| 蔡晓嘉,韩锦涛,刘晶,杨帆,王琦,李如迅.深度学习全模型迭代算法用于卵巢肿瘤术前低剂量CT[J].中国医学影像技术,2025,41(4):539~542 |
| 深度学习全模型迭代算法用于卵巢肿瘤术前低剂量CT |
| Artificial intelligence iterative reconstruction for preoperative low-dose CT of ovarian tumor |
| 投稿时间：2024-10-18 修订日期：2024-12-25 |
| DOI：10.13929/j.issn.1003-3289.2025.04.007 |
| 中文关键词: 卵巢肿瘤 人工智能 辐射剂量 体层摄影术,X线计算机 前瞻性研究 |
| 英文关键词:ovarian neoplasms artificial intelligence radiation dosage tomography, X-ray computed prospective studies |
| 基金项目: |
|
| 摘要点击次数: 535 |
| 全文下载次数: 114 |
| 中文摘要: |
| 目的 探讨深度学习全模型迭代算法(AIIR)用于卵巢肿瘤术前低剂量CT的价值。方法 前瞻性对70例卵巢肿瘤先后行腹盆腔常规(120 kVp,200 mAs)及低剂量(120 kVp,40 mAs)门静脉期扫描,对常规剂量图像采用混合迭代重建(HIR)(A组),对低剂量图像采用HIR(B组)及AIIR(C组);比较3组图像主、客观评价结果,记录基于各组图像诊断周围器官侵犯及腹膜转移准确率,以及低剂量与常规剂量扫描的辐射剂量。结果 图像显示肿瘤边界及分隔、肿瘤与周围器官分界清晰度的主观评分,以及肿瘤与腰大肌信噪比和对比度噪声比均依B组、A组、C组次序而升高(P均＜0.017);A组与C组显示肿瘤供血血管清晰度 主观评分差异无统计学意义(P=0.435),且均高于B组(P均＜0.017)。基于A、B、C组诊断周围器官侵犯的准确率分别为83.87%(52/62)、72.58%(45/62)及83.87%(52/62),诊断腹膜转移的准确率分别为85.71%(60/70)、78.57%(55/70)及84.29%(59/70)。相比常规剂量CT,低剂量CT有效剂量降低79.70%(2.60 mSv vs. 12.81 mSv, P＜0.001)。结论 AIIR可提高卵巢肿瘤低剂量CT图像质量和转移诊断效能。 |
| 英文摘要: |
| Objective To investigate the value of artificial intelligence iterative reconstruction (AIIR) in preoperative low-dose CT of ovarian tumor. Methods Seventy patients with ovarian tumor were prospectively enrolled. Routine-dose (120 kVp, 200 mAs) and low-dose (120 kVp, 40 mAs) contrast-enhanced abdominopelvic CT scanning at portal venous phase were sequentially performed. The routine-dose images were reconstructed with hybrid iterative reconstruction (HIR) (group A), while low-dose images were reconstructed with HIR (group B) and AIIR (group C), respectively. Subjective and objective evaluation of image quality were compared among groups, and the diagnostic accuracy of peripheral organ invasion and peritoneal metastasis based on group A, B and C, as well as radiation dose of routine- and low-dose scanning were recorded. Results In group B, A and C, the subjective scoring of definition of tumor margin and septation, boundary between tumor and surrounding organ, as well as the signal-to-noise ratio and contrast-to-noise ratio of ovarian tumor and psoas muscles, increased successively (all P＜0.017). No significant difference of subjective scoring of tumor feeding vessel clarity was found between group A and C (P=0.435), which were both higher than that in group B (P＜0.017). The accuracy for diagnosing peripheral organ invasion based on group A, B and C was 83.87% (52/62), 72.58% (45/62) and 83.87% (52/62), for diagnosing peritoneal metastasis was 85.71% (60/70), 78.57% (55/70) and 84.29% (59/70), respectively. Compared to routine-dose CT, the effective dose of low-dose CT was reduced by 79.70% (2.60 mSv vs. 12.81 mSv, P＜0.001). Conclusion AIIR could improve image quality and metastasis diagnostic efficacy in low-dose CT of ovarian tumors. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
