Rapid Mold Detection in Chinese Herbal Medicine Using Enhanced Deep Learning Technology

Area of Science:

• Food Science
• Analytical Chemistry
• Traditional Chinese Medicine

Background:

• Mold contamination is a major issue affecting Chinese herbal medicine (CHM) quality and efficacy.
• Accurate and rapid detection of mold in CHM is crucial for quality control.

Purpose of the Study:

• To develop a fast and precise method for detecting mold in CHM, specifically *Atractylodes macrocephala*.
• To utilize electronic nose (e-nose) technology combined with advanced deep learning models for mold identification.

Main Methods:

• Employing stochastic resonance (SR) to denoise e-nose data.
• Developing an eccentric temporal convolutional network (ETCN) model to analyze sensor data.
• Validating the SR-enhanced ETCN (SR-ETCN) model using data from eight sensors.

Main Results:

• The SR-ETCN method achieved a high accuracy of 94.3% for mold detection in CHM.
• The proposed model demonstrated superior performance compared to seven other models.
• The method effectively captured temporal and spatial information from e-nose signals.

Conclusions:

• The SR-ETCN method offers a reliable and efficient solution for detecting mold in Chinese herbal medicines.
• This technology can significantly expedite the quality assessment of CHM.
• Ensuring the safety and efficacy of traditional medicinal practices through improved quality control.