A Novel Subspace Alignment-Based Interference Suppression Method for the Transfer Caused by Different Sample Carriers...

Area of Science:

• Biomedical Engineering
• Sensor Technology
• Computational Biology

Background:

• Medical electronic noses (e-noses) show promise for wound infection detection by analyzing bacterial metabolites.
• A significant challenge is the dramatic drop in prediction accuracy when models trained on lab data are applied to human clinical samples due to "transfer caused by different sample carriers".

Purpose of the Study:

• To propose and evaluate a novel subspace alignment-based interference suppression (SAIS) method to address the sample carrier interference problem in medical e-nose applications.
• To enhance the recognition accuracy of human wound infection detection using e-nose technology.

Main Methods:

• Development of a subspace alignment-based interference suppression (SAIS) method incorporating domain correction.
• The SAIS method involves extracting subspaces from different data domains and aligning them to mitigate distribution differences.
• Experimental validation using infected rat samples to assess the method's effectiveness compared to no interference suppression.

Main Results:

• The SAIS method significantly improved the recognition accuracy for infected rat samples.
• Accuracy increased from 29.18% without interference suppression to 82.55% with SAIS.
• Demonstrated the capability of SAIS to suppress interference arising from different sample carriers.

Conclusions:

• The proposed SAIS method is effective in overcoming the "transfer caused by different sample carriers" interference in medical e-nose systems.
• SAIS technology offers a promising solution for improving the reliability and accuracy of e-nose-based wound infection detection in clinical settings.