Scholarlink

<p>Radiomics applied to two-dimensional breast ultrasound has emerged as a potential noninvasive approach for differentiating benign from malignant breast lesions; however, its diagnostic accuracy and methodological reliability remain uncertain. This PRISMA-DTA–compliant systematic review and meta-analysis (PROSPERO CRD420251149769) synthesized evidence from observational studies published between 2019 and 2026. A comprehensive search of PubMed, Scopus, Web of Science, and the Cochrane Library identified 27 eligible studies. Pooled sensitivity, specificity, and hierarchical summary ROC (HSROC) area under the curve were estimated using bivariate random-effects models; heterogeneity, threshold effects, and small-study effects were assessed with I<sup>2</sup>, Spearman correlation, and Deeks' funnel-plot asymmetry test. The meta-analysis pooled data from the validation or test sets of 9,627 histopathologically confirmed lesions (5,733 benign, 3,894 malignant). Pooled sensitivity and specificity were 0.85 (95% CI 0.82–0.88) and 0.87 (95% CI 0.83–0.91), with strong overall discrimination (HSROC AUC 0.92). Between-study heterogeneity was substantial (I<sup>2</sup> 74.0–85.5%), reflecting variability in study design, validation strategies, and ultrasound acquisition reporting. No evidence of small-study effects or publication bias was detected (Deeks' p = 0.5054; t = 0.676, df = 25); however, risk-of-bias assessment using QUADAS-2 revealed frequent high or unclear concerns in patient selection and index test domains. Although radiomics-based models demonstrate clinically meaningful diagnostic performance, pooled estimates likely represent upper-bound accuracy due to methodological limitations and optimism bias. Prospective, multicenter, multi-vendor validation, standardized acquisition and reporting frameworks, and evaluation of incremental clinical utility within BI-RADS-guided workflows are required to support safe clinical translation.</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

A Systematic Review and Meta-analysis of Two-Dimensional Breast Ultrasound Radiomics: Implications for Lesion Characterization and Diagnostic Accuracy

  • 期刊论文

  • 发表日期：

    2026年7月7日

查看全文
  • Omar Abd Al Mjed Allasasmeh,
  • Areej H. Al-Sarairah,
  • Muath A. Odeh Hurani,
  • Mus’ab S. Alkasasbeh,
  • Iza Nurzawani Che Isa

摘要

Radiomics applied to two-dimensional breast ultrasound has emerged as a potential noninvasive approach for differentiating benign from malignant breast lesions; however, its diagnostic accuracy and methodological reliability remain uncertain. This PRISMA-DTA–compliant systematic review and meta-analysis (PROSPERO CRD420251149769) synthesized evidence from observational studies published between 2019 and 2026. A comprehensive search of PubMed, Scopus, Web of Science, and the Cochrane Library identified 27 eligible studies. Pooled sensitivity, specificity, and hierarchical summary ROC (HSROC) area under the curve were estimated using bivariate random-effects models; heterogeneity, threshold effects, and small-study effects were assessed with I2, Spearman correlation, and Deeks' funnel-plot asymmetry test. The meta-analysis pooled data from the validation or test sets of 9,627 histopathologically confirmed lesions (5,733 benign, 3,894 malignant). Pooled sensitivity and specificity were 0.85 (95% CI 0.82–0.88) and 0.87 (95% CI 0.83–0.91), with strong overall discrimination (HSROC AUC 0.92). Between-study heterogeneity was substantial (I2 74.0–85.5%), reflecting variability in study design, validation strategies, and ultrasound acquisition reporting. No evidence of small-study effects or publication bias was detected (Deeks' p = 0.5054; t = 0.676, df = 25); however, risk-of-bias assessment using QUADAS-2 revealed frequent high or unclear concerns in patient selection and index test domains. Although radiomics-based models demonstrate clinically meaningful diagnostic performance, pooled estimates likely represent upper-bound accuracy due to methodological limitations and optimism bias. Prospective, multicenter, multi-vendor validation, standardized acquisition and reporting frameworks, and evaluation of incremental clinical utility within BI-RADS-guided workflows are required to support safe clinical translation.