A Vaginitis Classification Method Based on Multi-Spectral Image Feature Fusion

Area of Science:

• Medical Imaging
• Computational Pathology
• Gynecological Infections

Background:

• Vaginitis is a common female reproductive tract infection.
• Current diagnosis relies on manual microscopic examination, increasing workload.
• Existing computer-aided diagnosis using RGB images has limitations.

Purpose of the Study:

• To develop an improved vaginitis classification algorithm using multi-spectral imaging.
• To overcome the limitations of traditional RGB image-based diagnosis.
• To enhance the accuracy and efficiency of vaginitis diagnosis.

Main Methods:

• Utilized multi-spectral imaging technology for data acquisition.
• Developed a feature layer fusion algorithm for vaginitis classification.
• Compared multi-spectral approach against traditional RGB image analysis.

Main Results:

• Achieved an 11.39% improvement in classification accuracy.
• Increased precision by 15.82% and recall by 27.25% compared to RGB.
• Demonstrated distinctive spectral influences and the benefit of subdivided spectral images.

Conclusions:

• Multi-spectral imaging offers superior performance for vaginitis diagnosis.
• The proposed algorithm enhances diagnostic accuracy and efficiency.
• Subdivided spectral analysis is crucial for effective vaginitis image analysis.