KEY TAKEAWAYS
- The study aimed to investigate the diagnostic accuracy of ultrasound-based AI in predicting key molecular biomarkers in BC.
- Researchers noticed promising accuracy of AI in predicting key molecular biomarkers in BC; further investigation is ongoing.
Breast cancer (BC) diagnosis and treatment rely heavily on molecular markers such as HER2, Ki67, PR, and ER, typically identified through invasive methods. This meta-analysis explores ultrasound-based radiomics as an innovative method for predicting these markers, potentially offering a non-invasive alternative.
Yuxia Fu and the team aimed to assess the diagnostic accuracy of ultrasound-based radiomics in predicting key molecular markers—HER2, Ki67, PR, and ER—in patients with BC.
They performed a comprehensive search of PubMed, EMBASE, and Web of Science databases to identify studies evaluating ultrasound-based radiomics in BC. Inclusion criteria encompassed research focusing on HER2, Ki67, PR, and ER as key molecular markers. Quality assessment was conducted using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) and Radiomics Quality Score (RQS), with systematic data extraction.
They found that ultrasound-based radiomics achieved a sensitivity of 0.76 and specificity of 0.78 in predicting HER2 status, and 0.80 sensitivity with 0.76 specificity for Ki67 biomarkers. Quantitative analysis for PR and ER prediction was limited due to insufficient data from studies. The overall quality assessment using the RQS indicated moderate quality across the studies. Evaluation with the QUADAS-2 tool revealed unclear risk of bias in the flow and timing domain.
The study concluded that AI models show promising accuracy in predicting key molecular biomarkers’ status in patients with BC, particularly for HER2 and Ki67 biomarkers. However, there was insufficient data to conduct a meta-analysis for ER and PR prediction accuracy. Overall, the quality of the studies was deemed acceptable. Future research should prioritize thorough reporting of results and include more prospective studies from diverse centers.
The study received no funds.
Source: https://pubmed.ncbi.nlm.nih.gov/38820391/
Fu Y, Zhou J, Li J, 2024. “Diagnostic performance of ultrasound-based artificial intelligence for predicting key molecular markers in breast cancer: A systematic review and meta-analysis.” PLoS One. 2024 May 31;19(5):e0303669. doi: 10.1371/journal.pone.0303669. PMID: 38820391; PMCID: PMC11142607.
