Abstract
Biological control programs for fruit flies rely on the mass rearing and release of parasitoids to suppress pest populations. The success of this strategy depends on effective quality control, including the assessment of parasitism and insect survival. Conventional methods, such as dissection, Polymerase Chain Reaction (PCR), and adult emergence, are destructive, labor-intensive, and time-consuming. Hyperspectral imaging offers a rapid and non-destructive alternative for evaluating parasitism. Therefore, this study aimed to discriminate Anastrepha fraterculus (Wiedemann, 1830) (Diptera: Tephritidae) pupae parasitized by Diachasmimorpha longicaudata (Ashmead, 1905) (Hymenoptera: Braconidae) using hyperspectral imaging and identifying the most informative spectral ranges and wavelengths for this purpose. Hyperspectral images were acquired from pupae at different developmental stages expressed in degree days (DD), preserving insect viability. Principal component analysis (PCA) and clustering analysis indicated a tendency to separate parasitized from non-parasitized pupae. Spectral profiles showed lower reflectance in the visible region and higher reflectance in the near-infrared (NIR, > 700 nm) for parasitized pupae. Greater spectral variability was also observed throughout development, particularly at 7 and 12 days (135 and 189 DD, respectively), compared with non-parasitized pupae at the same chronological ages (260.1 and 321.3 DD, respectively), which exhibited a more homogeneous spectral pattern. The Random Forest model showed moderate classification performance across all developmental stages, with higher predictive accuracy at later stages. These findings demonstrate the potential of hyperspectral imaging as a non-destructive tool for detecting parasitism and improving quality control in mass-rearing programs for biological control.