Leveraging AI technology for distinguishing Eucommiae Cortex processing levels and evaluating anti-fatigue potential

Area of Science:

• Traditional Chinese Medicine
• Pharmacognosy
• Computational Biology

Background:

• Eucommiae Cortex (ECO) is a valued medicinal plant requiring specific processing for efficacy.
• Processing methods significantly impact ECO's therapeutic effects, necessitating quality control.
• Current identification methods for ECO are inefficient and invasive.

Purpose of the Study:

• To develop an accurate, rapid, and non-invasive method for classifying processed ECO using image analysis.
• To evaluate the anti-fatigue properties of ECO processed at different levels.
• To correlate image-based classification with functional anti-fatigue efficacy.

Main Methods:

• Deep learning models, including ResNet and Vision Transformer (ViT), were trained on images of ECO at various processing stages.
• Anti-fatigue efficacy was assessed in mice via swimming endurance, pole climbing, and biochemical markers (SDH, LDH, ATP, Na+-K+-ATPase, Ca2+-Mg2+-ATPase).
• Image classification accuracy and anti-fatigue performance across different processing levels were analyzed.

Main Results:

• The Vision Transformer model achieved over 95% accuracy in classifying ECO images based on processing levels.
• Image analysis effectively provided an automated and accurate method for ECO quality assessment.
• Mice treated with moderately processed ECO showed superior anti-fatigue benefits compared to other groups.

Conclusions:

• Deep learning-based image analysis offers a promising, non-invasive tool for quality control of processed Eucommiae Cortex.
• Processing level is critical for optimizing the anti-fatigue properties of ECO.
• Further research can leverage AI for standardization of traditional herbal medicines.