New Computational Model Achieves 98% Accuracy in Detecting Cervical Dysplasia

Led by Dr. Lipi B. Mahanta, the team at IASST focused on creating a model that excels in real-world applications, offering high accuracy while minimizing computation time.

A groundbreaking computational model developed by scientists from the Institute of Advanced Study in Science and Technology (IASST) promises to significantly improve the diagnosis of cervical dysplasia, a precursor to cervical cancer. This innovative model combines advanced machine learning (ML) techniques with precise pattern identification to enhance early detection of abnormal cell growth on the cervix.

Led by Dr. Lipi B. Mahanta, the team at IASST focused on creating a model that excels in real-world applications, offering high accuracy while minimizing computation time. The researchers experimented with various color models, transform techniques, feature representation schemes, and classification methods to devise a highly effective ML framework for detecting cervical dysplasia.

The model's robustness was validated through testing on two datasets: one from healthcare centers in India and another publicly available dataset. Utilizing the Non-subsampled Contourlet Transform (NSCT) and the YCbCr color model, the new system achieved an impressive average accuracy of 98.02%.

This advanced model represents a significant advancement in medical imaging and diagnostics, holding potential for widespread use in early cervical cancer detection and improved patient outcomes.