart and Raidomics workflow.

Figure 2

Texture feature selection using the absolute shrinkage and selection operator (LASSO) binary logistic regression model. (A) The penalty term (λ) in the LASSO model was selected through 10-fold cross-validation which is based on minimum criteria. Y-axis means binomial deviances. Down-X-axis means the log (λ), Up-X-axis means the average number of predictors. The red dots indicate average deviance values for each λ that different models have different deviance. Left-vertical-line shows the lowest deviance that the model showed the best fit to our training group. (B) LASSO coefficient profiles of the 16 features. The dotted vertical line was drawn at the value where log λ (0.008) resulted in eight non-zero coefficients.

Figure 3

Boxplots showed the Radscore, clinical model, and Rad_clin model value of SPs and ASPs in the training and test groups. (A,D) Radscore in the training group and the test group; (B,E) Radscore in the training group and the test group; (C,F) Radscore in the training group and the test group.

Figure 4

Diagnostic performance of univariate logistic regression model. (A,B) show the receiver operating characteristic curves of the Rad_clin model that show that the area under the curve is the highest among the Radscore and clinical model in the training and testing groups.

Figure 5

Barchart of the Rad_clin model in the training group and the test group. Yellow means symptomatic plaque (SP) and purple means asymptomatic plaque (ASP). When purple appears in the yellow area, this means the SP patients had been incorrectly predicted as ASP. When yellow appears in the purple area, this means the SP patients had been incorrectly predicted as ASP. (A) was the training group, (B) was the test group.

Figure 6

Diagnostic performance evaluation in the training group and test group. Calibration curves of the nomogram in the training (A) and the test groups (B). Calibration curves showed the calibration of the nomogram in terms of agreement between the predicted risk of vulnerable plaque and high-resolution magnetic resonance imaging (HRMRI)-observed vulnerable plaque. (C) Nomogram based on Rad_clin_model. (D) Decision curves for Radscore, clinical model, and Rad_clin_model; the Y-axis shows the model benefit. The red line represents the Rad_clin_model. The blue line represents the clinical model, and the green line represents the Radscore. The X-axis means the threshold probability. ROC, receiver operating characteristic; AUC, area under the curve.