Bengaluru: Cervical cancer rarely appears suddenly. In many women, the disease is preceded by subtle cellular changes that can linger unnoticed for years before turning invasive. A Bengaluru researcher believes artificial intelligence may be able to spot those warning signs long before a tumour appears.Lalasa Mukku, a researcher at Christ (Deemed-to-be-University), has developed a series of AI-based models aimed at identifying women at elevated risk of cervical cancer by analysing precancerous changes known as Cervical Intraepithelial Neoplasia (CIN).Mukku, who is from the department of artificial intelligence and data science engineering of Christ, also holds patents related to an AI model designed to predict cancer risk as much as five years before tumour formation.Cervical cancer remains among the leading cancers affecting women worldwide, with the burden falling disproportionately on low- and middle-income countries, where access to specialists and screening remains limited. Early diagnosis is known to improve outcomes substantially.Mukku’s approach combines images captured during colposcopy examinations with patient clinical information. During these tests, doctors examine the cervix after applying saline, acetic acid and iodine solutions. Each stage highlights tissue differently, producing clues that can reveal precancerous changes.In a 2024 study published in the International Journal of Advances in Intelligent Informatics, Mukku 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. The researchers said the model was intended to support clinicians and improve screening.Another challenge lies in the images themselves. Bright reflections caused by moisture on the cervix often resemble the white lesions doctors look for, potentially confusing computer systems.To address this, Mukku developed a separate technique, detailed in a paper published in Multimedia Tools and Applications, that removes these reflections and accurately isolates the cervical region before analysis, thus, improving diagnosis.More recently, in a paper presented at an IEEE conference in 2025, she proposed a quantum convolutional neural network architecture for analysing medical images. Tested on publicly available cervical cancer screening datasets, the model reported an overall accuracy of about 98.6%.The technology remains at the research stage and would require extensive clinical validation before being used in hospitals. But if successful, it can help identify the risk while there is still time to prevent it.
