Hyperspectral Imaging Coupled with Deep Learning Frameworks for Mycotoxin Detection in Wheat and Maize: Progress, Challenges, and Research Trends (2023-2025)

Mycotoxins, particularly in wheat and maize, pose serious global health and economic risks. Conventional detection techniques such as high-performance liquid chromatography (HPLC) and liquid chromatography mass spectrometry (LCMS) are even-though accurate but invasive, time-consuming, and costly. Hyperspectral imaging (HSI), coupled with deep learning (DL), offers a rapid, nondestructive alternative for detecting mycotoxin contamination. This review evaluates recent developments (2023-2025) in HSI-based detection frameworks, focusing on DL models including CNNs, YOLO, DeepMAC, SqueezeNet, and DenseNet. It explores advancements in spectral analysis, optimal wavelength identification, sample orientation effects, preprocessing strategies, and feature selection techniques. Key performance metrics across multiple studies are analyzed. Notably, the review highlights critical limitations such as dataset imbalance, overfitting risks, and the lack of transparency in reporting sample sizes and augmentation methods. It emphasizes the need for standardized HSI workflows and points to future directions, including portable sensor development, synthetic data generation, and integration of multimodal sensing with AI for enhanced detection accuracy.