Uso da inteligência artificial para o diagnóstico das alterações citológicas cérvico vaginais: uma revisão sistemática

Abstract

Cervical cancer is a major public health problem among people with an uterus, which is largely preventable through screening and early diagnosis. However, the traditional cytopathological test (Pap smear) has important limitations, such as variability in interpretation and reduced sensitivity in certain situations. In this scenario, artificial intelligence (AI) emerges as a promising alternative to optimize the detection of cytological alterations. This study carried out a systematic review of the literature with the aim of analyzing the use of AI in the diagnosis of cervicovaginal cytological alterations. The bibliographic search was conducted in the PubMed, SciELO and Elsevier databases, considering publications between 2005 and 2025. Approximately 136 articles were identified, of which 20 were selected based on inclusion criteria such as focus on AI applied to cytological images, and presentation of quantitative data (sensitivity and specificity). The results demonstrated that models based on convolutional neurais networks (CNN and Deep CNN) were widely used, with performances varying between studies. On average, AI models showed higher sensitivity than human professionals (88.77% versus 78.49%) with moderate to high standard deviation (11.01% versus 16.40%), while specificity was similar between groups (76.19% versus 73.94%), although with high standard deviation variability (27.08% versus 31.06%). AI proved to be effective especially in automated screening contexts, prioritizing the detection of alterations with high sensitivity, although with an increase in the number of false positives. It is concluded that, although it does not replace the cytopathologist, AI presents itself as a valuable tool to support diagnosis of cytopathological exams, with the potential to increase efficiency, standardize analyses and reduce human errors, as long as it is properly validated and integrated into existing clinical flows.