Cervical Cancer Diagnosis: Multiclass Classification with Transfer Learning for Enhanced Clinical Decision Support
Gunjan Shandilya, Vatsala Anand, Rahul Singh Chauhan, Hemant Singh Pokhariya, Sheifali Gupta · 2024
Worldwide, cervical cancer continues to be the primary cause of cancer-related deaths among women, underscoring the need of early and precise detection. While Pap smears are an essential tool for early diagnosis, the intricacy of cancer cell pictures frequently poses challenges to these labor-intensive and inconsistent assays. Although machine learning approaches for improved screening have been studied in previous publications, high accuracy with little computational loss is still unattainable. This study presents an innovative approach for categorizing cervical cancer into multiple classes. It involves utilizing a ResNet152V2 deep learning model that has been fine-tuned, together with transfer learning techniques. This study intends to improve the accuracy and efficiency of diagnosis by utilizing the comprehensive SipakMed dataset, which consists of 4049 pap smear images. The primary significance of this study resides in its pioneering utilization of transfer learning with ResNet152V2, resulting in an exceptional accuracy rate of 96.15% and an average precision of 97%. The results indicate that the model is highly effective in categorizing cervical cancer photos into five distinct classes with minimum loss. This highlights the potential of advanced computational tools to assist medical professionals in early detection and the creation of personalized treatment programs. This study addresses a significant deficiency by presenting a strong, effective, and extremely precise categorization technique, which will lead to enhanced clinical results in the screening of cervical cancer.