Pathologists and breeders assess Fusarium damaged kernels (FDKs) on spikes to evaluate FHB disease presence before harvest. However, the visible assessment alone cannot accurately quantify the disease severity. Additionally, the symptom on barley FDKs tend to fade as the grain matures. Rapid and accurate quantification of individual kernel DON levels, indicative of FDKs holds significant importance. Current detection methods, which rely on chromatography to analyze batch samples, are time-consuming and not realistic for a real-time measurement of individual kernels on spikes. Hyperspectral imaging (HSI) is capable of rapidly detecting toxins in kernels through infrared wave absorption at different wavelengths. The objective of this study is to establish a pilot-scale predictive model by combining HSI with machine learning data processing for identifying FDKs in malting barley. A total of 300 barley seeds is scanned using a HSI spectrum with a short-wave infrared camera (Specim, Oulu, Finland) at wavelengths of 1,000 - 2,500 nm. After preprocessing, each image is divided into 3×3 sub-pixel images using MATLAB 2023b (The MathWorks, Natick, Massachusetts). The scanned individual kernels are analyzed using LC-MS to quantify DON levels. Machine learning algorithms, such as Partial Least Squares Regression (PLSR), Multilayer Perceptron (MLP), Random Forest (RF) and Extreme Gradient Boosting (XGBoost), are applied to integrate LC-MS and hyperspectral data, establishing a comprehensive model for rapid and non-destructive analysis of kernel DON levels. A pilot-scale machine learning model is anticipated to predict the DON level in each kernel. Expected outcomes include a rapid, reliable, and non-destructive method for quantifying FDKs in a bulk barley sample. The HSI approach should provide accurate measurements for early detection of FDKs and mycotoxin contamination.