Prediction of Carotid In-Stent Restenosis by Computed Tomography Angiography Carotid Plaque-Based Radiomics

Abstract

In-stent restenosis (ISR) after carotid artery stenting (CAS) critically influences long-term CAS benefits and safety. The study was aimed at screening preoperative ISR-predictive features and developing predictive models. Thus, we retrospectively analyzed clinical and imaging data of 221 patients who underwent pre-CAS carotid computed tomography angiography (CTA) and whose digital subtraction angiography data for verifying ISR presence were available. Carotid plaque characteristics determined using CTA were used to build a traditional model. Backward elimination (likelihood ratio) was used for the radiomics model. Furthermore, a combined model was built using the traditional and radiomics features. Five-fold cross-validation was used to evaluate the accuracy of the trained classifier and stability of the selected features. Follow-up angiography showed ISR in 30 patients. Carotid plaque length and thickness were independently associated with ISR (multivariate analysis); regarding the conventional model, the area under the curve (AUC) was 0.84 and 0.82 in the training and validation cohorts, respectively. The corresponding AUC values for the radiomics-based model were 0.87 and 0.82, and those for the optimal combined model were 0.88 and 0.83. Plaque length and thickness could independently predict post-CAS ISR, and the combination of radiomics and plaque features afforded the best predictive performance.

Quality assessment of radiomics models in carotid plaque: a systematic review

“…For predictive power, the mean AUC of the final models in training group was 0.876±0.09, ranging from 0.741 to 0.989. Internal validation was performed in 14 studies ( 25 - 27 , 29 , 32 - 38 , 40 , 41 , 43 ), and 1 study employed external validation ( 42 ), but only 12 studies ( 25 , 26 , 29 , 32 - 34 , 36 - 38 , 40 , 41 , 43 ) reported the AUCs of the validation groups, ranging from 0.73 to 0.986. Model calibration was investigated in 7 studies (36.8%) and demonstrated good calibration performance ( 25 , 30 , 32 , 37 , 40 , 42 , 43 ).…”

“…Through radiomics or texture analysis, 10 studies tended to classify unstable or risk plaques ( 25 , 27 , 29 , 32 , 33 , 39 - 43 ). Moreover, 5 studies focused on predicting unfavorable cerebral and cardiovascular events ( 28 , 35 - 38 ), 2 on evaluating stenosis or in-stent restenosis ( 26 , 30 ), 1 on assessing the robustness of radiomics features ( 34 ), and 1 on evaluating the relationship between glycated hemoglobin and ultrasound plaque textures ( 31 ). The details are shown in Table 2 .…”

“…Only 1 study described the full prediction model ( 36 ), and 8 studies used a nomogram to explain how the models generated personalized predictions (item 15b). Finally, 2 studies failed to discuss the implications for further research ( 33 , 36 ), and appendix information was not available in 7 studies ( 26 , 27 , 30 , 31 , 35 , 37 , 39 ).…”

“…All studies exhibited a low events per variable (EPV) value due to the large amount of candidate features and the limited carotid plaque samples. Regarding missing data, 12 studies conducted complete-case analysis and manually eliminated individuals without satisfactory images or clinical data or those who were lost to follow-up ( 25 , 26 , 28 - 30 , 32 , 36 - 41 ). In contrast, 7 studies did not describe how missing data were handled or the analytical procedures used to assess missing data ( 27 , 31 , 33 - 35 , 42 , 43 ), 4 studies lacked validation ( 28 , 30 , 31 , 39 ), and 2 studies had a high risk of bias in domain 1 for the stated inappropriate inclusion or exclusion criteria ( 33 , 39 ).…”