| 刘松,田超,任涛,曹宸,靳松,夏爽.CT血管造影影像组学评估基底动脉尖动脉瘤破裂风险[J].中国医学影像技术,2025,41(1):20~24 |
| CT血管造影影像组学评估基底动脉尖动脉瘤破裂风险 |
| CT angiography radiomics for evaluating risk of basilar tip aneurysm rupture |
| 投稿时间：2024-08-30 修订日期：2025-01-20 |
| DOI：10.13929/j.issn.1003-3289.2025.01.005 |
| 中文关键词: 动脉瘤破裂 体层摄影术,X线计算机 机器学习 影像组学 |
| 英文关键词:aneurysm, ruptured tomography, X-ray computed machine learning radiomics |
| 基金项目:天津市卫生健康科技项目(TJWJ2022QN061)、国家自然科学基金(82171916)、天津市自然科学基金(21CYBJC01580)、天津市卫生科技专项(重点学科专项)项目(TJWJ2022XK019)。 |
|
| 摘要点击次数: 404 |
| 全文下载次数: 110 |
| 中文摘要: |
| 目的 观察CT血管造影(CTA)影像组学评估基底动脉尖动脉瘤(BTA)破裂风险的价值。方法 回顾性收集133例BTA患者,根据BTA破裂与否分为破裂组(n=39)与未破裂组(n=94),并按7∶3比例划分训练集(n=93)与测试集(n=40)。针对CTA提取BTA 影像组学特征并筛选其中最优者,计算影像组学评分(Radscore);分别利用逻辑回归(LR)、随机森林(RF)、决策树(DT)及K近邻(KNN)算法基于最优影像组学特征建立影像组学机器学习(ML)模型,基于Radscore建立Radscore模型;再利用后者联合临床及常规影像学表现建立联合模型。对比分析上述各模型鉴别破裂与未破裂BTA的效能。结果 最终选出4个BTA最优影像组学特征,以之建立的LR、RF、DT及KNN 4种影像组学模型鉴别训练集破裂与未破裂BTA的曲线下面积(AUC)分别为0.770、0.816、0.817及0.795,在测试集分别为0.795、0.793、0.786及0.824;4种模型AUC差异均无统计学意义(P均＞0.05)。性别、饮酒史、BTA形态及Radscore均为BTA破裂的独立影响因素(P均＜0.05),以之建立临床-常规影像学模型。针对全部133例,联合模型鉴别BTA破裂与未破裂的AUC为0.877,Radscore模型为0.775,而临床-常规影像学模型为0.677,前者明显高于后两者(P均＜0.05)。结论 CTA影像组学有助于评估BTA破裂风险;联合临床及常规影像学表现可进一步提高其效能。 |
| 英文摘要: |
| Objective To observe the value of CT angiography (CTA) radiomics for evaluating the risk of basilar tip aneurysm (BTA) rupture. Methods Totally 133 BTA patients were retrospectively enrolled and divided into ruptured group (n=39) and unruptured group (n=94) based on BTA ruptured or not,also divided into training set (n=93) and test set (n=40) at the ratio of 7∶3. CTA radiomics features of BTA were extracted, the best radiomics features were screened, and the radiomics score (Radscore) was calculated. Then machine learning (ML) models were established with logistic regression (LR), random forest (RF), decision tree (DT) and K-nearest neighbor (KNN) algorithms, respectively. Radscore model was also established, and finally a combined model was constructed based on clinical data, routine imaging findings and Radscore. The efficacy of the above models for evaluating the risk of BTA rupture were comparatively analyzed. Results Finally 4 radiomics features of BTA were obtained. The area under the curve (AUC) of LR, RF, DT and KNN radiomics models for differentiating ruptured and unruptured BTA in training set was 0.770, 0.816, 0.817 and 0.795, respectively, while that in test set was 0.795, 0.793, 0.786 and 0.824, respectively, both being not significant different (both P＞0.05). Patient's gender, alcohol consumption history, BTA morphology and Radscore were all independent impact factors of BTA rupture (all P＜0.05), which were used to establish a clinical-routine imaging model. For all 133 cases, AUC of the combination model for differentiating ruptured and unruptured BTA was 0.877, of Radscore model was 0.775, while that of clinical-routine imaging model was 0.677, of the former was significantly higher than of the last two (both P＜0.05). Conclusion CTA radiomics was helpful for evaluating the risk of BTA rupture. Combining with clinical data and routine imaging findings could further improve the value of CTA radiomics. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
