This meta-analysis assesses the diagnostic performance of deep learning (DL) algorithms for early breast and cervical cancer detection. Thirty-five studies were reviewed, with 20 meta-analyzed, showing pooled sensitivity of 88%, specificity of 84%, and AUC of 0.92. DL algorithms showed comparable performance across subgroups (cancer type, validation type, imaging modality). However, limitations in study design and reporting suggest potential bias and overestimation of algorithm performance. Standardized guidelines for study methods and reporting are needed to improve DL research quality.
