Page 60 - eBook_Proceedings of the International Conference on Digital Manufacturing V2

Proceedings of the International Conference on Digital Manufacturing –
Volume 2

This work establishes Swin Transformers as a paradigm-
shifting tool for cervical cancer screening, combining diagnostic
accuracy with computational efficiency for scalable deployment
in resource-constrained settings.

Keywords: Mask2former, Swin Transformer, Cervical cancer,
Detection, Classification.

INTRODUCTION

Cervical cancer continues to pose a significant global health
challenge, ranking as the fourth most common cancer among
women worldwide (Rutili de Lima, Khan, Shah & Ferri, 2023). Its
prevalence is disproportionately high in low- and middle-income
countries, where access to routine healthcare and cervical
screening programs remains limited (Glučina, Lorencin, Anđelić
& Lorencin, 2023). The disease typically progresses from pre-
neoplastic transformations within the cervical epithelium. The
cells involved are typically located deep within surrounding
tissues, as depicted in Figure 20, making early detection critical
for effective intervention and improved patient outcomes
(Ghoneim, Muhammad & Hossain, 2020; Pacal & Kılıcarslan,
2023). Among various screening methods, the Papanicolaou (Pap)
smear test remains the gold standard due to its non-invasive nature
and cost-effectiveness (Meza Ramirez, Greenop, Almoshawah,
Martin Hirsch & Rehman, 2023). The procedure involves
collecting and examining exfoliated cervical cells to identify
potential abnormalities. However, the manual interpretation of
Pap smear slides is labour-intensive and error-prone (Xiang et al.,
2020; Yaman & Tuncer, 2022). Cytopathologists must distinguish
between a wide range of normal and abnormal cellular
morphologies, with considerable variation in cell size, shape, and
staining intensity. These diagnostic challenges underscore the
need for reliable, automated tools to assist in cervical cancer
screening (Win, Kitjaidure, Hamamoto & Aung, 2020; Rehman,
Ali, Taj, Sajid & Karimov, 2020).

44