CERVIA: A Hybrid Deep Learning Framework With Explainable  AI for Automated Cervical Cancer Classification From Pap Smear Imagesys

The primary objective of the CERVIA framework is to develop an automated system capable of accurately classifying cervical cells from Pap smear images into normal, precancerous, and malignant categories. To achieve this, CERVIA integrates three complementary deep learning models—HybridDenseNetInception, KaryoFormer, and CytomeshModel—enabling the extraction of multi-scale, global, and structural cellular features for enhanced classification performance. The framework also incorporates explainable AI techniques to highlight the morphological and cytological factors influencing predictions, thereby improving clinical transparency and trust. Ultimately, CERVIA aims to reduce pathologists’ workload, facilitate early detection, and ensure robust, reliable performance across diverse datasets.