Construction And Evaluation Of Multi-Radiomics Models For Identifying Intracranial Aneurysm Instability Based On CTA

Background and Aims:For therapeutic decision-making,recognizing unruptured intracranial aneurysm instability is essential.By building and evaluating several models,this study aims to evaluate the contribution of radiomics and traditional morphological features to the identification of aneurysm instability.Materials and Methods:Retrospective data collection was done on patients who were diagnosed with intracranial aneurysms at Sichuan Provincial People’s Hospital from November 2015 to December 2019.Comprehensive criteria were utilized to separate aneurysms into unstable and stable groups.The unstable group was defined as aneurysms with near-term rupture,growth during follow-up,or caused compressive symptoms;those without the aforementioned conditions were grouped as stable aneurysms.Aneurysms were randomly divided into training and test sets at a 1:1 ratio(According to the“sampling method”proposed by Muhammed et al.^[1]when the sample size of the positive group was small).Radiomics and traditional morphological features（maximum diameter,irregular shape,aspect ratio,size ratio,location,etc.）were extracted.Three basic models and two integrated models were constructed after corresponding statistical analysis.Model A was based on significant traditional morphological parameters identified by multivariate analysis.Model B was based on Radiomics features optimized using least absolute shrinkage and selection operator（LASSO）regression.Model C used merely the significant Radiomics features related to aneurysm morphology.Furthermore,combined models of traditional and Radiomics features were built（model A+B,model A+C）.The area under curves（AUC）of each model was calculated and compared.Results:There were 31（13.7%）patients harboring 36（14.2%）unstable aneurysms,15 of which ruptured post-imaging,16 with growth on serial imaging,and 5 with compressive symptoms,respectively.Four traditional morphological features,six Radiomics features,and three Radiomics-derived morphological features were identified.The performance of the classification of aneurysm stability was:[AUC:0.921（95%CI:0.862–0.981）on the training set and 0.909（95%CI:0.853–0.965）on the test set]in Model A,[AUC:0.865（95%CI:0.778–0.951）on the training set and 0.739（95%CI:0.636–0.841）on the test set]in Model B,and[AUC:0.605（95%CI:0.470–0.739）on the training set and 0.552（95%CI:0.401–0.703）on the test set]in Model C,respectively.The AUC（0.868,95%CI:0.799-0.937）for model A+B was numerically slightly higher than any of the individual models,but the AUC（0.855,95%CI:0.790-0.921）for model A+C was not.Conclusions:Models constructed only using radiomics features of CTA or Radiomics-derived morphological features,were not superior to the model constructed by traditional morphological features.The model combining radiomics with traditional morphology could be an effective tool for identifying the instability of intracranial aneurysms and serving as the foundation for subsequent clinical diagnosis and treatment strategies.