Scholarlink

<p>AI-assisted radiomics is an emerging tool for precision oncology, and many studies have recently shown promising results. However, there are still differences in whether it can be applied in clinical practice. This study aimed to evaluate the diagnostic accuracy of PET-based radiomics in ML models for histological subtype classification of lung cancer through a systematic review and meta-analysis. The study protocol was registered in PROSPERO, CRD42024603590. Methodological quality and risk of bias were assessed using QUADAS-2 and RQS. For the meta-analysis, validation data statistics were extracted from the studies with Type 2a or above as per the TRIPOD statement. A random-effects model was used to estimate the overall effect size. Statistical heterogeneity was assessed using the <i>I</i><sup>2</sup> value. Fourteen studies were included in the systematic review, of which eight were eligible for meta-analysis. All the studies were performed with internal validation. The average RQS across studies was 10.47 ± 4.72. Pooled sensitivity, specificity, and diagnostic odds ratio (DOR) were 0.81 (95% CI 0.65–0.90), 0.79 (95% CI 0.75–0.83), and 22.42 (95% CI 9.04–55.59), respectively. The SROC curve suggested good diagnostic performance with an AUC of 0.89 (95% CI 0.83–0.95). The meta-analysis revealed significant heterogeneity using Cochrane’s <i>Q</i> test with <i>p</i> &lt; 0.001. ML models utilizing <sup>18</sup>F-FDG PET radiomics have the potential to predict histological subtypes of non-small cell lung cancer. External validation studies could provide stronger evidence for the generalizability.</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

Artificial Intelligence Assisted 18F-FDG PET Radiomics in Classifying Histological Subtypes of Lung Cancer: Systematic Review and Meta-analysis

  • 期刊论文

  • 发表日期：

    2025年5月17日

查看全文
  • Pooja Dwivedi,
  • Sagar Barage,
  • Ashish Jha,
  • Archi Agrawal,
  • Rajshri Singh,
  • Sayak Choudhury,
  • Venkatesh Rangarajan

摘要

AI-assisted radiomics is an emerging tool for precision oncology, and many studies have recently shown promising results. However, there are still differences in whether it can be applied in clinical practice. This study aimed to evaluate the diagnostic accuracy of PET-based radiomics in ML models for histological subtype classification of lung cancer through a systematic review and meta-analysis. The study protocol was registered in PROSPERO, CRD42024603590. Methodological quality and risk of bias were assessed using QUADAS-2 and RQS. For the meta-analysis, validation data statistics were extracted from the studies with Type 2a or above as per the TRIPOD statement. A random-effects model was used to estimate the overall effect size. Statistical heterogeneity was assessed using the I2 value. Fourteen studies were included in the systematic review, of which eight were eligible for meta-analysis. All the studies were performed with internal validation. The average RQS across studies was 10.47 ± 4.72. Pooled sensitivity, specificity, and diagnostic odds ratio (DOR) were 0.81 (95% CI 0.65–0.90), 0.79 (95% CI 0.75–0.83), and 22.42 (95% CI 9.04–55.59), respectively. The SROC curve suggested good diagnostic performance with an AUC of 0.89 (95% CI 0.83–0.95). The meta-analysis revealed significant heterogeneity using Cochrane’s Q test with p < 0.001. ML models utilizing 18F-FDG PET radiomics have the potential to predict histological subtypes of non-small cell lung cancer. External validation studies could provide stronger evidence for the generalizability.