Scholarlink

Purpose <p>Minimal ablative margins (MAM) are critical for local tumour control after microwave ablation, yet their assessment remains&#xa0;subjective. We evaluated whether software improves the precision of ablation margin assessment after liver MWA.</p> Materials and Methods <p>In this retrospective study, patients undergoing CT-guided liver MWA between January 2019 and January 2025 were included. Two radiologists (with 4 and 5&#xa0;years of experience in abdominal imaging) assessed MAM in two steps: first by side-by-side visual comparison of pre- and post-ablation CT using anatomical landmarks, then with dedicated software providing semi-automatic 3D co-registration and margin quantification. Each method included two reading sessions per reader, six weeks apart. Precision was assessed using intraclass correlation coefficients (ICC), Cohen’s kappa (κ) for categorical classification (&lt; 5, 5–10, &gt; 10&#xa0;mm), and Bland–Altman analysis.</p> Results <p>Sixty-seven patients with 87 liver tumours were included. Visual assessment showed poor reliability, with low intrareader agreement (ICC 0.10 and 0.33) and near-random interreader agreement (κ =  − 0.10; ICC − 0.05), with limits of agreement approaching 17&#xa0;mm. Software-assisted assessment improved reliability, with higher intrareader ICCs (0.54 and 0.83) and improved interreader agreement (κ = 0.38; ICC = 0.54). Interreader bias decreased from − 2.84 to + 0.91&#xa0;mm, and limits of agreement narrowed to − 4.10 to + 5.92&#xa0;mm (<i>p</i> &lt; 0.001). The median absolute interreader difference decreased from 3.0 to 1.0&#xa0;mm (<i>p</i> &lt; 0.001).</p> Conclusion <p>Software-assisted assessment improves&#xa0;reliability of MAM&#xa0;evaluation compared with visual methods, but interreader agreement remains only fair-to-moderate; further validation is required before routine clinical adoption.</p> Graphical Abstract <p></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

Precision of Ablation Margin Assessment After Liver Microwave Ablation: A Head-to-Head Comparison of Visual and Software-Assisted Methods

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Coline Morisot,
  • Idriss Chami Khazraji,
  • Vincent Habouzit,
  • Nicolas Magand,
  • Bertrand Le Roy,
  • Ilies El Boukili,
  • Claire Boutet,
  • Sylvain Grange,
  • Rémi Grange

摘要

Purpose

Minimal ablative margins (MAM) are critical for local tumour control after microwave ablation, yet their assessment remains subjective. We evaluated whether software improves the precision of ablation margin assessment after liver MWA.

Materials and Methods

In this retrospective study, patients undergoing CT-guided liver MWA between January 2019 and January 2025 were included. Two radiologists (with 4 and 5 years of experience in abdominal imaging) assessed MAM in two steps: first by side-by-side visual comparison of pre- and post-ablation CT using anatomical landmarks, then with dedicated software providing semi-automatic 3D co-registration and margin quantification. Each method included two reading sessions per reader, six weeks apart. Precision was assessed using intraclass correlation coefficients (ICC), Cohen’s kappa (κ) for categorical classification (< 5, 5–10, > 10 mm), and Bland–Altman analysis.

Results

Sixty-seven patients with 87 liver tumours were included. Visual assessment showed poor reliability, with low intrareader agreement (ICC 0.10 and 0.33) and near-random interreader agreement (κ =  − 0.10; ICC − 0.05), with limits of agreement approaching 17 mm. Software-assisted assessment improved reliability, with higher intrareader ICCs (0.54 and 0.83) and improved interreader agreement (κ = 0.38; ICC = 0.54). Interreader bias decreased from − 2.84 to + 0.91 mm, and limits of agreement narrowed to − 4.10 to + 5.92 mm (p < 0.001). The median absolute interreader difference decreased from 3.0 to 1.0 mm (p < 0.001).

Conclusion

Software-assisted assessment improves reliability of MAM evaluation compared with visual methods, but interreader agreement remains only fair-to-moderate; further validation is required before routine clinical adoption.

Graphical Abstract