Morphology guided hybrid deep learning for species identification of root knot nematodes

Root-knot nematodes ( Meloidogyne spp.) are a major global constraint on crop productivity. The second-stage juvenile (J2) is the infective stage of nematode that actively seeks out and infects plant roots; thus, accurate identification at this stage is important for effective nematode management and control. However, current identification practices remain challenging and is time-consuming, due to their reliance on expert-dependent microscopic examinations. This challenge creates significant barriers to scalability and rapid decision-making. To address this, we propose a morphology guided hybrid deep learning framework for identifying Meloidogyne graminicola and M. incognita at the J2 stage. The proposed framework integrates biologically relevant shape descriptors with deep visual representations to improve species discrimination. Standardized morphological features are fused with convolutional embeddings via a modified MobileNetV2 classifier head, to enable joint and discriminative representation learning. The proposed approach was evaluated against morphology-based machine learning models and end-to-end deep learning baselines. The results demonstrate that the proposed morphology guided deep learning improves classification performance and interpretability. It provides a lightweight and biologically informed solution for automated identification of J2 root-knot nematode species.