Reflectance and Interactance Spectroscopy for Detecting Root Knot Nematode Galls in Harvested Sweetpotatoes

Authors

• Hoang M. Nguyen, North Carolina State University
• Adrienne M. Gorny, North Carolina State University
• Simon Fraher, North Carolina State University
• Ali Altaqui, North Carolina State University
• Jerome Maleski, North Carolina State University
• Cranos Williams, North Carolina State University
• Daniela S. Jones, North Carolina State University
• Yuzhen Lu, Michigan State University
• Mohammed Kamruzzaman, University of Illinois at Urbana-Champaign
• Nuwan K. Wijewardane, Mississippi State University
• Michael W. Kudenov, North Carolina State University

ORCID

Wijewardane: https://orcid.org/0000-0001-8962-9451

MSU Affiliation

College of Agriculture and Life Sciences; Department of Agricultural and Biological Engineering; James Worth Bagley College of Engineering

Creation Date

2026-06-30

Abstract

Root-knot nematodes (RKN) are a major pathogen in sweetpotato production, with current detection relying on labor-intensive, time-consuming laboratory assays. This study compared non-contact near-infrared reflectance and interactance imaging for rapid RKN gall detection, using three laser wavelengths (750, 808, and 905 nm) identified in prior work. Experimental imaging was paired with Monte Carlo scattering simulations to quantify spectral contrast between galled and healthy tissues. Additionally, six convolutional neural network (CNN) architectures were evaluated for automated classification, and a cost-based analysis was performed to assess performance trade-offs. Spectral analysis revealed consistent trends across measured and simulated data: galled tissues had lower 750/808 nm ratios and higher 905/808 nm ratios than healthy tissues, with interactance generally providing higher absolute contrast than reflectance. However, classification results showed that spatial features in reflectance images maintained similar accuracy to interactance data despite lower spectral contrast. The highest detection accuracy (97.5%) and lowest false positive rate (1.46%) were achieved using a combined (hybrid) reflectance–interactance metric with ResNet 101. Single-modality reflectance or interactance achieved ≈[jls-end-space/]95.7% accuracy. Cost modeling indicated that the optimal network–metric choice depends on the relative costs of false positives (field remediation) and false negatives (rejected deliveries). When remediation costs were higher, ResNet 50 with a hybrid metric performed best. However, when costs were equal or false negatives were more costly, ResNet 101 with a hybrid metric was optimal. Limitations include evaluation on a single sweetpotato variety, simplified cost modeling, and exclusion of other surface/subsurface defects.

Publication Date

5-9-2026

Publication Title

Postharvest Biology and Technology

Publisher

Elsevier

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.

Rights

© 2026 The Authors

Recommended Citation

Nguyen, H. M., Gorny, A. M., Fraher, S., Altaqui, A., Maleski, J., Williams, C., Jones, D. S., Lu, Y., Kamruzzaman, M., Wijewardane, N. K., & Kudenov, M. W. (2026). Reflectance and interactance spectroscopy for detecting root knot nematode galls in harvested sweetpotatoes. Postharvest Biology and Technology, 240, 114398. https://doi.org/10.1016/j.postharvbio.2026.114398