Application of electron beam irradiation in kiwifruit preservation and construction of a predictive model for its effectiveness

Abstract

The effects of electron beam irradiation on postharvest physiological and biochemical quality and electrical parameters of kiwifruit were investigated, and a machine learning model for its quality prediction was developed based on electrical properties. Kiwifruit was irradiated with 0 (control), 200 and 400 Gy of electron beam and stored in a refrigerator (0–1 °C, 90–95 % relative humidity). Results indicated that the 400 Gy irradiated group reduced decay rate and weight loss. Additionally, this group exhibited lower O₂^−·generation rate and H₂O₂ content, higher gene expression and activity of peroxidase, superoxide dismutase, and ascorbate peroxidase, and down-regulated lipoxygenase gene expression and activity compared to the control group. Microscopic observations revealed better-preserved and more intact cellular structures in the irradiated groups at the end of storage. A strong correlation was found between quality and electrical parameters, with equivalent parallel resistance (Rp) identified as a common key feature across all groups. The Random Forest model, built on these electrical characteristics, outperformed the linear model, improving the maximum R² from 0.5130 to 0.8830. Overall, 400 Gy electron beam irradiation effectively preserved kiwifruit quality, and the proposed electrical-property-based machine learning approach offers a promising non-destructive method for quality evaluation.