Cervical cancer detection using multimodal medical images
Madhura Malhari Kalbhor, Bal Singh Virdee, Ashish K. Khanna, Prathamesh Bachhav, Vedant Bijwe, Vaishnav Raundal, Sayali S. Pawar · 2025
A potentially deadly condition, cervical cancer affects millions of women globally. Survival rates are greatly increased by early discovery made possible by medical imaging, which allows for prompt intervention. By combining Pap smear (cytology) and colposcopy images, this study aims to develop a deep learning-based system that employs a multimodal approach to accurately diagnose cervical cancer. To efficiently analyse cervical pictures, the suggested system makes use of an attention network and a multimodal encoder. The Malhari dataset, which includes Pap smear and colposcopy images classified as normal and severe instances, is used to train the models. To improve classification performance, sophisticated feature extraction techniques are used. The multimodal model effectively differentiates between cases that are malignant and those that are not, exhibiting excellent classification accuracy. The suggested method’s dependability is confirmed by important performance indicators like accuracy, recall, precision, and F1-score. The method improves the accuracy of cervical cancer detection by combining various imaging modalities and utilising deep learning algorithms. The findings demonstrate how multimodal learning may enhance patient outcomes and early diagnosis.