Radiomics-based prediction of pathological subtypes in peripheral neuroblastic tumors using 123I-MIBG SPECT/CT imaging: an observational study
摘要
The purpose of the study is to employ a radiomics approach based on 123I-MIBG SPECT/CT imaging to predict pathological subtypes in peripheral neuroblastic tumors (pNTs). A retrospective and exploratory study was conducted involving 67 pediatric patients with pNTs, who were randomly divided into training and validation cohorts at a ratio of 7:3. Clinical and radiomics features were selected using univariate feature selection and recursive feature elimination methods. By integrating clinical and radiomics features, a combined model based on logistic regression and a voting classifier incorporating four algorithms were constructed to optimize prediction accuracy. A total of 1702 features were extracted from SPECT and CT features. Ultimately, six clinical and nine radiomic features were included in our analysis. The combined model integrating clinical and radiomic features achieved a macro-average area under the curve (AUC) of 0.871 and an overall accuracy of 80.4% in the training set, and a macro-average AUC of 0.836 with an overall accuracy of 81.0% in the test set. The voting classifier significantly improved performance, achieving a macro-average AUC of 0.968 with an overall accuracy of 87.0% in the training set, and achieved a macro-average AUC of 0.879 in the test set, demonstrating robust stability and high accuracy. Conclusions:The study demonstrates the potential of radiomics as a non-invasive diagnostic tool for differentiating pathological subtypes of pNTs, which could significantly influence treatment planning and surgical decisions.