Deep Learning-Based Detection of Aflatoxin B1 Contamination in Almonds Using Hyperspectral Imaging: A Focus on...

Area of Science:

• Food Science
• Analytical Chemistry
• Computer Science

Background:

• Aflatoxin B1 is a toxic carcinogen found in food products like almonds, posing significant health risks.
• Rapid and non-destructive detection methods are essential for ensuring food safety and preventing the spread of aflatoxin B1 contamination.

Purpose of the Study:

• To develop and evaluate a novel deep learning approach for the accurate classification of aflatoxin B1-contaminated almonds using hyperspectral imaging.
• To compare the performance of the proposed deep learning model against traditional machine learning methods for aflatoxin B1 detection.

Main Methods:

• Utilized a 3D Inception-ResNet deep learning architecture, fine-tuned for classification tasks.
• Employed hyperspectral imaging to capture spectral data from almonds.
• Implemented a feature selection algorithm to optimize processing efficiency and reduce spectral dimensionality.

Main Results:

• The proposed Lightweight 3D Inception-ResNet model achieved a validation accuracy of 90.81%, an F1-score of 0.899, and an AUC of 0.964.
• The deep learning model significantly outperformed traditional methods like SVM, RF, QDA, and DT in classifying aflatoxin B1-contaminated almonds.
• The computationally efficient Lightweight 3D Inception-ResNet model is suitable for real-time industrial applications.

Conclusions:

• Hyperspectral imaging combined with deep learning provides a highly accurate method for detecting aflatoxin B1 in almonds.
• The developed deep learning approach supports the creation of real-time automated screening systems for enhanced food safety.
• This research contributes to reducing health risks associated with aflatoxin B1 contamination in the food supply chain.