摘要: |
目的 探讨经直肠超声(TRUS)征象联合纹理分析对克罗恩肛瘘(PFCD)和腺源性肛瘘(CAF)的鉴别诊断价值。方法 回顾性分析110例肛瘘患者的TRUS图像(按确诊病史分为PFCD组,n=42;CAF组,n=68),对比其病灶TRUS征象。使用3D Slicer提取病灶的纹理特征,经套索回归和交叉验证降维筛选后与TRUS征象分别构建TRUS征象模型、纹理特征模型及联合模型,并通过ROC曲线评估各模型的诊断效能。结果 联合模型共筛选出克罗恩超声瘘管征(CUFS)、分叉支管、峰度、自相关性四个特征(p<0.05),曲线下面积(AUC)为0.761、0.676、0.730、0.770,其中CUFS的特异度最高为92.6%,PFCD组的峰度和自相关性都高于CAF组(OR=5.1、5.2,p<0.02),联合模型的AUC为0.927(95%CI 0.880~0.974)诊断效能强于TRUS征象模型与纹理特征模型(p<0.05)。结论 CUFS为PFCD的特异性征象,高峰度与高自相关性可能是兼有肠道炎症的瘘管的纹理特征表现,联合TRUS征象与纹理分析对PFCD与CAF有较好的鉴别诊断价值。 |
关键词: TRUS 纹理分析 克罗恩病 肛瘘 鉴别诊断 |
DOI: |
投稿时间:2023-04-04修订日期:2023-11-26 |
基金项目:2022年省科技计划专项资金(BK20221178);2020年度省中医药科技发展计划项目(YB2020029);2021年宿迁市科技计划项目(自然科学资金)(K202144);南京市“十三五”规划名中医工作室建设项目(ZXP-2019-NJ) |
|
Differential diagnosis value of TRUS findings combined with texture analysis on perianal fistulizing Crohn’s disease and cryptoglandular anal fistula |
xuchengmu,dingkang,tanyanyan |
(Nanjing Hospital of Chinese Medicine Affiliated to Nanjing University of Chinese Medicine) |
Abstract: |
Objective To explore the differential diagnosis value of transrectal ultrasonography (TRUS) signs combined with texture analysis on perianal fistulizing Crohn’s disease (PFCD) and cryptoglandular anal fistula (CAF). Methods The TRUS images of 110 patients with anal fistula were analyzed retrospectively (according to diagnosed medical history divided into PFCD group, n=42; CAF group, n=68), and the TRUS findings of the lesions were compared. The texture features of lesions were extracted by 3D Slicer. TRUS sign models, texture feature models and combined models were constructed with texture features after lasso regression and cross-validation dimensionality reduction screening, and the diagnostic performance of each model was evaluated by ROC curve. Results The combined model screened out four features of Crohn’s Ultrasound Fistula Sign (CUFS), fistula bifurcation, kurtosis, and autocorrelation (p<0.05) with areas under the curve (AUC) of 0.761, 0.676, 0.730, 0.770 respectively. CUFS had the highest specificity at 92.6%. PFCD group had higher kurtosis and autocorrelation than the CAF group (OR=5.1, 5.2, p<0.02). The combined model showed a greater diagnostic efficiency with an AUC of 0.927 (95%CI 0.880~0.974) compared to the TRUS sign model and the texture feature model (p<0.05). Conclusion CUFS is a specific sign of PFCD. High kurtosis and autocorrelation may be characteristic indicators of the textural features of fistulas coexisting with intestinal inflammation. Combining TRUS findings and texture analysis has good differential diagnosis value for PFCD and CAF. |
Key words: TRUS texture analysis Crohn’s disease anal fistula differential diagnosis |