Algorithm mediated early detection of oral cancer from image analysis

Area of Science:

• Medical Imaging
• Computational Pathology
• Oncology

Background:

• Oral cancer poses a significant global health challenge, particularly in developing countries.
• Early detection of premalignant lesions is crucial for improving patient outcomes and reducing mortality rates.
• Current diagnostic methods can be subjective and may benefit from objective, automated tools.

Purpose of the Study:

• To develop an Automatic Oral Cancer Detection algorithm.
• To identify and differentiate premalignant lesions (erythroplakia, leukoplakia) from normal buccal cavity images.
• To facilitate early oral cancer detection and potentially reduce fatalities.

Main Methods:

• Collected and processed 60 oral cavity images (normal, erythroplakia, leukoplakia) using MATLAB.
• Employed image processing techniques including red value analysis (maximum and mean) and YCbCr color space.
• Utilized Gray-Level Co-occurrence Matrix (GLCM) based features, specifically entropy, for final classification.
• Randomly divided images into training and testing sets for algorithm validation.

Main Results:

• Achieved 100% efficiency in separating normal from abnormal images using R value distribution.
• Successfully differentiated abnormal images using Y plane and Cr plane segmentation based on mean R values.
• The GLCM entropy feature achieved 89% efficiency in differentiating premalignant lesions.

Conclusions:

• The developed algorithm effectively differentiates premalignant lesions from normal oral tissue.
• A user-friendly graphic interface was created to display detection outcomes with notable accuracy.
• This automated approach holds potential for early oral cancer diagnosis and intervention.