PERAN ARTIFICIAL INTELLIGENCE (AI) DALAM MENINGKATKAN AKURASI DIAGNOSIS SITOLOGI NON-GINEKOLOGI: SYSTEMATIC LITERATURE REVIEW

Abstrak

Sitologi non-ginekologi memegang peranan penting dalam skrining dan diagnosis berbagai keganasan sel di luar traktus genitalia wanita, namun interpretasinya masih sangat bergantung pada penilaian subjektif ahli patologi sehingga rentan terhadap variabilitas antar-pemeriksa. Kecerdasan buatan (Artificial Intelligence/AI), khususnya yang berbasis pembelajaran mendalam (deep learning) sudah berkembang pesat sebagai alat bantu diagnostik yang berpotensi meningkatkan akurasi, konsistensi, dan efisiensi pemeriksaan sitologi. Penelitian ini bertujuan merangkum dan mengevaluasi bukti ilmiah mengenai peran AI dalam meningkatkan akurasi diagnosis sitologi non-ginekologi melalui pendekatan Tinjauan Pustaka (SLR). Pencarian literatur dilakukan pada basis data Google Scholar dan PubMed serta pencarian manual mengikuti kerangka kerja PRISMA 2020, dengan rentang publikasi tahun 2016–2026. Setelah proses skrining berjenjang berdasarkan judul, abstrak, dan teks lengkap, diperoleh dua belas studi yang memenuhi kriteria inklusi, mencakup pemeriksaan sitologi tiroid, kandung kemih, paru-paru, dan pankreas. Hasil tinjauan menunjukkan bahwa model AI berbasis convolutional neural network (CNN) dan arsitektur turunannya mampu mencapai nilai sensitivitas, spesifisitas, dan akurasi yang tinggi, bahkan pada beberapa studi setara dengan penilaian ahli patologi berpengalaman. Meskipun demikian, keterbatasan seperti ukuran sampel yang kecil, desain retrospektif, dan minimnya standarisasi masih perlu diatasi sebelum penerapan AI dapat dilakukan secara luas dalam praktik klinis rutin.

Non-gynecological cytology plays an important role in the screening and diagnosis of various malignancies outside the female genital tract, yet its interpretation remains highly dependent on the subjective assessment of cytopathologists, making it prone to inter-observer variability. Artificial intelligence (AI), particularly deep learning-based approaches, has developed rapidly as a diagnostic aid with the potential to improve the accuracy, consistency, and efficiency of cytological examination. This study aims to summarize and evaluate the scientific evidence regarding the role of AI in improving the diagnostic accuracy of non-gynecological cytology through a Systematic Literature Review (SLR) approach. The literature search was conducted on the Google Scholar and PubMed databases as well as manual searching, following the PRISMA 2020 framework, covering publications from 2016–2026. After a staged screening process based on title, abstract, and full text, twelve studies met the inclusion criteria, covering thyroid, urinary bladder, lung, and pancreatic cytology. The review results show that AI models based on convolutional neural networks (CNNs) and their derivative architectures achieved high sensitivity, specificity, and accuracy, in several studies comparable to the assessment of experienced cytopathologists. Nevertheless, limitations such as small sample sizes, retrospective designs, and a lack of standardization still need to be addressed before AI can be widely implemented in routine clinical practice.