| 王沄,王天娇,陈钰,苏童,冯逢,金征宇.深度学习重建算法用于改善颈部肿瘤加速T2WI成像质量[J].中国医学影像技术,2025,41(9):1573~1576 |
| 深度学习重建算法用于改善颈部肿瘤加速T2WI成像质量 |
| Deep learning reconstruction algorithm for improving imaging quality of accelerated T2WI of cervical tumors |
| 投稿时间：2025-07-31 修订日期：2025-09-15 |
| DOI：10.13929/j.issn.1003-3289.2025.09.025 |
| 中文关键词: 头颈部肿瘤|深度学习|磁共振成像 |
| 英文关键词:head and neck neoplasms|deep learning|magnetic resonance imaging |
| 基金项目:国家自然科学基金(82371962)、中央高水平医院临床科研专项(2022-PUMCH-B-067)、2021SKY北京影像科研基金(Z-2014-07-2101)。 |
|
| 摘要点击次数: 103 |
| 全文下载次数: 34 |
| 中文摘要: |
| 目的 观察深度学习(DL)重建算法用于改善颈部肿瘤加速T2WI成像质量的价值。方法 前瞻性纳入43例疑诊颈部肿瘤患者,采集颈部常规T2WI和基于DL重建的加速T2WI(DL-T2WI)。由2名医师以4分法针对整体图像质量、伪影、噪声、锐利度及病变可检测性对图像质量进行主观评估;比较常规T2WI与DL-T2WI主观评分。结果 采集常规T2WI耗时116 s,DL-T2WI耗时101 s。观察者间对于整体图像质量、伪影、噪声、锐利度及病变可检测性评分结果的一致性均非常好(Kappa=0.851~0.969);常规T2WI与DL-T2WI整体图像质量及病变可检测性评分差异均无统计学意义(P均＞0.05);DL-T2WI伪影及锐利度评分均显著高于、而噪声评分显著低于常规T2WI(P均＜0.05)。结论 DL重建算法有助于改善颈部肿瘤加速T2WI成像质量。 |
| 英文摘要: |
| Objective To observe the value of deep learning (DL) reconstruction algorithm for improving imaging quality of accelerated T2WI of cervical tumors. Methods A total of 43 patients with suspected cervical tumors were prospectively enrolled. Cervical conventional T2WI and accelerated T2WI based on DL reconstruction (DL-T2WI) were acquired. The imaging quality was subjectively assessed by 2 physicians using a 4-point system, including overall image quality, artifact, noise, sharpness and lesion detectability scores, and then were compared between conventional T2WI and DL-T2WI. Results The acquisition of conventional T2WI took 116 s, while of DL-T2WI took 101 s. The inter-observer consistency of subjective evaluation results on the overall image quality, artifact, noise, sharpness and lesion detectability scores were all excellent (Kappa=0.851—0.969). No significant difference of subjective evaluation results on overall image quality nor lesion detectability scores was found between conventional T2WI and DL-T2WI (both P＞0.05), while the artifact and sharpness scores of DL-T2WI were significantly higher but the noise score was significantly lower than those of conventional T2WI (all P＜0.05). Conclusion DL reconstruction algorithm was helpful for improving imaging quality of accelerated T2WI of cervical tumors. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
