Transfer Learning Assisted Cervical Cancer Categorization from Pap Smear Images Via the Multihead Attention Technique

Cervical cancer, characterized by the abnormal proliferation of cervical cells in the cervix, remains a significant global health concern, with approximately 660,000 new cases diagnosed annually in women worldwide. The Pap test serves as a vital clinical tool for diagnosing cervical cancer with precision. This study introduces a custom transfer learning model designed to classify cervical cancer efficiently from Pap smear images. Existing pre-trained architectures, including MobileNetV2, AlexNet, DenseNet121, and VGG-19, have been utilized for this task. However, these models often struggle to effectively capture critical pixel-level features in the images. To address this limitation, we propose a lightweight model combining Convolutional Neural Networks (CNNs) with Multi-Head Attention mechanisms. We used many types of transfer learning architectures, including MobileNetV2, ResNet101V2, ResNet152V2, and InceptionV3, with the following layers: GlobalAveragepooling2D,MultiHead Attention Block, Droput, and Dense layer. This configuration reduces computational complexity while improving the model’s ability to identify key image features. Our method achieves a classification accuracy of 96.23%, significantly outperforming existing algorithms in both accuracy and computational efficiency.