Enhancing Interpretability in Medical Image Classification by Integrating Formal Concept Analysis with Convolutional...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Convolutional Neural Networks (CNNs) are widely used for medical image classification.
• Interpreting CNN decision-making remains a significant challenge in clinical settings.
• Current visualization methods lack explicit links between high-level features and class labels across datasets.

Purpose of the Study:

• To enhance the interpretability of medical image classification.
• To bridge the gap between CNN feature learning and class labels using Formal Concept Analysis (FCA).
• To develop a transparent image classification model for clinical decision-making.

Main Methods:

• Integration of Formal Concept Analysis (FCA) with Convolutional Neural Networks (CNNs).
• Utilizing FCA as an image classification model to understand feature-label relationships.
• Evaluation on histopathological image datasets (Warwick-QU, BreakHIS) and comparison with deep neural classifiers.

Main Results:

• The FCA-based classifier achieved accuracy comparable to deep neural classifiers.
• The proposed method provides transparency into the classification process.
• Demonstrated effectiveness on complex histopathological image datasets.

Conclusions:

• Formal Concept Analysis (FCA) offers a transparent alternative for medical image classification.
• This approach enhances clinical decision-making by providing interpretable AI models.
• The FCA framework shows promise for advancing explainable AI in healthcare.