目的:比较对比增强计算机断层扫描(CECT)和非对比磁共振成像(MRI)的影像组学在评估小儿周围神经母细胞瘤(PNTs)的细胞行为方面的性能。
方法:回顾性分析81例接受静脉期CECT的PNT患者,T1加权成像(T1WI),进行T2加权成像(T2WI)扫描。根据病理亚型将患者分为神经母细胞瘤和神经节神经母细胞瘤/神经节神经瘤。此外,根据国际神经母细胞瘤病理学分类(INPC),将其分为有利组织学和不利组织学.在CECT上分割感兴趣的肿瘤区域,轴向T1WI,和轴向T2WI图像,和影像组学模型是根据选定的影像组学特征开发的。经过五次交叉验证,使用受试者工作特征曲线下面积(AUC)和准确性比较了CECT和MRI的影像组学模型的性能.
结果:为了区分病理亚型,CECT影像组学模型的AUC范围为0.765~0.870,准确度范围为0.728~0.815.相比之下,MRI影像组学模型的AUC范围为0.549~0.748,准确度范围为0.531~0.778.关于INPC分组的歧视,CECT影像组学模型的AUC范围为0.503~0.759,准确度范围为0.432~0.741.同时,MRI影像组学模型的AUC范围为0.512~0.739,准确度范围为0.605~0.815.
结论:CECT影像组学在评估病理亚型方面优于非对比MRI影像组学。在评估INPC分组时,CECT影像组学与非对比MRI影像组学具有可比性。
OBJECTIVE: To compare the performance of radiomics from contrast-enhanced computed tomography (CECT) and non-contrast magnetic resonance imaging (MRI) in assessing cellular behavior in pediatric peripheral neuroblastic tumors (PNTs).
METHODS: A retrospective analysis of 81 PNT patients who underwent venous phase CECT, T1-weighted imaging (T1WI), and T2-weighted imaging (T2WI) scans was conducted. The patients were classified into neuroblastoma and ganglioneuroblastoma/
ganglioneuroma based on their pathological subtypes. Additionally, they were categorized into favorable histology and unfavorable histology according to the International Neuroblastoma Pathology Classification (INPC). Tumor regions of interest were segmented on CECT, axial T1WI, and axial T2WI images, and radiomics models were developed based on the selected radiomics features. Following five-fold cross-validation, the performance of the radiomics models derived from CECT and MRI was compared using the area under the receiver operating characteristic curve (AUC) and accuracy.
RESULTS: For discriminating pathological subtypes, the AUC for CECT radiomics models ranged from 0.765 to 0.870, with an accuracy range of 0.728 to 0.815. In contrast, the AUC for MRI radiomics models ranged from 0.549 to 0.748, with an accuracy range of 0.531 to 0.778. Regarding the discrimination of INPC subgroups, the AUC for CECT radiomics models ranged from 0.503 to 0.759, with an accuracy range of 0.432 to 0.741. Meanwhile, the AUC for MRI radiomics models ranged from 0.512 to 0.739, with an accuracy range of 0.605 to 0.815.
CONCLUSIONS: CECT radiomics outperforms non-contrast MRI radiomics in evaluating pathological subtypes. When assessing INPC subgroups, CECT radiomics demonstrates comparability with non-contrast MRI radiomics.