Improved two-view interactional fuzzy learning based on mutual-rectification and knowledge-mergence.

Area of Science:

• Medical imaging analysis
• Artificial intelligence in oncology
• Fuzzy logic systems

Background:

• Nasopharyngeal carcinoma (NPC) is challenging to diagnose due to its hidden location and varied symptoms.
• Current imaging methods struggle to fully delineate NPC's complex anatomical relationships.
• Existing fuzzy classifiers for NPC have limitations in rule reusability.

Purpose of the Study:

• To propose a novel two-view mutual rectification and knowledge mergence Takagi-Sugeno-Kang fuzzy classifier (TVRM-TFC).
• To address challenges in using imaging for fine-tuning organ tissue analysis in NPC diagnosis.
• To improve decision-making accuracy by integrating complementary information from multiple imaging modalities.

Main Methods:

• Utilizing Kullback-Leibler divergence (KLIC) for feature selection from imaging data.
• Employing interpretable zero-order Takagi-Sugeno-Kang (TSK) fuzzy classifiers as basic training units.
• Implementing a two-view mutual rectification and knowledge mergence strategy to integrate information from different imaging perspectives.

Main Results:

• The proposed TVRM-TFC achieved satisfactory accuracies and linguistic interpretability.
• The method effectively fine-tuned information for decision-making between different imaging means.
• Comparative analysis on CT and MRI data demonstrated the classifier's merits.

Conclusions:

• The TVRM-TFC successfully merges decision-making knowledge from different imaging views.
• This integration compensates for information gaps and optimizes diagnostic decisions for NPC.
• The study highlights the potential of advanced fuzzy logic classifiers in improving NPC detection.