Page 78 - eBook_Proceedings of the International Conference on Digital Manufacturing V2

Proceedings of the International Conference on Digital Manufacturing –
Volume 2

High recall and pixel accuracy underscore its suitability for
clinical adoption. Nonetheless, critical areas requiring further
enhancement have been identified, particularly concerning
precision, boundary delineation accuracy, and management of
class imbalances.

Future research should focus on implementing sophisticated
data augmentation techniques, refining annotation standards, and
leveraging advanced loss functions to enhance model precision
and segmentation accuracy. Investigating hybrid methodologies
combining transformer architectures with complementary
segmentation approaches or post-processing techniques could also
be beneficial.

Additionally, validating the model with larger, diverse clinical
datasets, and conducting prospective clinical studies will be
crucial for confirming its real-world applicability and
effectiveness, ultimately aiming to improve early detection and
clinical outcomes in cervical cancer screening programs.

REFERENCES

Ahmadzadeh Sarhangi, Hamed, Beigifard, Davoud, Farmani,
Ehsan, & Bolhasani, Hossein. (2024). Deep learning
techniques for cervical cancer diagnosis based on pathology
and colposcopy images. Informatics in Medicine Unlocked,
Article 101600.
AlMohimeed, Abdullah, Saleh, Hadeel, Mostafa, Sayed, Saad,
Rania M. A., & Talaat, Ahmed S. (2023). Cervical cancer
diagnosis using stacked ensemble model and optimised
feature selection: An explainable artificial intelligence
approach. Computers, 12(10), Article 200.
Alyafeai, Zaid, & Ghouti, Layla. (2020). A fully-automated deep
learning pipeline for cervical cancer classification. Expert
Systems with Applications, 141, Article 112951.
Attallah, Osama. (2023). Cervical cancer diagnosis based on
multi-domain features using deep learning enhanced by
handcrafted descriptors. Applied Sciences, 13(3), 1805.

62