bangalore: Cervical cancer rarely develops suddenly. In many women, the disease is preceded by subtle cellular changes that can remain unnoticed for years before becoming invasive. Researchers in Bengaluru believe that artificial intelligence may be able to spot these warning signs long before tumors appear.Christ (Deemed University) researcher Lalasa Mukku has developed a series of AI-based models aimed at identifying women at high risk of cervical cancer by analyzing precancerous changes known as cervical intraepithelial neoplasia (CIN).Mr Mook, from Christ’s School of Artificial Intelligence and Data Science Engineering, also holds a patent on an AI model designed to predict cancer risk as much as five years before tumor formation.Cervical cancer remains one of the leading cancers affecting women worldwide, with the burden falling disproportionately on low- and middle-income countries where access to specialists and testing remains limited. Early diagnosis is known to significantly improve outcomes.Mukku’s approach combines images taken during a colposcopy exam with the patient’s clinical information. During these tests, your doctor examines your cervix after applying a saline, acetic acid, and iodine solution. Each stage highlights tissue in a different way, providing clues that reveal precancerous changes.In a 2024 study published in the International Journal of Advances in Intelligent Informatics, Mook developed a model called CMT-CNN that combines these sequential images with clinical data. The system achieved a classification accuracy of 92.3% in identifying CIN. Researchers said the model aims to support clinicians and improve screening.Another challenge lies in the images themselves. Bright reflections caused by moisture in the cervix look a lot like the white lesions doctors look for, and can confuse computer systems.To address this, Mukku developed another technique detailed in a paper published in Multimedia Tools and Applications. This technique eliminates these reflections and accurately isolates the cervical region before analysis, improving diagnosis.Most recently, in a paper presented at the 2025 IEEE Conference, she proposed a quantum convolutional neural network architecture for analyzing medical images. The model, tested on a publicly available cervical cancer screening dataset, reported an overall accuracy of approximately 98.6%.The technology is still in the research phase and will require extensive clinical validation before it can be used in hospitals. However, if successful, it can help identify risks while there is still time to prevent them.
