An Interpretable Deep Learning Model for Covid-19 Detection With Chest X-Ray Images

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computer Science

Background:

• Accurate epidemic detection is crucial for public health.
• Deep learning models offer potential for disease detection.
• Interpretability of deep learning in healthcare is essential for trust and adoption.

Purpose of the Study:

• Introduce Gen-ProtoPNet, a novel interpretable deep learning model.
• Enhance transparency in deep learning for medical image classification.
• Improve disease detection accuracy while maintaining model interpretability.

Main Methods:

• Developed Gen-ProtoPNet, an extension of ProtoPNet and NP-ProtoPNet.
• Utilized a generalized distance function to support diverse prototype dimensions (square and rectangular).
• Evaluated model performance on a dataset of X-ray images for classification tasks.

Main Results:

• Gen-ProtoPNet achieved an accuracy of 87.27% in classifying three image classes.
• The model's performance is comparable to state-of-the-art non-interpretable deep learning models.
• Achieved accuracy close to the non-interpretable model's 88.42%.

Conclusions:

• Gen-ProtoPNet offers a transparent and accurate deep learning solution for medical image analysis.
• The generalized distance function allows for flexible prototype utilization.
• This interpretable model advances the application of AI in disease detection.