Scholarlink

Purpose <p>To evaluate the depiction of cerebral aneurysms on magnetic resonance angiography (MRA) reconstructed using super-resolution deep learning reconstruction (SR-DLR).</p> Methods <p>We retrospectively reviewed the MRA images of 79 patients (49 with cerebral aneurysms and 30 without). The MRA images were subjected to SR-DLR and compared with the original images using both quantitative (signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for the basilar artery (BA) and aneurysms; full width at half maximum (FWHM), edge rise distance (ERD), and edge rise slope (ERS) for the BA) and qualitative metrics (depiction of aneurysms and BA, image sharpness, noise, artifacts, and overall image quality performed by three radiologists). Statistical comparisons were performed using a paired t-test for quantitative metrics and the Wilcoxon signed-rank test for qualitative metrics.</p> Results <p>SR-DLR significantly improved SNR, CNR, and ERD compared with the original images. ERS values showed a trend toward improvement with SR-DLR, whereas FWHM showed no significant difference. Depiction of the vasculature, image sharpness, noise, and overall image quality were rated as significantly better for SR-DLR by all readers. Most readers rated that the depiction of aneurysms was improved with SR-DLR, particularly for aneurysms &lt; 5&#xa0;mm. The assessment of artifacts varied among readers.</p> Conclusion <p>SR-DLR significantly improves the image quality of MRA and enhances the visualization of cerebral aneurysms, with greater improvements in the depiction of smaller aneurysms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Super-resolution deep learning reconstruction enhances visualization of cerebral aneurysms on magnetic resonance angiography

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月14日

查看全文
  • Jun Kanzawa,
  • Koichiro Yasaka,
  • Masayoshi Kato,
  • Noriko Kanemaru,
  • Yusuke Watanabe,
  • Yusuke Asari,
  • Yuki Sonoda,
  • Shigeru Kiryu,
  • Osamu Abe

摘要

Purpose

To evaluate the depiction of cerebral aneurysms on magnetic resonance angiography (MRA) reconstructed using super-resolution deep learning reconstruction (SR-DLR).

Methods

We retrospectively reviewed the MRA images of 79 patients (49 with cerebral aneurysms and 30 without). The MRA images were subjected to SR-DLR and compared with the original images using both quantitative (signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for the basilar artery (BA) and aneurysms; full width at half maximum (FWHM), edge rise distance (ERD), and edge rise slope (ERS) for the BA) and qualitative metrics (depiction of aneurysms and BA, image sharpness, noise, artifacts, and overall image quality performed by three radiologists). Statistical comparisons were performed using a paired t-test for quantitative metrics and the Wilcoxon signed-rank test for qualitative metrics.

Results

SR-DLR significantly improved SNR, CNR, and ERD compared with the original images. ERS values showed a trend toward improvement with SR-DLR, whereas FWHM showed no significant difference. Depiction of the vasculature, image sharpness, noise, and overall image quality were rated as significantly better for SR-DLR by all readers. Most readers rated that the depiction of aneurysms was improved with SR-DLR, particularly for aneurysms < 5 mm. The assessment of artifacts varied among readers.

Conclusion

SR-DLR significantly improves the image quality of MRA and enhances the visualization of cerebral aneurysms, with greater improvements in the depiction of smaller aneurysms.